Choosing Wisely: Things We Do For No Reason

The “Things We Do for No Reason” series reviews practices which have become common parts of hospital care, but which may provide little value to our patients. Practices reviewed in the TWDFNR series do not represent “black and white” conclusions or clinical practice standards, but are meant as a starting place for research and active discussions among hospitalists and patients. We invite you to be part of that discussion. https://www.choosingwisely.org/

Bedside thoracentesis can cause serious complications, such as pneumothorax, re-expansion pulmonary edema, or hemorrhage. These rare complications have led many hospitalists to routinely order chest radiographs (CXRs) following thoracentesis. However, post-thoracentesis CXRs are usually not indicated and can lead to unnecessary radiation exposure and expense. Rather than obtaining routine CXRs, hospitalists should use postprocedural signs and symptoms to identify the occasional patients who require imaging. A risk-stratified approach is a safe and cost-effective way to avoid unnecessary radiographs.

A 52-year-old man with decompensated liver disease and hepatic hydrothorax is hospitalized for increasing dyspnea caused by a recurrent pleural effusion. Diuretics do not improve his dyspnea, and his hospitalist recommends a therapeutic thoracentesis for symptom relief. The patient does not have any significant procedural risk factors: He does not have preexisting pulmonary or pleural disease, his platelet count is 105,000 × 10³/µl, and his international normalized ratio is 1.3. Bedside sonography demonstrates a large, free-flowing, right-sided pleural effusion. The hospitalist performs an uncomplicated ultrasound-guided removal of 1.5 L of straw-colored fluid with a catheter-over-needle kit. The patient does not have any pain or increased shortness of breath during or after the procedure. The hospitalist reflexively orders a routine chest radiograph to assess for pneumothorax.

Pleural effusions are newly diagnosed in more than 1.5 million Americans annually,¹ and hospitalists frequently care for patients requiring thoracentesis. Internal medicine residents traditionally learn to perform this procedure during residency, and thoracentesis remains a common task for both residents and hospitalists.² Patients typically tolerate thoracentesis well, but they can develop serious complications such as pneumothorax, re-expansion pulmonary edema, or hemothorax. Before the advent of bedside ultrasound, these complications occurred relatively commonly; a 2010 systematic review, for example, found that the rate of pneumothorax from thoracentesis performed without ultrasound was 9.3%.³ Other studies have identified even higher rates of complications, including two case series in which investigators found a 14% rate of major complications⁴ and a pneumothorax rate of nearly 30%.⁵ Postprocedure radiographs became common practice because of the high rate of complications, and this practice has persisted for many practitioners despite the substantial safety improvements introduced by bedside ultrasonography.⁶

Hospitalists might think routine CXRs are helpful after ultrasound-guided thoracentesis for additional reasons. First, modern guidelines reflecting the low risk of complications after ultrasound-guided procedures have not been released by United States pulmonary medicine societies, and some clinicians may continue to follow practices acquired during the era of unguided thoracentesis. Second, performing postprocedure imaging has become ingrained as a standard part of some institutional procedure checklists⁶ and some prominent textbooks continue to recommend the practice.⁷ For some hospitalists, this testing reflex may be reinforced by other common procedures, such as placing a nasogastric tube or a central venous catheter, for which a postprocedure CXR is standard practice. Thus, ordering postprocedure imaging can become internalized as the safe, checklist-based final step of a procedure. Third, hospitalists may order a postprocedure CXR for reasons other than detecting procedural complications. The pleural effusion might be thought to obscure a parenchymal or endobronchial lesion for which a postprocedure CXR may reveal an important finding. Finally, a CXR also may also satisfy the clinician’s curiosity regarding the completeness of drainage.

A routine post-thoracentesis CXR is not necessary for three reasons. First, the use of ultrasound marking or guidance has substantially improved site selection and reduced the rate of complications for experienced operators. For example, a 2010 systematic review found an overall rate of pneumothorax of 4% for ultrasound-guided procedures performed between 1986 and 2006,³ whereas more recently published data suggest the current rate of pneumothorax is closer to 1% when ultrasound marking or guidance is used.^8,9 One study of 462 consecutive patients with malignant pleural effusions, for example, showed that the rate of pneumothorax with ultrasound-guided needle-over-catheter thoracentesis was 0.97% (3/310 patients), compared with a rate of 8.89% (12/135 patients) when the procedure was performed without ultrasound.⁹ Another prospective, randomized study of 160 patients with various causes of pleural effusion showed that the rate of pneumothorax with ultrasound-marked thoracentesis was 1.25% (1/80 patients), compared with 12.5% (10/80 patients) for procedures performed without ultrasound.⁸ Hospitalists who competently use ultrasound guidance should act on modern estimates of complications and may also choose to incorporate postprocedure ultrasound into their practice. Indeed, the Society of Hospital Medicine recommends against routine chest radiography in asymptomatic patients when sliding lung is visualized on postprocedure ultrasound.¹⁰

Second, procedural factors and postprocedural symptoms (new chest pain, dyspnea, or persistent cough) reliably identify patients with high risk of clinically meaningful complications. On one hand, only 1% to 2% of asymptomatic patients have a postprocedure pneumothorax, and clinical monitoring does not lead to chest tube placement in almost all of these cases.¹¹ On the other hand, 67% to 72% of symptomatic patients are found to have complications.¹² Doyle et al¹³ showed that the use of symptoms and procedure-specific factors (such as the aspiration of air, difficult procedure, multiple needle passes, or high operator suspicion of pneumothorax) could obviate the need for routine CXRs in approximately 60% of their procedures without any serious consequences.

Third, postprocedural CXRs very rarely reveal new or unexpected findings. For example, in one series,¹² only 3.8% of postdrainage radiographs uncovered new findings, none of which clarified the underlying diagnosis or changed management. To assess the utility of an initial thoracentesis and decide about repeat procedures, begin by asking the patient about symptoms and perform a physical exam.

CXRs might be helpful in certain scenarios, even when a complication is not suspected. For example, a postprocedure CXR to detect nonexpandable lung or evaluate the rate of recurrence may guide definitive management of patients with recurrent or malignant pleural effusion. Determining completeness of drainage may also assist with planning for palliative measures such as pleurodesis or indwelling pleural catheter placement. A postprocedure CXR is also helpful in patients with a technically difficult procedure or in those with symptoms during or immediately after the procedure. This recommendation is consistent with the 2010 British Thoracic Society guidelines, which recommend CXRs for procedures where air was withdrawn, the procedure was difficult, multiple needle passes were required, or the patient became symptomatic.¹⁴ The Society of Hospital Medicine’s recent Position Statement concurs with these guidelines and recommends against routine chest radiography in asymptomatic patients when sliding lung is visualized by postprocedure ultrasound.¹⁰

Hospitalists should not rountinely obtain post-thoracentesis CXRs in asymptomatic patients. Clinical monitoring with subsequent symptom-guided evaluation lowers costs, avoids unnecessary radiation exposure, and has been shown to be successful in a large case series of more than 9,300 patients.¹⁵ Some coughing should be expected with all large-volume thoracenteses as a normal response to re-expansion of atelectatic lung. The coughing should not persist past the immediate postprocedure period. If symptoms arise or if a complication is expected, the test of choice is either CXR or, if the hospitalist is a competent sonographer, bedside sonography. Bedside sonography is a low-cost, noninvasive method and has been well studied in the diagnosis of post-thoracentesis pneumothorax.¹⁶ CXRs may still be needed to confirm findings by sonography, to investigate postprocedural symptoms in those with pleural adhesions or other lung/pleural diseases (because ultrasonography is less reliable in these patients), or if reexpansion pulmonary edema or other complications are suspected. A robust quality improvement strategy to reduce unnecessary post-thoracentesis CXRs could result in cost savings and spare patients from radiation exposure, because a recent study of almost 1,000 thoracenteses performed at an academic medical center demonstrated that internal medicine residents, pulmonologists, and interventional radiologists order a CXR following 95% of thoracenteses.¹⁷ For a hypothetical hospital that orders 100 unnecessary post-thoracentesis CXRs annually, hospitalists could avoid approximately $7,000 in wasted expense per year.¹⁸

• Do not routinely order post-thoracentesis CXRs.
• Order a post-thoracentesis CXR if (1) the patient had new chest pain, dyspnea, or persistent cough during or after the procedure; (2) procedural features suggest increased risk of a complication (multiple needle passes, aspiration of air, difficulty obtaining fluid); or (3) a definitive palliative procedure will be arranged based on lung expansion.
• If qualified, use bedside sonography as a first step in the diagnosis of pneumothorax, reserving CXRs for those patients in whom accurate sonography is not possible, an alternative diagnosis is suspected, or when sonography findings are equivocal.

Following the uncomplicated thoracentesis, the hospitalist reconsidered the initial decision to order a CXR and rapidly assessed the patient’s risk of complications. Because the procedure required only one needle pass, air was not aspirated, and the patient did not experience prolonged coughing or pain, the CXR order was canceled. The patient recovered uneventfully and was spared the cost and radiation associated with the proposed CXR.

Do you think this is a low-value practice? Is this truly a “Thing We Do for No Reason?” Share what you do in your practice and join in the conversation online by retweeting it on Twitter (#TWDFNR) and liking it on Facebook. We invite you to propose ideas for other “Things We Do for No Reason” topics by emailing TWDFNR@hospitalmedicine.org.cknowledgments

Acknowledgements

The authors would like to thank Patricia Kritek and Somnath Mookherjee for their comments on an early version of this manuscript.

Disclosures

The authors have nothing to disclose.

The “Things We Do for No Reason” series reviews practices which have become common parts of hospital care, but which may provide little value to our patients. Practices reviewed in the TWDFNR series do not represent “black and white” conclusions or clinical practice standards, but are meant as a starting place for research and active discussions among hospitalists and patients. We invite you to be part of that discussion. https://www.choosingwisely.org/

Bedside thoracentesis can cause serious complications, such as pneumothorax, re-expansion pulmonary edema, or hemorrhage. These rare complications have led many hospitalists to routinely order chest radiographs (CXRs) following thoracentesis. However, post-thoracentesis CXRs are usually not indicated and can lead to unnecessary radiation exposure and expense. Rather than obtaining routine CXRs, hospitalists should use postprocedural signs and symptoms to identify the occasional patients who require imaging. A risk-stratified approach is a safe and cost-effective way to avoid unnecessary radiographs.

A 52-year-old man with decompensated liver disease and hepatic hydrothorax is hospitalized for increasing dyspnea caused by a recurrent pleural effusion. Diuretics do not improve his dyspnea, and his hospitalist recommends a therapeutic thoracentesis for symptom relief. The patient does not have any significant procedural risk factors: He does not have preexisting pulmonary or pleural disease, his platelet count is 105,000 × 10³/µl, and his international normalized ratio is 1.3. Bedside sonography demonstrates a large, free-flowing, right-sided pleural effusion. The hospitalist performs an uncomplicated ultrasound-guided removal of 1.5 L of straw-colored fluid with a catheter-over-needle kit. The patient does not have any pain or increased shortness of breath during or after the procedure. The hospitalist reflexively orders a routine chest radiograph to assess for pneumothorax.

Pleural effusions are newly diagnosed in more than 1.5 million Americans annually,¹ and hospitalists frequently care for patients requiring thoracentesis. Internal medicine residents traditionally learn to perform this procedure during residency, and thoracentesis remains a common task for both residents and hospitalists.² Patients typically tolerate thoracentesis well, but they can develop serious complications such as pneumothorax, re-expansion pulmonary edema, or hemothorax. Before the advent of bedside ultrasound, these complications occurred relatively commonly; a 2010 systematic review, for example, found that the rate of pneumothorax from thoracentesis performed without ultrasound was 9.3%.³ Other studies have identified even higher rates of complications, including two case series in which investigators found a 14% rate of major complications⁴ and a pneumothorax rate of nearly 30%.⁵ Postprocedure radiographs became common practice because of the high rate of complications, and this practice has persisted for many practitioners despite the substantial safety improvements introduced by bedside ultrasonography.⁶

Hospitalists might think routine CXRs are helpful after ultrasound-guided thoracentesis for additional reasons. First, modern guidelines reflecting the low risk of complications after ultrasound-guided procedures have not been released by United States pulmonary medicine societies, and some clinicians may continue to follow practices acquired during the era of unguided thoracentesis. Second, performing postprocedure imaging has become ingrained as a standard part of some institutional procedure checklists⁶ and some prominent textbooks continue to recommend the practice.⁷ For some hospitalists, this testing reflex may be reinforced by other common procedures, such as placing a nasogastric tube or a central venous catheter, for which a postprocedure CXR is standard practice. Thus, ordering postprocedure imaging can become internalized as the safe, checklist-based final step of a procedure. Third, hospitalists may order a postprocedure CXR for reasons other than detecting procedural complications. The pleural effusion might be thought to obscure a parenchymal or endobronchial lesion for which a postprocedure CXR may reveal an important finding. Finally, a CXR also may also satisfy the clinician’s curiosity regarding the completeness of drainage.

A routine post-thoracentesis CXR is not necessary for three reasons. First, the use of ultrasound marking or guidance has substantially improved site selection and reduced the rate of complications for experienced operators. For example, a 2010 systematic review found an overall rate of pneumothorax of 4% for ultrasound-guided procedures performed between 1986 and 2006,³ whereas more recently published data suggest the current rate of pneumothorax is closer to 1% when ultrasound marking or guidance is used.^8,9 One study of 462 consecutive patients with malignant pleural effusions, for example, showed that the rate of pneumothorax with ultrasound-guided needle-over-catheter thoracentesis was 0.97% (3/310 patients), compared with a rate of 8.89% (12/135 patients) when the procedure was performed without ultrasound.⁹ Another prospective, randomized study of 160 patients with various causes of pleural effusion showed that the rate of pneumothorax with ultrasound-marked thoracentesis was 1.25% (1/80 patients), compared with 12.5% (10/80 patients) for procedures performed without ultrasound.⁸ Hospitalists who competently use ultrasound guidance should act on modern estimates of complications and may also choose to incorporate postprocedure ultrasound into their practice. Indeed, the Society of Hospital Medicine recommends against routine chest radiography in asymptomatic patients when sliding lung is visualized on postprocedure ultrasound.¹⁰

Second, procedural factors and postprocedural symptoms (new chest pain, dyspnea, or persistent cough) reliably identify patients with high risk of clinically meaningful complications. On one hand, only 1% to 2% of asymptomatic patients have a postprocedure pneumothorax, and clinical monitoring does not lead to chest tube placement in almost all of these cases.¹¹ On the other hand, 67% to 72% of symptomatic patients are found to have complications.¹² Doyle et al¹³ showed that the use of symptoms and procedure-specific factors (such as the aspiration of air, difficult procedure, multiple needle passes, or high operator suspicion of pneumothorax) could obviate the need for routine CXRs in approximately 60% of their procedures without any serious consequences.

Third, postprocedural CXRs very rarely reveal new or unexpected findings. For example, in one series,¹² only 3.8% of postdrainage radiographs uncovered new findings, none of which clarified the underlying diagnosis or changed management. To assess the utility of an initial thoracentesis and decide about repeat procedures, begin by asking the patient about symptoms and perform a physical exam.

CXRs might be helpful in certain scenarios, even when a complication is not suspected. For example, a postprocedure CXR to detect nonexpandable lung or evaluate the rate of recurrence may guide definitive management of patients with recurrent or malignant pleural effusion. Determining completeness of drainage may also assist with planning for palliative measures such as pleurodesis or indwelling pleural catheter placement. A postprocedure CXR is also helpful in patients with a technically difficult procedure or in those with symptoms during or immediately after the procedure. This recommendation is consistent with the 2010 British Thoracic Society guidelines, which recommend CXRs for procedures where air was withdrawn, the procedure was difficult, multiple needle passes were required, or the patient became symptomatic.¹⁴ The Society of Hospital Medicine’s recent Position Statement concurs with these guidelines and recommends against routine chest radiography in asymptomatic patients when sliding lung is visualized by postprocedure ultrasound.¹⁰

Hospitalists should not rountinely obtain post-thoracentesis CXRs in asymptomatic patients. Clinical monitoring with subsequent symptom-guided evaluation lowers costs, avoids unnecessary radiation exposure, and has been shown to be successful in a large case series of more than 9,300 patients.¹⁵ Some coughing should be expected with all large-volume thoracenteses as a normal response to re-expansion of atelectatic lung. The coughing should not persist past the immediate postprocedure period. If symptoms arise or if a complication is expected, the test of choice is either CXR or, if the hospitalist is a competent sonographer, bedside sonography. Bedside sonography is a low-cost, noninvasive method and has been well studied in the diagnosis of post-thoracentesis pneumothorax.¹⁶ CXRs may still be needed to confirm findings by sonography, to investigate postprocedural symptoms in those with pleural adhesions or other lung/pleural diseases (because ultrasonography is less reliable in these patients), or if reexpansion pulmonary edema or other complications are suspected. A robust quality improvement strategy to reduce unnecessary post-thoracentesis CXRs could result in cost savings and spare patients from radiation exposure, because a recent study of almost 1,000 thoracenteses performed at an academic medical center demonstrated that internal medicine residents, pulmonologists, and interventional radiologists order a CXR following 95% of thoracenteses.¹⁷ For a hypothetical hospital that orders 100 unnecessary post-thoracentesis CXRs annually, hospitalists could avoid approximately $7,000 in wasted expense per year.¹⁸

• Do not routinely order post-thoracentesis CXRs.
• Order a post-thoracentesis CXR if (1) the patient had new chest pain, dyspnea, or persistent cough during or after the procedure; (2) procedural features suggest increased risk of a complication (multiple needle passes, aspiration of air, difficulty obtaining fluid); or (3) a definitive palliative procedure will be arranged based on lung expansion.
• If qualified, use bedside sonography as a first step in the diagnosis of pneumothorax, reserving CXRs for those patients in whom accurate sonography is not possible, an alternative diagnosis is suspected, or when sonography findings are equivocal.

Following the uncomplicated thoracentesis, the hospitalist reconsidered the initial decision to order a CXR and rapidly assessed the patient’s risk of complications. Because the procedure required only one needle pass, air was not aspirated, and the patient did not experience prolonged coughing or pain, the CXR order was canceled. The patient recovered uneventfully and was spared the cost and radiation associated with the proposed CXR.

Do you think this is a low-value practice? Is this truly a “Thing We Do for No Reason?” Share what you do in your practice and join in the conversation online by retweeting it on Twitter (#TWDFNR) and liking it on Facebook. We invite you to propose ideas for other “Things We Do for No Reason” topics by emailing TWDFNR@hospitalmedicine.org.cknowledgments

Acknowledgements

The authors would like to thank Patricia Kritek and Somnath Mookherjee for their comments on an early version of this manuscript.

Disclosures

The authors have nothing to disclose.

The “Things We Do for No Reason” series reviews practices which have become common parts of hospital care, but which may provide little value to our patients. Practices reviewed in the TWDFNR series do not represent “black and white” conclusions or clinical practice standards, but are meant as a starting place for research and active discussions among hospitalists and patients. We invite you to be part of that discussion. https://www.choosingwisely.org/

Bedside thoracentesis can cause serious complications, such as pneumothorax, re-expansion pulmonary edema, or hemorrhage. These rare complications have led many hospitalists to routinely order chest radiographs (CXRs) following thoracentesis. However, post-thoracentesis CXRs are usually not indicated and can lead to unnecessary radiation exposure and expense. Rather than obtaining routine CXRs, hospitalists should use postprocedural signs and symptoms to identify the occasional patients who require imaging. A risk-stratified approach is a safe and cost-effective way to avoid unnecessary radiographs.

A 52-year-old man with decompensated liver disease and hepatic hydrothorax is hospitalized for increasing dyspnea caused by a recurrent pleural effusion. Diuretics do not improve his dyspnea, and his hospitalist recommends a therapeutic thoracentesis for symptom relief. The patient does not have any significant procedural risk factors: He does not have preexisting pulmonary or pleural disease, his platelet count is 105,000 × 10³/µl, and his international normalized ratio is 1.3. Bedside sonography demonstrates a large, free-flowing, right-sided pleural effusion. The hospitalist performs an uncomplicated ultrasound-guided removal of 1.5 L of straw-colored fluid with a catheter-over-needle kit. The patient does not have any pain or increased shortness of breath during or after the procedure. The hospitalist reflexively orders a routine chest radiograph to assess for pneumothorax.

Pleural effusions are newly diagnosed in more than 1.5 million Americans annually,¹ and hospitalists frequently care for patients requiring thoracentesis. Internal medicine residents traditionally learn to perform this procedure during residency, and thoracentesis remains a common task for both residents and hospitalists.² Patients typically tolerate thoracentesis well, but they can develop serious complications such as pneumothorax, re-expansion pulmonary edema, or hemothorax. Before the advent of bedside ultrasound, these complications occurred relatively commonly; a 2010 systematic review, for example, found that the rate of pneumothorax from thoracentesis performed without ultrasound was 9.3%.³ Other studies have identified even higher rates of complications, including two case series in which investigators found a 14% rate of major complications⁴ and a pneumothorax rate of nearly 30%.⁵ Postprocedure radiographs became common practice because of the high rate of complications, and this practice has persisted for many practitioners despite the substantial safety improvements introduced by bedside ultrasonography.⁶

Hospitalists might think routine CXRs are helpful after ultrasound-guided thoracentesis for additional reasons. First, modern guidelines reflecting the low risk of complications after ultrasound-guided procedures have not been released by United States pulmonary medicine societies, and some clinicians may continue to follow practices acquired during the era of unguided thoracentesis. Second, performing postprocedure imaging has become ingrained as a standard part of some institutional procedure checklists⁶ and some prominent textbooks continue to recommend the practice.⁷ For some hospitalists, this testing reflex may be reinforced by other common procedures, such as placing a nasogastric tube or a central venous catheter, for which a postprocedure CXR is standard practice. Thus, ordering postprocedure imaging can become internalized as the safe, checklist-based final step of a procedure. Third, hospitalists may order a postprocedure CXR for reasons other than detecting procedural complications. The pleural effusion might be thought to obscure a parenchymal or endobronchial lesion for which a postprocedure CXR may reveal an important finding. Finally, a CXR also may also satisfy the clinician’s curiosity regarding the completeness of drainage.

A routine post-thoracentesis CXR is not necessary for three reasons. First, the use of ultrasound marking or guidance has substantially improved site selection and reduced the rate of complications for experienced operators. For example, a 2010 systematic review found an overall rate of pneumothorax of 4% for ultrasound-guided procedures performed between 1986 and 2006,³ whereas more recently published data suggest the current rate of pneumothorax is closer to 1% when ultrasound marking or guidance is used.^8,9 One study of 462 consecutive patients with malignant pleural effusions, for example, showed that the rate of pneumothorax with ultrasound-guided needle-over-catheter thoracentesis was 0.97% (3/310 patients), compared with a rate of 8.89% (12/135 patients) when the procedure was performed without ultrasound.⁹ Another prospective, randomized study of 160 patients with various causes of pleural effusion showed that the rate of pneumothorax with ultrasound-marked thoracentesis was 1.25% (1/80 patients), compared with 12.5% (10/80 patients) for procedures performed without ultrasound.⁸ Hospitalists who competently use ultrasound guidance should act on modern estimates of complications and may also choose to incorporate postprocedure ultrasound into their practice. Indeed, the Society of Hospital Medicine recommends against routine chest radiography in asymptomatic patients when sliding lung is visualized on postprocedure ultrasound.¹⁰

Second, procedural factors and postprocedural symptoms (new chest pain, dyspnea, or persistent cough) reliably identify patients with high risk of clinically meaningful complications. On one hand, only 1% to 2% of asymptomatic patients have a postprocedure pneumothorax, and clinical monitoring does not lead to chest tube placement in almost all of these cases.¹¹ On the other hand, 67% to 72% of symptomatic patients are found to have complications.¹² Doyle et al¹³ showed that the use of symptoms and procedure-specific factors (such as the aspiration of air, difficult procedure, multiple needle passes, or high operator suspicion of pneumothorax) could obviate the need for routine CXRs in approximately 60% of their procedures without any serious consequences.

Third, postprocedural CXRs very rarely reveal new or unexpected findings. For example, in one series,¹² only 3.8% of postdrainage radiographs uncovered new findings, none of which clarified the underlying diagnosis or changed management. To assess the utility of an initial thoracentesis and decide about repeat procedures, begin by asking the patient about symptoms and perform a physical exam.

CXRs might be helpful in certain scenarios, even when a complication is not suspected. For example, a postprocedure CXR to detect nonexpandable lung or evaluate the rate of recurrence may guide definitive management of patients with recurrent or malignant pleural effusion. Determining completeness of drainage may also assist with planning for palliative measures such as pleurodesis or indwelling pleural catheter placement. A postprocedure CXR is also helpful in patients with a technically difficult procedure or in those with symptoms during or immediately after the procedure. This recommendation is consistent with the 2010 British Thoracic Society guidelines, which recommend CXRs for procedures where air was withdrawn, the procedure was difficult, multiple needle passes were required, or the patient became symptomatic.¹⁴ The Society of Hospital Medicine’s recent Position Statement concurs with these guidelines and recommends against routine chest radiography in asymptomatic patients when sliding lung is visualized by postprocedure ultrasound.¹⁰

Hospitalists should not rountinely obtain post-thoracentesis CXRs in asymptomatic patients. Clinical monitoring with subsequent symptom-guided evaluation lowers costs, avoids unnecessary radiation exposure, and has been shown to be successful in a large case series of more than 9,300 patients.¹⁵ Some coughing should be expected with all large-volume thoracenteses as a normal response to re-expansion of atelectatic lung. The coughing should not persist past the immediate postprocedure period. If symptoms arise or if a complication is expected, the test of choice is either CXR or, if the hospitalist is a competent sonographer, bedside sonography. Bedside sonography is a low-cost, noninvasive method and has been well studied in the diagnosis of post-thoracentesis pneumothorax.¹⁶ CXRs may still be needed to confirm findings by sonography, to investigate postprocedural symptoms in those with pleural adhesions or other lung/pleural diseases (because ultrasonography is less reliable in these patients), or if reexpansion pulmonary edema or other complications are suspected. A robust quality improvement strategy to reduce unnecessary post-thoracentesis CXRs could result in cost savings and spare patients from radiation exposure, because a recent study of almost 1,000 thoracenteses performed at an academic medical center demonstrated that internal medicine residents, pulmonologists, and interventional radiologists order a CXR following 95% of thoracenteses.¹⁷ For a hypothetical hospital that orders 100 unnecessary post-thoracentesis CXRs annually, hospitalists could avoid approximately $7,000 in wasted expense per year.¹⁸

• Do not routinely order post-thoracentesis CXRs.
• Order a post-thoracentesis CXR if (1) the patient had new chest pain, dyspnea, or persistent cough during or after the procedure; (2) procedural features suggest increased risk of a complication (multiple needle passes, aspiration of air, difficulty obtaining fluid); or (3) a definitive palliative procedure will be arranged based on lung expansion.
• If qualified, use bedside sonography as a first step in the diagnosis of pneumothorax, reserving CXRs for those patients in whom accurate sonography is not possible, an alternative diagnosis is suspected, or when sonography findings are equivocal.

Following the uncomplicated thoracentesis, the hospitalist reconsidered the initial decision to order a CXR and rapidly assessed the patient’s risk of complications. Because the procedure required only one needle pass, air was not aspirated, and the patient did not experience prolonged coughing or pain, the CXR order was canceled. The patient recovered uneventfully and was spared the cost and radiation associated with the proposed CXR.

Do you think this is a low-value practice? Is this truly a “Thing We Do for No Reason?” Share what you do in your practice and join in the conversation online by retweeting it on Twitter (#TWDFNR) and liking it on Facebook. We invite you to propose ideas for other “Things We Do for No Reason” topics by emailing TWDFNR@hospitalmedicine.org.cknowledgments

Acknowledgements

The authors would like to thank Patricia Kritek and Somnath Mookherjee for their comments on an early version of this manuscript.

Disclosures

The authors have nothing to disclose.

The “Things We Do for No Reason” (TWDFNR) series reviews practices that have become common parts of hospital care but may provide little value to our patients. Practices reviewed in the TWDFNR series do not represent “black and white” conclusions or clinical practice standards but are meant as a starting place for research and active discussions among hospitalists and patients. We invite you to be part of that discussion. https://www.choosingwisely.org/

A nticoagulation is the cornerstone of acute venous thromboembolism (VTE) management. Nonetheless, the use of inferior vena cava (IVC) filters in addition to anticoagulation is increasing, with wide variation in practice patterns and a growing recognition of filter-related complications. Rigorous randomized controlled data demonstrating that IVC filters, particularly the increasingly commonly placed retrievable filters, provide a mortality benefit are sparse. Given our review of IVC filter use and the lack of evidence demonstrating that IVC filters provide a mortality benefit, we recommend using anticoagulation alone for stable medical service patients admitted with acute VTE. In nuanced cases, hospitalists should engage in multidisciplinary care to develop individualized treatment options.

A 65-year-old woman with a history of diabetes mellitus, metastatic breast cancer, and peptic ulcer disease presents to the Emergency Department for the evaluation of right thigh swelling, chest pain, and dyspnea after a transcontinental flight. Physical examination is notable for a pulse of 114 beats per minute, blood pressure of 136/93 mm Hg, respiratory rate of 14 breaths per minute, oxygen saturation of 95% on room air, and swelling of the right thigh. Computerized tomography imaging demonstrates multiple bilateral pulmonary emboli. Emergency department physicians begin anticoagulation and inform you that they have ordered the placement of a retrievable inferior vena cava (IVC) filter.

Acute venous thromboembolism (VTE) accounts for more than 500,000 hospitalizations in the United States each year.¹ Although the management of VTE centers around anticoagulation, the concurrent use of IVC filters has increased over the past several decades.² Several observational studies have attempted to quantify IVC filter usage and have shown that overall filter placement has increased at an impressive rate. Within two decades, the number of patients undergoing IVC filter placement has increased nearly 25 times from 2,000 in 1979 to 49,000 in 1999.² Recent Medicare data show that claims for IVC filter placement procedures have increased from 30,756 in 1999 to 65,041 in 2008.³ IVC filter placement rates are higher in the US than in other developed countries; one review projected that in 2012, the IVC filter placement rate in a given population in the US is 25 times higher than that in a similar population in Europe.⁴

The guidelines for IVC filter usage are largely based on expert opinion, and solid data regarding this intervention are lacking. This combination is problematic, especially because the practice is becoming commonplace, and filter-related complications are increasingly recognized. Additionally, the appropriateness of filter use varies among providers, as evidenced by a retrospective study in which three VTE experts reviewed medical records to determine the appropriateness of filter placement. They unanimously agreed that filter use was appropriate in 51% of the cases, unanimously agreed that filter use was inappropriate in 26% of the cases, and lacked consensus on the appropriateness of filter use in 23% of the cases.⁵ The striking lack of consensus among experts underscores the wide range of opinion regarding the appropriateness of IVC filter placement on a case-by-case basis. Moreover, evidence suggests that physician adherence to guidelines for appropriate IVC filter use is suboptimal. One single-center study showed that only 43.5% of filters placed by interventional radiology practitioners met the guidelines established by the American College of Chest Physicians (ACCP), with a slightly increased percentage of filter placement meeting guidelines if the requesting provider is an IM-trained physician.⁶

In theory, the concept of IVC filters makes intuitive sense—filters block the ascent of any thrombus from the lower extremities to prevent the feared complication of a pulmonary embolism (PE). Unfortunately, rigorous data are limited, and consensus guidelines vary between different specialty organizations, further obfuscating the role of IVC filter placement in the management of VTE. For example, the ACCP recommends against the use of IVC filters in most patients with VTE receiving anticoagulation and does not list any prophylactic indications.^7,8 Meanwhile, the Society of Interventional Radiology lists prophylactic indications for IVC filter placement in certain patient populations, such patients with a risk of VTE and a high risk of bleeding, and notes numerous relative indications for IVC filter placement.⁸ Notably, these differences in expert opinion likely influence practice patterns, as evidenced by the increase in IVC filter placement for relative indications.^9,10

The Prevention du Risque d’Embolie Pulmonaire par Interruption Cave (PRECIP) trial is the most robust study supporting the 2016 ACCP recommendation against IVC filter use in patients that can receive anticoagulation.^7,11 This study randomized 400 patients with deep vein thrombosis (DVT) at high risk for PE to anticoagulation with or without permanent filter placement to address VTE and mortality rates associated with IVC filter placement. The trial showed that the VTE burden shifts in the presence of IVC filters. At 2-year follow-up, the group with IVC filters had nonsignificantly fewer PEs than the control group and an increased incidence of DVT. Mortality rates did not differ between groups.¹¹ At eight-year follow-up this shift in VTE burden is again seen given that the number of PEs in patients who received IVC filters decreased and the incidence of DVTs increased. Again, mortality did not differ between groups.¹² A subsequent study randomized 399 patients with DVT and acute symptomatic PE with at least one additional marker of severity to anticoagulation with or without retrievable IVC filter placement and showed no difference in recurrent PE or mortality at 3 or 6 months.¹³ These results argue against placing retrievable filters in patients receiving anticoagulation.

The identification of associated adverse events further favor the judicious use of IVC filters. A retrospective review of the long-term complications of IVC filters based on imaging data showed a 14% fracture rate, 13% IVC thrombosis rate, and a 48% perforation rate.¹⁴ Multiple studies have shown that the associated complication rates of retrievable filters are higher than those of permanent filters; such an association is concerning given that retrievable filter usage exceeds permanent filter usage.^14,15 The increase in retrievable filter usage is likely attributable to their attractive risk-benefit calculation. In theory, retrievable IVC filters should be perfect for patients who have conditions that increase VTE risk but create temporary contraindications, such as trauma or major surgery, to anticoagulation. However, anticoagulation is preferred over IVC filters in the long term because the complication rates of IVC filters increase with dwell time.¹⁶ Given the reports of adverse events and concern that IVC filters are not appropriately removed, the Food and Drug Administration recommends removing retrievable IVC filters once the risk of filters outweighs the benefits, which appears to be 29-54 days after implantation.¹⁷ However, successful retrieval rates are low, both because of the low rates of removal attempts and because of the interference of complications, such as embedded or thrombosed filters, with removal.^10,18 As an example, in a retrospective review of all patients who received an IVC filter at an academic medical center over the period of 2003-2011, nearly 25% of patients were discharged on anticoagulation after IVC filter placement.¹⁰ This suggests that their contraindication to anticoagulation and need for IVC placement have passed by the time of discharge. Nevertheless, clinicians attempted filter retrieval in only 9.6% of these patients, representing a significant missed opportunity of treatment with anticoagulation rather than IVC filters.¹⁰

Factors such as filter plan documentation, hematology involvement, patient age ≤70 years, and establishment of dedicated IVC filter clinics are correlated with improved rates of filter removal; these correlations emphasize the importance of a clear follow-up plan in the timely removal of these devices.^18,19

IVC filter placement is inappropriate in the vast majority of patients with VTE who can be anticoagulated. However the ACCP does acknowledge that a small subset of patients – specifically, those with severe or massive PE – may fall outside this guideline.⁷ Clinicians fear that these patients have low cardiopulmonary reserve and may experience hemodynamic collapse and death with another “hit” from a recurrent PE. This recommendation is consistent with the evidence that in unstable patients with PE, IVC filter placement is associated with decreased in-hospital mortality.²⁰ Data remain limited for this situation, and the decision to place an IVC filter in anticoagulated but unstable patients is an individualized one.

In stable patients admitted to the medical service with VTE and who can be anticoagulated, there is little evidence that placement of an IVC filter will improve short- or long-term mortality. Hospitalists should anticoagulate these patients with a vitamin-K antagonist, heparin product, or novel oral anticoagulants.

• Anticoagulate hemodynamically stable patients who are admitted to the medical service with VTE and who do not have a contraindication to anticoagulation. Do not place a permanent or retrievable IVC filter.
• IVC filter placement may benefit unstable patients who may experience hemodynamic collapse with an increased PE burden. IVC filter placement should be discussed with a multidisciplinary team.
• When discharging a patient with an IVC filter, hospitalists should improve retrieval rates by scheduling subsequent removal. The discharge summary should contain information about the IVC filter, as well as clear instructions regarding the plan for removal. The instructions should include radiology follow-up information and the designation of responsible physicians in case of questions.

Although IVC filter use is increasing, the evidence does not support their use in hemodynamically stable patients who can be anticoagulated. The patient described in the initial case has no contraindication to systemic anticoagulation. Therefore, she should be started on anticoagulation, and an IVC filter should not be placed.

Do you think this is a low-value practice? Is this truly a “Thing We Do for No Reason?” Share what you do in your practice and join in the conversation online by retweeting it on Twitter (#TWDFNR) and liking it on Facebook. We invite you to propose ideas for other “Things We Do for No Reason” topics by emailingTWDFNR@hospitalmedicine.org.

Disclosures

The authors do not have any conflicts of interest to disclose

The “Things We Do for No Reason” (TWDFNR) series reviews practices that have become common parts of hospital care but may provide little value to our patients. Practices reviewed in the TWDFNR series do not represent “black and white” conclusions or clinical practice standards but are meant as a starting place for research and active discussions among hospitalists and patients. We invite you to be part of that discussion. https://www.choosingwisely.org/

A nticoagulation is the cornerstone of acute venous thromboembolism (VTE) management. Nonetheless, the use of inferior vena cava (IVC) filters in addition to anticoagulation is increasing, with wide variation in practice patterns and a growing recognition of filter-related complications. Rigorous randomized controlled data demonstrating that IVC filters, particularly the increasingly commonly placed retrievable filters, provide a mortality benefit are sparse. Given our review of IVC filter use and the lack of evidence demonstrating that IVC filters provide a mortality benefit, we recommend using anticoagulation alone for stable medical service patients admitted with acute VTE. In nuanced cases, hospitalists should engage in multidisciplinary care to develop individualized treatment options.

A 65-year-old woman with a history of diabetes mellitus, metastatic breast cancer, and peptic ulcer disease presents to the Emergency Department for the evaluation of right thigh swelling, chest pain, and dyspnea after a transcontinental flight. Physical examination is notable for a pulse of 114 beats per minute, blood pressure of 136/93 mm Hg, respiratory rate of 14 breaths per minute, oxygen saturation of 95% on room air, and swelling of the right thigh. Computerized tomography imaging demonstrates multiple bilateral pulmonary emboli. Emergency department physicians begin anticoagulation and inform you that they have ordered the placement of a retrievable inferior vena cava (IVC) filter.

Acute venous thromboembolism (VTE) accounts for more than 500,000 hospitalizations in the United States each year.¹ Although the management of VTE centers around anticoagulation, the concurrent use of IVC filters has increased over the past several decades.² Several observational studies have attempted to quantify IVC filter usage and have shown that overall filter placement has increased at an impressive rate. Within two decades, the number of patients undergoing IVC filter placement has increased nearly 25 times from 2,000 in 1979 to 49,000 in 1999.² Recent Medicare data show that claims for IVC filter placement procedures have increased from 30,756 in 1999 to 65,041 in 2008.³ IVC filter placement rates are higher in the US than in other developed countries; one review projected that in 2012, the IVC filter placement rate in a given population in the US is 25 times higher than that in a similar population in Europe.⁴

The guidelines for IVC filter usage are largely based on expert opinion, and solid data regarding this intervention are lacking. This combination is problematic, especially because the practice is becoming commonplace, and filter-related complications are increasingly recognized. Additionally, the appropriateness of filter use varies among providers, as evidenced by a retrospective study in which three VTE experts reviewed medical records to determine the appropriateness of filter placement. They unanimously agreed that filter use was appropriate in 51% of the cases, unanimously agreed that filter use was inappropriate in 26% of the cases, and lacked consensus on the appropriateness of filter use in 23% of the cases.⁵ The striking lack of consensus among experts underscores the wide range of opinion regarding the appropriateness of IVC filter placement on a case-by-case basis. Moreover, evidence suggests that physician adherence to guidelines for appropriate IVC filter use is suboptimal. One single-center study showed that only 43.5% of filters placed by interventional radiology practitioners met the guidelines established by the American College of Chest Physicians (ACCP), with a slightly increased percentage of filter placement meeting guidelines if the requesting provider is an IM-trained physician.⁶

In theory, the concept of IVC filters makes intuitive sense—filters block the ascent of any thrombus from the lower extremities to prevent the feared complication of a pulmonary embolism (PE). Unfortunately, rigorous data are limited, and consensus guidelines vary between different specialty organizations, further obfuscating the role of IVC filter placement in the management of VTE. For example, the ACCP recommends against the use of IVC filters in most patients with VTE receiving anticoagulation and does not list any prophylactic indications.^7,8 Meanwhile, the Society of Interventional Radiology lists prophylactic indications for IVC filter placement in certain patient populations, such patients with a risk of VTE and a high risk of bleeding, and notes numerous relative indications for IVC filter placement.⁸ Notably, these differences in expert opinion likely influence practice patterns, as evidenced by the increase in IVC filter placement for relative indications.^9,10

The Prevention du Risque d’Embolie Pulmonaire par Interruption Cave (PRECIP) trial is the most robust study supporting the 2016 ACCP recommendation against IVC filter use in patients that can receive anticoagulation.^7,11 This study randomized 400 patients with deep vein thrombosis (DVT) at high risk for PE to anticoagulation with or without permanent filter placement to address VTE and mortality rates associated with IVC filter placement. The trial showed that the VTE burden shifts in the presence of IVC filters. At 2-year follow-up, the group with IVC filters had nonsignificantly fewer PEs than the control group and an increased incidence of DVT. Mortality rates did not differ between groups.¹¹ At eight-year follow-up this shift in VTE burden is again seen given that the number of PEs in patients who received IVC filters decreased and the incidence of DVTs increased. Again, mortality did not differ between groups.¹² A subsequent study randomized 399 patients with DVT and acute symptomatic PE with at least one additional marker of severity to anticoagulation with or without retrievable IVC filter placement and showed no difference in recurrent PE or mortality at 3 or 6 months.¹³ These results argue against placing retrievable filters in patients receiving anticoagulation.

The identification of associated adverse events further favor the judicious use of IVC filters. A retrospective review of the long-term complications of IVC filters based on imaging data showed a 14% fracture rate, 13% IVC thrombosis rate, and a 48% perforation rate.¹⁴ Multiple studies have shown that the associated complication rates of retrievable filters are higher than those of permanent filters; such an association is concerning given that retrievable filter usage exceeds permanent filter usage.^14,15 The increase in retrievable filter usage is likely attributable to their attractive risk-benefit calculation. In theory, retrievable IVC filters should be perfect for patients who have conditions that increase VTE risk but create temporary contraindications, such as trauma or major surgery, to anticoagulation. However, anticoagulation is preferred over IVC filters in the long term because the complication rates of IVC filters increase with dwell time.¹⁶ Given the reports of adverse events and concern that IVC filters are not appropriately removed, the Food and Drug Administration recommends removing retrievable IVC filters once the risk of filters outweighs the benefits, which appears to be 29-54 days after implantation.¹⁷ However, successful retrieval rates are low, both because of the low rates of removal attempts and because of the interference of complications, such as embedded or thrombosed filters, with removal.^10,18 As an example, in a retrospective review of all patients who received an IVC filter at an academic medical center over the period of 2003-2011, nearly 25% of patients were discharged on anticoagulation after IVC filter placement.¹⁰ This suggests that their contraindication to anticoagulation and need for IVC placement have passed by the time of discharge. Nevertheless, clinicians attempted filter retrieval in only 9.6% of these patients, representing a significant missed opportunity of treatment with anticoagulation rather than IVC filters.¹⁰

Factors such as filter plan documentation, hematology involvement, patient age ≤70 years, and establishment of dedicated IVC filter clinics are correlated with improved rates of filter removal; these correlations emphasize the importance of a clear follow-up plan in the timely removal of these devices.^18,19

IVC filter placement is inappropriate in the vast majority of patients with VTE who can be anticoagulated. However the ACCP does acknowledge that a small subset of patients – specifically, those with severe or massive PE – may fall outside this guideline.⁷ Clinicians fear that these patients have low cardiopulmonary reserve and may experience hemodynamic collapse and death with another “hit” from a recurrent PE. This recommendation is consistent with the evidence that in unstable patients with PE, IVC filter placement is associated with decreased in-hospital mortality.²⁰ Data remain limited for this situation, and the decision to place an IVC filter in anticoagulated but unstable patients is an individualized one.

In stable patients admitted to the medical service with VTE and who can be anticoagulated, there is little evidence that placement of an IVC filter will improve short- or long-term mortality. Hospitalists should anticoagulate these patients with a vitamin-K antagonist, heparin product, or novel oral anticoagulants.

• Anticoagulate hemodynamically stable patients who are admitted to the medical service with VTE and who do not have a contraindication to anticoagulation. Do not place a permanent or retrievable IVC filter.
• IVC filter placement may benefit unstable patients who may experience hemodynamic collapse with an increased PE burden. IVC filter placement should be discussed with a multidisciplinary team.
• When discharging a patient with an IVC filter, hospitalists should improve retrieval rates by scheduling subsequent removal. The discharge summary should contain information about the IVC filter, as well as clear instructions regarding the plan for removal. The instructions should include radiology follow-up information and the designation of responsible physicians in case of questions.

Although IVC filter use is increasing, the evidence does not support their use in hemodynamically stable patients who can be anticoagulated. The patient described in the initial case has no contraindication to systemic anticoagulation. Therefore, she should be started on anticoagulation, and an IVC filter should not be placed.

Do you think this is a low-value practice? Is this truly a “Thing We Do for No Reason?” Share what you do in your practice and join in the conversation online by retweeting it on Twitter (#TWDFNR) and liking it on Facebook. We invite you to propose ideas for other “Things We Do for No Reason” topics by emailingTWDFNR@hospitalmedicine.org.

Disclosures

The authors do not have any conflicts of interest to disclose

The “Things We Do for No Reason” (TWDFNR) series reviews practices that have become common parts of hospital care but may provide little value to our patients. Practices reviewed in the TWDFNR series do not represent “black and white” conclusions or clinical practice standards but are meant as a starting place for research and active discussions among hospitalists and patients. We invite you to be part of that discussion. https://www.choosingwisely.org/

A nticoagulation is the cornerstone of acute venous thromboembolism (VTE) management. Nonetheless, the use of inferior vena cava (IVC) filters in addition to anticoagulation is increasing, with wide variation in practice patterns and a growing recognition of filter-related complications. Rigorous randomized controlled data demonstrating that IVC filters, particularly the increasingly commonly placed retrievable filters, provide a mortality benefit are sparse. Given our review of IVC filter use and the lack of evidence demonstrating that IVC filters provide a mortality benefit, we recommend using anticoagulation alone for stable medical service patients admitted with acute VTE. In nuanced cases, hospitalists should engage in multidisciplinary care to develop individualized treatment options.

A 65-year-old woman with a history of diabetes mellitus, metastatic breast cancer, and peptic ulcer disease presents to the Emergency Department for the evaluation of right thigh swelling, chest pain, and dyspnea after a transcontinental flight. Physical examination is notable for a pulse of 114 beats per minute, blood pressure of 136/93 mm Hg, respiratory rate of 14 breaths per minute, oxygen saturation of 95% on room air, and swelling of the right thigh. Computerized tomography imaging demonstrates multiple bilateral pulmonary emboli. Emergency department physicians begin anticoagulation and inform you that they have ordered the placement of a retrievable inferior vena cava (IVC) filter.

Acute venous thromboembolism (VTE) accounts for more than 500,000 hospitalizations in the United States each year.¹ Although the management of VTE centers around anticoagulation, the concurrent use of IVC filters has increased over the past several decades.² Several observational studies have attempted to quantify IVC filter usage and have shown that overall filter placement has increased at an impressive rate. Within two decades, the number of patients undergoing IVC filter placement has increased nearly 25 times from 2,000 in 1979 to 49,000 in 1999.² Recent Medicare data show that claims for IVC filter placement procedures have increased from 30,756 in 1999 to 65,041 in 2008.³ IVC filter placement rates are higher in the US than in other developed countries; one review projected that in 2012, the IVC filter placement rate in a given population in the US is 25 times higher than that in a similar population in Europe.⁴

The guidelines for IVC filter usage are largely based on expert opinion, and solid data regarding this intervention are lacking. This combination is problematic, especially because the practice is becoming commonplace, and filter-related complications are increasingly recognized. Additionally, the appropriateness of filter use varies among providers, as evidenced by a retrospective study in which three VTE experts reviewed medical records to determine the appropriateness of filter placement. They unanimously agreed that filter use was appropriate in 51% of the cases, unanimously agreed that filter use was inappropriate in 26% of the cases, and lacked consensus on the appropriateness of filter use in 23% of the cases.⁵ The striking lack of consensus among experts underscores the wide range of opinion regarding the appropriateness of IVC filter placement on a case-by-case basis. Moreover, evidence suggests that physician adherence to guidelines for appropriate IVC filter use is suboptimal. One single-center study showed that only 43.5% of filters placed by interventional radiology practitioners met the guidelines established by the American College of Chest Physicians (ACCP), with a slightly increased percentage of filter placement meeting guidelines if the requesting provider is an IM-trained physician.⁶

In theory, the concept of IVC filters makes intuitive sense—filters block the ascent of any thrombus from the lower extremities to prevent the feared complication of a pulmonary embolism (PE). Unfortunately, rigorous data are limited, and consensus guidelines vary between different specialty organizations, further obfuscating the role of IVC filter placement in the management of VTE. For example, the ACCP recommends against the use of IVC filters in most patients with VTE receiving anticoagulation and does not list any prophylactic indications.^7,8 Meanwhile, the Society of Interventional Radiology lists prophylactic indications for IVC filter placement in certain patient populations, such patients with a risk of VTE and a high risk of bleeding, and notes numerous relative indications for IVC filter placement.⁸ Notably, these differences in expert opinion likely influence practice patterns, as evidenced by the increase in IVC filter placement for relative indications.^9,10

The Prevention du Risque d’Embolie Pulmonaire par Interruption Cave (PRECIP) trial is the most robust study supporting the 2016 ACCP recommendation against IVC filter use in patients that can receive anticoagulation.^7,11 This study randomized 400 patients with deep vein thrombosis (DVT) at high risk for PE to anticoagulation with or without permanent filter placement to address VTE and mortality rates associated with IVC filter placement. The trial showed that the VTE burden shifts in the presence of IVC filters. At 2-year follow-up, the group with IVC filters had nonsignificantly fewer PEs than the control group and an increased incidence of DVT. Mortality rates did not differ between groups.¹¹ At eight-year follow-up this shift in VTE burden is again seen given that the number of PEs in patients who received IVC filters decreased and the incidence of DVTs increased. Again, mortality did not differ between groups.¹² A subsequent study randomized 399 patients with DVT and acute symptomatic PE with at least one additional marker of severity to anticoagulation with or without retrievable IVC filter placement and showed no difference in recurrent PE or mortality at 3 or 6 months.¹³ These results argue against placing retrievable filters in patients receiving anticoagulation.

The identification of associated adverse events further favor the judicious use of IVC filters. A retrospective review of the long-term complications of IVC filters based on imaging data showed a 14% fracture rate, 13% IVC thrombosis rate, and a 48% perforation rate.¹⁴ Multiple studies have shown that the associated complication rates of retrievable filters are higher than those of permanent filters; such an association is concerning given that retrievable filter usage exceeds permanent filter usage.^14,15 The increase in retrievable filter usage is likely attributable to their attractive risk-benefit calculation. In theory, retrievable IVC filters should be perfect for patients who have conditions that increase VTE risk but create temporary contraindications, such as trauma or major surgery, to anticoagulation. However, anticoagulation is preferred over IVC filters in the long term because the complication rates of IVC filters increase with dwell time.¹⁶ Given the reports of adverse events and concern that IVC filters are not appropriately removed, the Food and Drug Administration recommends removing retrievable IVC filters once the risk of filters outweighs the benefits, which appears to be 29-54 days after implantation.¹⁷ However, successful retrieval rates are low, both because of the low rates of removal attempts and because of the interference of complications, such as embedded or thrombosed filters, with removal.^10,18 As an example, in a retrospective review of all patients who received an IVC filter at an academic medical center over the period of 2003-2011, nearly 25% of patients were discharged on anticoagulation after IVC filter placement.¹⁰ This suggests that their contraindication to anticoagulation and need for IVC placement have passed by the time of discharge. Nevertheless, clinicians attempted filter retrieval in only 9.6% of these patients, representing a significant missed opportunity of treatment with anticoagulation rather than IVC filters.¹⁰

Factors such as filter plan documentation, hematology involvement, patient age ≤70 years, and establishment of dedicated IVC filter clinics are correlated with improved rates of filter removal; these correlations emphasize the importance of a clear follow-up plan in the timely removal of these devices.^18,19

IVC filter placement is inappropriate in the vast majority of patients with VTE who can be anticoagulated. However the ACCP does acknowledge that a small subset of patients – specifically, those with severe or massive PE – may fall outside this guideline.⁷ Clinicians fear that these patients have low cardiopulmonary reserve and may experience hemodynamic collapse and death with another “hit” from a recurrent PE. This recommendation is consistent with the evidence that in unstable patients with PE, IVC filter placement is associated with decreased in-hospital mortality.²⁰ Data remain limited for this situation, and the decision to place an IVC filter in anticoagulated but unstable patients is an individualized one.

In stable patients admitted to the medical service with VTE and who can be anticoagulated, there is little evidence that placement of an IVC filter will improve short- or long-term mortality. Hospitalists should anticoagulate these patients with a vitamin-K antagonist, heparin product, or novel oral anticoagulants.

• Anticoagulate hemodynamically stable patients who are admitted to the medical service with VTE and who do not have a contraindication to anticoagulation. Do not place a permanent or retrievable IVC filter.
• IVC filter placement may benefit unstable patients who may experience hemodynamic collapse with an increased PE burden. IVC filter placement should be discussed with a multidisciplinary team.
• When discharging a patient with an IVC filter, hospitalists should improve retrieval rates by scheduling subsequent removal. The discharge summary should contain information about the IVC filter, as well as clear instructions regarding the plan for removal. The instructions should include radiology follow-up information and the designation of responsible physicians in case of questions.

Although IVC filter use is increasing, the evidence does not support their use in hemodynamically stable patients who can be anticoagulated. The patient described in the initial case has no contraindication to systemic anticoagulation. Therefore, she should be started on anticoagulation, and an IVC filter should not be placed.

Do you think this is a low-value practice? Is this truly a “Thing We Do for No Reason?” Share what you do in your practice and join in the conversation online by retweeting it on Twitter (#TWDFNR) and liking it on Facebook. We invite you to propose ideas for other “Things We Do for No Reason” topics by emailingTWDFNR@hospitalmedicine.org.

Disclosures

The authors do not have any conflicts of interest to disclose

The “Things We Do for No Reason” series reviews practices which have become common parts of hospital care but which may provide little value to our patients. Practices reviewed in the TWDFNR series do not represent “black and white” conclusions or clinical practice standards, but are meant as a starting place for research and active discussions among hospitalists and patients. We invite you to be part of that discussion. https://www.choosingwisely.org/

CASE PRESENTATION

A 12-month-old female is admitted for acute bronchiolitis with increased work of breathing and decreased oral intake. She is mildly dehydrated upon exam with a sodium level of 139 mEq/L and is given a 20 mL/kg bolus of 0.9% saline. Given the patient’s poor oral intake, the admitting intern orders maintenance intravenous (IV) fluids and asks her senior resident which IV fluid should be used. The medical student on the team wonders if a different IV fluid would be selected for a 2-week-old with a similar presentation.

Maintenance IV fluids are continuously infused to preserve extracellular volume and electrolyte balance when fluids cannot be taken orally. In contrast, resuscitation IV fluids are given as a bolus to patients in states of hypoperfusion to restore extracellular volume. The given IV fluid concentration can be categorized as approximately equal to (isotonic) or less than (hypotonic) the plasma sodium concentration. Refer to Table 1 for the electrolyte composition of commonly used IV fluids. Dextrose is rapidly metabolized upon infusion and does not affect tonicity.

Why You Might Think Hypotonic Maintenance IV Fluids Are The Right Choice

A 1957 publication by Holliday and Segar laid the foundation for maintenance IV fluid and electrolyte requirements in children and was the initial catalyst for the use of hypotonic maintenance IV fluids.¹ This manuscript contended that hypotonic IV fluids could supply the water and sodium needed to meet maintenance dietary requirements. This claim led to the predominant use of hypotonic maintenance IV fluids in children. By contrast, isotonic IV fluids have been avoided given the apprehension over electrolytes exceeding maintenance needs.

Concerns about the unintended consequences of fluid overload – edema, hypernatremia, and hypertension secondary to increased sodium load – have led some to avoid isotonic IV fluids.² When presented with common clinical scenarios of patients at risk for excess antidiuretic hormone (ADH; also known as arginine vasopressin), pediatric residents chose hypotonic (instead of isotonic) IV fluids 78% of the time.³

Why Isotonic Maintenance IV Fluids Are Usually The Right Choice For Children

General recommendations for hypotonic IV fluids are primarily based on theoretical calculations from the fluid and electrolyte requirements of healthy individuals, and studies have not validated the use of hypotonic IV fluids in clinical practice.¹ Acutely ill patients are at risk for excessive levels of ADH from numerous causes (see Table 2).² As a result, nearly every hospitalized patient is at risk for excess ADH release, thus making them vulnerable to the development of hyponatremia. The syndrome of inappropriate secretion of ADH (SIADH) occurs when nonosmotic/nonhemodynamic stimuli trigger ADH release, which leads to excessive free-water retention and resultant hyponatremia. Schwartz and Bartter reported the first two cases of SIADH in 1957 when hyponatremia developed in the setting of bronchogenic carcinoma.⁴ Although the publication by Holliday and Seger did acknowledge the potential for water intoxication, it was written before this report and before the effects of ADH on the sodium levels of hospitalized patients were clearly understood.² SIADH is now recognized as one of the most common causes of hyponatremia in hospitalized patients.^{5, 6}

Numerous studies have demonstrated that patients who receive hypotonic IV fluids have a significantly higher risk of developing hyponatremia than patients who receive isotonic IV fluids.^7,8 An infrequent, yet serious, complication of iatrogenic hyponatremia is hyponatremic encephalopathy, which carries a high rate of morbidity or mortality.⁹ The prevention of hyponatremia is essential as the early symptoms of hyponatremic encephalopathy are nonspecific and can be easily missed.²

More than 15 prospective randomized controlled trials (RCTs) involving over 2,000 children have demonstrated that isotonic IV fluids are more effective in preventing hospital-acquired hyponatremia than hypotonic IV fluids and are not associated with the development of fluid overload or hypernatremia. A 2014 metaanalysis comprising 10 RCTs and involving over 800 children found that when compared with isotonic IV fluids, hypotonic IV fluids present a relative risk of 2.37 for sodium levels to drop below 135 mEq/L and a relative risk of 6.1 for levels to drop below 130 mEq/L. The numbers needed to treat (NNT) with isotonic IV fluids to prevent hyponatremia in each group were 6 and 17, respectively.⁷ A Cochrane review published in 2014 presented comparable findings, demonstrating that hypotonic IV fluids had a 34% risk of causing hyponatremia; by comparison, isotonic IV fluids had a 17% risk of causing hyponatremia and a NNT of six to prevent hyponatremia.⁸ In a large RCT conducted in 2015 with 676 pediatric patients, McNabb et al. found that when compared with patients receiving isotonic IV fluids, those receiving hypotonic IV fluids had a higher incidence of developing hyponatremia (10.9% versus 3.8%) with a NNT of 15 to prevent hyponatremia with the use of isotonic fluids.¹⁰ Published trials have likely been underpowered to detect a difference in the infrequent adverse hyponatremia outcomes of seizures and mortality.

On the basis of these data, patient safety alerts have recommended the avoidance of hypotonic IV fluids in the United Kingdom (UK) and Australia, and the 2015 UK guidelines for children now recommend isotonic IV fluids for maintenance needs.¹¹ Although many of the aforementioned studies included predominantly critically ill or surgical pediatric patients, the risk of hyponatremia with hypotonic IV fluids seems similarly increased in nonsurgical and noncritically ill pediatric patients.¹⁰

For patients at risk for excess ADH release, some have supported the use of hypotonic IV fluids at a lower than maintenance rate to theoretically decrease the risk of hyponatremia, but this practice has not been effective in preventing hyponatremia.^2,12 Unless a patient is in a fluid overload state, such as in congestive heart failure, cirrhosis, or renal failure; isotonic maintenance IV fluids should not result in fluid overload.³ Available evidence for guiding maintenance IV fluid choice in neonates or young infants is limited. Nevertheless, given the aforementioned reasons, we generally recommend the prescription of isotonic IV fluids for most in this population.

Which Isotonic IV Fluid Should Be Used?

The sodium concentration (154 mmol/L) of 0.9% saline, an isotonic IV fluid, is approximately equal to the tonicity of the aqueous phase of plasma. The majority of studies evaluating the risk of hyponatremia with maintenance IV fluids have used 0.9% saline as the studied isotonic IV fluid. Plasma-Lyte and Ringer’s lactate are low-chloride, buffered/balanced solutions. Plasma-Lyte ([Na] = 140 mmol/L) has been demonstrated to be effective in preventing hyponatremia. Ringers’ lactate is slightly hypotonic ([Na] = 130 mmol/L), and its administration is associated with a decrease in serum sodium.¹³ A resultant dilutional and hyperchloremic metabolic acidosis is more likely to develop with the use of large volumes of 0.9% saline in resuscitation than with the use of balanced solutions.² Whether the prolonged use of 0.9% saline maintenance IV fluids can lead to this same side effect remains unknown given insufficient evidence.² Retrospective studies using balanced solutions have shown an association with decreased rates of acute kidney injury (AKI) and mortality when compared with 0.9% saline. However, a RCT with over 2,000 adult ICU patients showed no change in rates of AKI in those that received Plasma-Lyte compared with those who received 0.9% saline.¹⁴

Two recent, single-center, prospective studies compared the use of Ringer’s lactate or Plasma-Lyte for resuscitation with that of 0.9% saline. One study was comprised of 15,802 critically ill adults, and the other was comprised of 13,347 noncritically adults. Both studies showed that balanced solutions decreased the rate of major adverse kidney events (defined as a composite of death from any cause, new renal-replacement therapy, or persistent renal injury) within 30 days.^15,16 Available published pediatric studies indicate that 0.9% saline is an effective maintenance IV fluid for the prevention of hyponatremia that is not associated with hypernatremia or fluid overload. Further pediatric studies comparing 0.9% saline with balanced solutions are needed.

When Should We Use Hypotonic IV Fluids?

Hypotonic IV fluids may be needed for patients with hypernatremia and a free-water deficit or a renal-concentrating defect with ongoing urinary free-water losses.² Special care should be taken when choosing maintenance IV fluids for patients with renal disease, liver disease, or heart failure given that these groups have been excluded from some studies.¹² These patients may be at risk for increased salt and fluid retention with any IV fluid, and fluid rates need to be restricted. The fluid intake of patients with hyponatremia secondary to SIADH needs close management; these patients benefit from total fluid restriction instead of standard maintenance IV fluid rates.²

What We Should Do Instead?

Maintenance IV fluids should only be used when necessary and should be stopped as soon as they are no longer required, especially in light of the recent shortages in 0.9% saline.¹⁷ Similar to all medications, maintenance IV fluids should be individualized to the patient’s needs on the basis of the indication for IV fluids and the patient’s comorbidities.² Consideration should be given to checking the patient’s electrolyte levels to monitor response to IV fluids, especially during the first 24 hours of admission when risk of hyponatremia is highest. Isotonic IV fluids with 5% dextrose should be used as the maintenance IV fluid in the majority of hospitalized children given its proven benefit in decreasing the rate of hospital-acquired hyponatremia.^7,8 Hypotonic IV fluids should be avoided as the default maintenance IV fluid and should only be utilized under specific circumstances.

• When needed, maintenance IV fluids should always be tailored to each individual patient.
• For most acutely ill hospitalized children, isotonic IV fluids should be the maintenance IV fluid of choice.
• Consider monitoring electrolytes to determine the effects of maintenance IV fluids.

Enteral maintenance fluids should be used first-line if possible. Although hypotonic IV fluids have historically been the maintenance IV fluid of choice, this class of IV fluids should be avoided for most hospitalized children to decrease the significant risk of iatrogenic hyponatremia, which can be severe and have catastrophic complications. When necessary, isotonic IV fluids should be used for the majority of hospitalized children given that these fluids present a significantly decreased risk for causing hyponatremia. Returning to our case presentation, to decrease the risk of hyponatremia, the senior resident should recommend starting isotonic IV fluids in the 12-month-old and theoretical 2-week-old until oral intake can be maintained.

Do you think this is a low-value practice? Is this truly a “Thing We Do for No Reason”? Let us know what you do in your practice and propose ideas for other “Things We Do for No Reason” topics. Please join in the conversation online at Twitter (#TWDFNR)/Facebook and don’t forget to “Like It” on Facebook or retweet it on Twitter.

Disclosure

The authors have no relevant conflicts of interest to report. No payment or services from a 3rd party were received for any aspect of this submitted work. The authors have no financial relationships with entities in the biomedical arena that could be perceived to influence, or that give the appearance of potentially influencing, what was written in this submitted work.

The “Things We Do for No Reason” series reviews practices which have become common parts of hospital care but which may provide little value to our patients. Practices reviewed in the TWDFNR series do not represent “black and white” conclusions or clinical practice standards, but are meant as a starting place for research and active discussions among hospitalists and patients. We invite you to be part of that discussion. https://www.choosingwisely.org/

CASE PRESENTATION

A 12-month-old female is admitted for acute bronchiolitis with increased work of breathing and decreased oral intake. She is mildly dehydrated upon exam with a sodium level of 139 mEq/L and is given a 20 mL/kg bolus of 0.9% saline. Given the patient’s poor oral intake, the admitting intern orders maintenance intravenous (IV) fluids and asks her senior resident which IV fluid should be used. The medical student on the team wonders if a different IV fluid would be selected for a 2-week-old with a similar presentation.

Maintenance IV fluids are continuously infused to preserve extracellular volume and electrolyte balance when fluids cannot be taken orally. In contrast, resuscitation IV fluids are given as a bolus to patients in states of hypoperfusion to restore extracellular volume. The given IV fluid concentration can be categorized as approximately equal to (isotonic) or less than (hypotonic) the plasma sodium concentration. Refer to Table 1 for the electrolyte composition of commonly used IV fluids. Dextrose is rapidly metabolized upon infusion and does not affect tonicity.

Why You Might Think Hypotonic Maintenance IV Fluids Are The Right Choice

A 1957 publication by Holliday and Segar laid the foundation for maintenance IV fluid and electrolyte requirements in children and was the initial catalyst for the use of hypotonic maintenance IV fluids.¹ This manuscript contended that hypotonic IV fluids could supply the water and sodium needed to meet maintenance dietary requirements. This claim led to the predominant use of hypotonic maintenance IV fluids in children. By contrast, isotonic IV fluids have been avoided given the apprehension over electrolytes exceeding maintenance needs.

Concerns about the unintended consequences of fluid overload – edema, hypernatremia, and hypertension secondary to increased sodium load – have led some to avoid isotonic IV fluids.² When presented with common clinical scenarios of patients at risk for excess antidiuretic hormone (ADH; also known as arginine vasopressin), pediatric residents chose hypotonic (instead of isotonic) IV fluids 78% of the time.³

Why Isotonic Maintenance IV Fluids Are Usually The Right Choice For Children

General recommendations for hypotonic IV fluids are primarily based on theoretical calculations from the fluid and electrolyte requirements of healthy individuals, and studies have not validated the use of hypotonic IV fluids in clinical practice.¹ Acutely ill patients are at risk for excessive levels of ADH from numerous causes (see Table 2).² As a result, nearly every hospitalized patient is at risk for excess ADH release, thus making them vulnerable to the development of hyponatremia. The syndrome of inappropriate secretion of ADH (SIADH) occurs when nonosmotic/nonhemodynamic stimuli trigger ADH release, which leads to excessive free-water retention and resultant hyponatremia. Schwartz and Bartter reported the first two cases of SIADH in 1957 when hyponatremia developed in the setting of bronchogenic carcinoma.⁴ Although the publication by Holliday and Seger did acknowledge the potential for water intoxication, it was written before this report and before the effects of ADH on the sodium levels of hospitalized patients were clearly understood.² SIADH is now recognized as one of the most common causes of hyponatremia in hospitalized patients.^{5, 6}

Numerous studies have demonstrated that patients who receive hypotonic IV fluids have a significantly higher risk of developing hyponatremia than patients who receive isotonic IV fluids.^7,8 An infrequent, yet serious, complication of iatrogenic hyponatremia is hyponatremic encephalopathy, which carries a high rate of morbidity or mortality.⁹ The prevention of hyponatremia is essential as the early symptoms of hyponatremic encephalopathy are nonspecific and can be easily missed.²

More than 15 prospective randomized controlled trials (RCTs) involving over 2,000 children have demonstrated that isotonic IV fluids are more effective in preventing hospital-acquired hyponatremia than hypotonic IV fluids and are not associated with the development of fluid overload or hypernatremia. A 2014 metaanalysis comprising 10 RCTs and involving over 800 children found that when compared with isotonic IV fluids, hypotonic IV fluids present a relative risk of 2.37 for sodium levels to drop below 135 mEq/L and a relative risk of 6.1 for levels to drop below 130 mEq/L. The numbers needed to treat (NNT) with isotonic IV fluids to prevent hyponatremia in each group were 6 and 17, respectively.⁷ A Cochrane review published in 2014 presented comparable findings, demonstrating that hypotonic IV fluids had a 34% risk of causing hyponatremia; by comparison, isotonic IV fluids had a 17% risk of causing hyponatremia and a NNT of six to prevent hyponatremia.⁸ In a large RCT conducted in 2015 with 676 pediatric patients, McNabb et al. found that when compared with patients receiving isotonic IV fluids, those receiving hypotonic IV fluids had a higher incidence of developing hyponatremia (10.9% versus 3.8%) with a NNT of 15 to prevent hyponatremia with the use of isotonic fluids.¹⁰ Published trials have likely been underpowered to detect a difference in the infrequent adverse hyponatremia outcomes of seizures and mortality.

On the basis of these data, patient safety alerts have recommended the avoidance of hypotonic IV fluids in the United Kingdom (UK) and Australia, and the 2015 UK guidelines for children now recommend isotonic IV fluids for maintenance needs.¹¹ Although many of the aforementioned studies included predominantly critically ill or surgical pediatric patients, the risk of hyponatremia with hypotonic IV fluids seems similarly increased in nonsurgical and noncritically ill pediatric patients.¹⁰

For patients at risk for excess ADH release, some have supported the use of hypotonic IV fluids at a lower than maintenance rate to theoretically decrease the risk of hyponatremia, but this practice has not been effective in preventing hyponatremia.^2,12 Unless a patient is in a fluid overload state, such as in congestive heart failure, cirrhosis, or renal failure; isotonic maintenance IV fluids should not result in fluid overload.³ Available evidence for guiding maintenance IV fluid choice in neonates or young infants is limited. Nevertheless, given the aforementioned reasons, we generally recommend the prescription of isotonic IV fluids for most in this population.

Which Isotonic IV Fluid Should Be Used?

The sodium concentration (154 mmol/L) of 0.9% saline, an isotonic IV fluid, is approximately equal to the tonicity of the aqueous phase of plasma. The majority of studies evaluating the risk of hyponatremia with maintenance IV fluids have used 0.9% saline as the studied isotonic IV fluid. Plasma-Lyte and Ringer’s lactate are low-chloride, buffered/balanced solutions. Plasma-Lyte ([Na] = 140 mmol/L) has been demonstrated to be effective in preventing hyponatremia. Ringers’ lactate is slightly hypotonic ([Na] = 130 mmol/L), and its administration is associated with a decrease in serum sodium.¹³ A resultant dilutional and hyperchloremic metabolic acidosis is more likely to develop with the use of large volumes of 0.9% saline in resuscitation than with the use of balanced solutions.² Whether the prolonged use of 0.9% saline maintenance IV fluids can lead to this same side effect remains unknown given insufficient evidence.² Retrospective studies using balanced solutions have shown an association with decreased rates of acute kidney injury (AKI) and mortality when compared with 0.9% saline. However, a RCT with over 2,000 adult ICU patients showed no change in rates of AKI in those that received Plasma-Lyte compared with those who received 0.9% saline.¹⁴

Two recent, single-center, prospective studies compared the use of Ringer’s lactate or Plasma-Lyte for resuscitation with that of 0.9% saline. One study was comprised of 15,802 critically ill adults, and the other was comprised of 13,347 noncritically adults. Both studies showed that balanced solutions decreased the rate of major adverse kidney events (defined as a composite of death from any cause, new renal-replacement therapy, or persistent renal injury) within 30 days.^15,16 Available published pediatric studies indicate that 0.9% saline is an effective maintenance IV fluid for the prevention of hyponatremia that is not associated with hypernatremia or fluid overload. Further pediatric studies comparing 0.9% saline with balanced solutions are needed.

When Should We Use Hypotonic IV Fluids?

Hypotonic IV fluids may be needed for patients with hypernatremia and a free-water deficit or a renal-concentrating defect with ongoing urinary free-water losses.² Special care should be taken when choosing maintenance IV fluids for patients with renal disease, liver disease, or heart failure given that these groups have been excluded from some studies.¹² These patients may be at risk for increased salt and fluid retention with any IV fluid, and fluid rates need to be restricted. The fluid intake of patients with hyponatremia secondary to SIADH needs close management; these patients benefit from total fluid restriction instead of standard maintenance IV fluid rates.²

What We Should Do Instead?

Maintenance IV fluids should only be used when necessary and should be stopped as soon as they are no longer required, especially in light of the recent shortages in 0.9% saline.¹⁷ Similar to all medications, maintenance IV fluids should be individualized to the patient’s needs on the basis of the indication for IV fluids and the patient’s comorbidities.² Consideration should be given to checking the patient’s electrolyte levels to monitor response to IV fluids, especially during the first 24 hours of admission when risk of hyponatremia is highest. Isotonic IV fluids with 5% dextrose should be used as the maintenance IV fluid in the majority of hospitalized children given its proven benefit in decreasing the rate of hospital-acquired hyponatremia.^7,8 Hypotonic IV fluids should be avoided as the default maintenance IV fluid and should only be utilized under specific circumstances.

• When needed, maintenance IV fluids should always be tailored to each individual patient.
• For most acutely ill hospitalized children, isotonic IV fluids should be the maintenance IV fluid of choice.
• Consider monitoring electrolytes to determine the effects of maintenance IV fluids.

Enteral maintenance fluids should be used first-line if possible. Although hypotonic IV fluids have historically been the maintenance IV fluid of choice, this class of IV fluids should be avoided for most hospitalized children to decrease the significant risk of iatrogenic hyponatremia, which can be severe and have catastrophic complications. When necessary, isotonic IV fluids should be used for the majority of hospitalized children given that these fluids present a significantly decreased risk for causing hyponatremia. Returning to our case presentation, to decrease the risk of hyponatremia, the senior resident should recommend starting isotonic IV fluids in the 12-month-old and theoretical 2-week-old until oral intake can be maintained.

Do you think this is a low-value practice? Is this truly a “Thing We Do for No Reason”? Let us know what you do in your practice and propose ideas for other “Things We Do for No Reason” topics. Please join in the conversation online at Twitter (#TWDFNR)/Facebook and don’t forget to “Like It” on Facebook or retweet it on Twitter.

Disclosure

The authors have no relevant conflicts of interest to report. No payment or services from a 3rd party were received for any aspect of this submitted work. The authors have no financial relationships with entities in the biomedical arena that could be perceived to influence, or that give the appearance of potentially influencing, what was written in this submitted work.

The “Things We Do for No Reason” series reviews practices which have become common parts of hospital care but which may provide little value to our patients. Practices reviewed in the TWDFNR series do not represent “black and white” conclusions or clinical practice standards, but are meant as a starting place for research and active discussions among hospitalists and patients. We invite you to be part of that discussion. https://www.choosingwisely.org/

CASE PRESENTATION

A 12-month-old female is admitted for acute bronchiolitis with increased work of breathing and decreased oral intake. She is mildly dehydrated upon exam with a sodium level of 139 mEq/L and is given a 20 mL/kg bolus of 0.9% saline. Given the patient’s poor oral intake, the admitting intern orders maintenance intravenous (IV) fluids and asks her senior resident which IV fluid should be used. The medical student on the team wonders if a different IV fluid would be selected for a 2-week-old with a similar presentation.

Maintenance IV fluids are continuously infused to preserve extracellular volume and electrolyte balance when fluids cannot be taken orally. In contrast, resuscitation IV fluids are given as a bolus to patients in states of hypoperfusion to restore extracellular volume. The given IV fluid concentration can be categorized as approximately equal to (isotonic) or less than (hypotonic) the plasma sodium concentration. Refer to Table 1 for the electrolyte composition of commonly used IV fluids. Dextrose is rapidly metabolized upon infusion and does not affect tonicity.

Why You Might Think Hypotonic Maintenance IV Fluids Are The Right Choice

A 1957 publication by Holliday and Segar laid the foundation for maintenance IV fluid and electrolyte requirements in children and was the initial catalyst for the use of hypotonic maintenance IV fluids.¹ This manuscript contended that hypotonic IV fluids could supply the water and sodium needed to meet maintenance dietary requirements. This claim led to the predominant use of hypotonic maintenance IV fluids in children. By contrast, isotonic IV fluids have been avoided given the apprehension over electrolytes exceeding maintenance needs.

Concerns about the unintended consequences of fluid overload – edema, hypernatremia, and hypertension secondary to increased sodium load – have led some to avoid isotonic IV fluids.² When presented with common clinical scenarios of patients at risk for excess antidiuretic hormone (ADH; also known as arginine vasopressin), pediatric residents chose hypotonic (instead of isotonic) IV fluids 78% of the time.³

Why Isotonic Maintenance IV Fluids Are Usually The Right Choice For Children

General recommendations for hypotonic IV fluids are primarily based on theoretical calculations from the fluid and electrolyte requirements of healthy individuals, and studies have not validated the use of hypotonic IV fluids in clinical practice.¹ Acutely ill patients are at risk for excessive levels of ADH from numerous causes (see Table 2).² As a result, nearly every hospitalized patient is at risk for excess ADH release, thus making them vulnerable to the development of hyponatremia. The syndrome of inappropriate secretion of ADH (SIADH) occurs when nonosmotic/nonhemodynamic stimuli trigger ADH release, which leads to excessive free-water retention and resultant hyponatremia. Schwartz and Bartter reported the first two cases of SIADH in 1957 when hyponatremia developed in the setting of bronchogenic carcinoma.⁴ Although the publication by Holliday and Seger did acknowledge the potential for water intoxication, it was written before this report and before the effects of ADH on the sodium levels of hospitalized patients were clearly understood.² SIADH is now recognized as one of the most common causes of hyponatremia in hospitalized patients.^{5, 6}

Numerous studies have demonstrated that patients who receive hypotonic IV fluids have a significantly higher risk of developing hyponatremia than patients who receive isotonic IV fluids.^7,8 An infrequent, yet serious, complication of iatrogenic hyponatremia is hyponatremic encephalopathy, which carries a high rate of morbidity or mortality.⁹ The prevention of hyponatremia is essential as the early symptoms of hyponatremic encephalopathy are nonspecific and can be easily missed.²

More than 15 prospective randomized controlled trials (RCTs) involving over 2,000 children have demonstrated that isotonic IV fluids are more effective in preventing hospital-acquired hyponatremia than hypotonic IV fluids and are not associated with the development of fluid overload or hypernatremia. A 2014 metaanalysis comprising 10 RCTs and involving over 800 children found that when compared with isotonic IV fluids, hypotonic IV fluids present a relative risk of 2.37 for sodium levels to drop below 135 mEq/L and a relative risk of 6.1 for levels to drop below 130 mEq/L. The numbers needed to treat (NNT) with isotonic IV fluids to prevent hyponatremia in each group were 6 and 17, respectively.⁷ A Cochrane review published in 2014 presented comparable findings, demonstrating that hypotonic IV fluids had a 34% risk of causing hyponatremia; by comparison, isotonic IV fluids had a 17% risk of causing hyponatremia and a NNT of six to prevent hyponatremia.⁸ In a large RCT conducted in 2015 with 676 pediatric patients, McNabb et al. found that when compared with patients receiving isotonic IV fluids, those receiving hypotonic IV fluids had a higher incidence of developing hyponatremia (10.9% versus 3.8%) with a NNT of 15 to prevent hyponatremia with the use of isotonic fluids.¹⁰ Published trials have likely been underpowered to detect a difference in the infrequent adverse hyponatremia outcomes of seizures and mortality.

On the basis of these data, patient safety alerts have recommended the avoidance of hypotonic IV fluids in the United Kingdom (UK) and Australia, and the 2015 UK guidelines for children now recommend isotonic IV fluids for maintenance needs.¹¹ Although many of the aforementioned studies included predominantly critically ill or surgical pediatric patients, the risk of hyponatremia with hypotonic IV fluids seems similarly increased in nonsurgical and noncritically ill pediatric patients.¹⁰

For patients at risk for excess ADH release, some have supported the use of hypotonic IV fluids at a lower than maintenance rate to theoretically decrease the risk of hyponatremia, but this practice has not been effective in preventing hyponatremia.^2,12 Unless a patient is in a fluid overload state, such as in congestive heart failure, cirrhosis, or renal failure; isotonic maintenance IV fluids should not result in fluid overload.³ Available evidence for guiding maintenance IV fluid choice in neonates or young infants is limited. Nevertheless, given the aforementioned reasons, we generally recommend the prescription of isotonic IV fluids for most in this population.

Which Isotonic IV Fluid Should Be Used?

The sodium concentration (154 mmol/L) of 0.9% saline, an isotonic IV fluid, is approximately equal to the tonicity of the aqueous phase of plasma. The majority of studies evaluating the risk of hyponatremia with maintenance IV fluids have used 0.9% saline as the studied isotonic IV fluid. Plasma-Lyte and Ringer’s lactate are low-chloride, buffered/balanced solutions. Plasma-Lyte ([Na] = 140 mmol/L) has been demonstrated to be effective in preventing hyponatremia. Ringers’ lactate is slightly hypotonic ([Na] = 130 mmol/L), and its administration is associated with a decrease in serum sodium.¹³ A resultant dilutional and hyperchloremic metabolic acidosis is more likely to develop with the use of large volumes of 0.9% saline in resuscitation than with the use of balanced solutions.² Whether the prolonged use of 0.9% saline maintenance IV fluids can lead to this same side effect remains unknown given insufficient evidence.² Retrospective studies using balanced solutions have shown an association with decreased rates of acute kidney injury (AKI) and mortality when compared with 0.9% saline. However, a RCT with over 2,000 adult ICU patients showed no change in rates of AKI in those that received Plasma-Lyte compared with those who received 0.9% saline.¹⁴

Two recent, single-center, prospective studies compared the use of Ringer’s lactate or Plasma-Lyte for resuscitation with that of 0.9% saline. One study was comprised of 15,802 critically ill adults, and the other was comprised of 13,347 noncritically adults. Both studies showed that balanced solutions decreased the rate of major adverse kidney events (defined as a composite of death from any cause, new renal-replacement therapy, or persistent renal injury) within 30 days.^15,16 Available published pediatric studies indicate that 0.9% saline is an effective maintenance IV fluid for the prevention of hyponatremia that is not associated with hypernatremia or fluid overload. Further pediatric studies comparing 0.9% saline with balanced solutions are needed.

When Should We Use Hypotonic IV Fluids?

Hypotonic IV fluids may be needed for patients with hypernatremia and a free-water deficit or a renal-concentrating defect with ongoing urinary free-water losses.² Special care should be taken when choosing maintenance IV fluids for patients with renal disease, liver disease, or heart failure given that these groups have been excluded from some studies.¹² These patients may be at risk for increased salt and fluid retention with any IV fluid, and fluid rates need to be restricted. The fluid intake of patients with hyponatremia secondary to SIADH needs close management; these patients benefit from total fluid restriction instead of standard maintenance IV fluid rates.²

What We Should Do Instead?

Maintenance IV fluids should only be used when necessary and should be stopped as soon as they are no longer required, especially in light of the recent shortages in 0.9% saline.¹⁷ Similar to all medications, maintenance IV fluids should be individualized to the patient’s needs on the basis of the indication for IV fluids and the patient’s comorbidities.² Consideration should be given to checking the patient’s electrolyte levels to monitor response to IV fluids, especially during the first 24 hours of admission when risk of hyponatremia is highest. Isotonic IV fluids with 5% dextrose should be used as the maintenance IV fluid in the majority of hospitalized children given its proven benefit in decreasing the rate of hospital-acquired hyponatremia.^7,8 Hypotonic IV fluids should be avoided as the default maintenance IV fluid and should only be utilized under specific circumstances.

• When needed, maintenance IV fluids should always be tailored to each individual patient.
• For most acutely ill hospitalized children, isotonic IV fluids should be the maintenance IV fluid of choice.
• Consider monitoring electrolytes to determine the effects of maintenance IV fluids.

Enteral maintenance fluids should be used first-line if possible. Although hypotonic IV fluids have historically been the maintenance IV fluid of choice, this class of IV fluids should be avoided for most hospitalized children to decrease the significant risk of iatrogenic hyponatremia, which can be severe and have catastrophic complications. When necessary, isotonic IV fluids should be used for the majority of hospitalized children given that these fluids present a significantly decreased risk for causing hyponatremia. Returning to our case presentation, to decrease the risk of hyponatremia, the senior resident should recommend starting isotonic IV fluids in the 12-month-old and theoretical 2-week-old until oral intake can be maintained.

Do you think this is a low-value practice? Is this truly a “Thing We Do for No Reason”? Let us know what you do in your practice and propose ideas for other “Things We Do for No Reason” topics. Please join in the conversation online at Twitter (#TWDFNR)/Facebook and don’t forget to “Like It” on Facebook or retweet it on Twitter.

Disclosure

The authors have no relevant conflicts of interest to report. No payment or services from a 3rd party were received for any aspect of this submitted work. The authors have no financial relationships with entities in the biomedical arena that could be perceived to influence, or that give the appearance of potentially influencing, what was written in this submitted work.

The “Things We Do for No Reason” series reviews practices which have become common parts of hospital care but which may provide little value to our patients. Practices reviewed in the TWDFR series do not represent “black and white” conclusions or clinical practice standards, but are meant as a starting place for research and active discussions among hospitalists and patients. We invite you to be part of that discussion. https://www.choosingwisely.org/

A 67-year-old man with acute myeloid leukemia who has recently completed a cycle of consolidation chemotherapy presents to the emergency room with fatigue and bruising. He is found to have pancytopenia due to chemotherapy. His absolute neutrophil count (ANC) is 380/mm³,and he has no symptoms or signs of infection. He is admitted for transfusion support and asks for a dinner tray. The provider reflexively prescribes a neutropenic diet.

Although aggressive chemotherapy regimens have significantly improved survival rates in patients with cancer, these intensive regimens put patients at risk for a number of complications, including severe, prolonged neutropenia. Patients with neutropenia, particularly those with ANC< 500/mm³, are at a significantly increased risk for infection. Common sites of infection include the blood stream, skin, lungs, urinary tract, and, particularly, the gastrointestinal tract.¹ Oncologists and dieticians first designed neutropenic diets, or low-bacteria diets, to limit the introduction of pathogenic microbes to the gastrointestinal system. Neutropenic diets typically limit the intake of fresh fruits, fresh vegetables, raw or undercooked meats and fish, and soft cheese made from unpasteurized milk. Despite the widespread recommendation of the neutropenic diet, no standardized guidelines exist, and the utilization of the diet and its contents vary widely among and within institutions.²

The neutropenic diet is a national phenomenon. A survey of 156 United States members of the Association of Community Cancer Centers revealed that 120 (78%) of the members had placed patients with neutropenia on restricted diets.² The triggers for prescription (neutropenia, or starting chemotherapy), ANC threshold for prescription, and duration of prescription (throughout chemotherapy or just when neutropenic) were not uniform. A majority of centers restricted fresh fruits, fresh vegetables, and raw eggs, while some locations also restricted tap water, herbs and spices, and alcoholic beverages.² Similarly, a study of practices in 29 countries across 6 continents found that 88% of centers have some version of a neutropenic diet guideline with significant heterogeneity in their prescription and content. For example, dried fruits were unrestricted in 23% of centers but were forbidden in 43%.³

The rationale behind the neutropenic diet is to limit the bacterial load delivered to the gut. Studies have shown that organisms such as Enterobacter, Pseudomonas, and Klebsiella have been isolated from food, particularly fruits and vegetables.^4,5 The ingestion of contaminated food products may serve as a source of pathogenic bacteria, which may cause potentially life-threatening infections. Mucositis, a common complication among cancer patients receiving therapy, predisposes patients to infection by disrupting the mucosal barrier, allowing bacteria to translocate from the gut to the bloodstream. Given that neutropenia and mucositis often occur simultaneously, these patients are at an increased risk of infections.⁶ Cooking destroys bacteria if present, rendering cooked foods safe. Thus, the avoidance of fresh fruits and vegetables and other foods considered to have high bacterial loads should theoretically decrease the risk of infections in these patients.

Researchers have investigated the ability of the neutropenic diet to reduce infection in adult and pediatric neutropenic patients. A study involving 153 patients receiving chemotherapy for acute myeloid leukemia or myelodysplastic syndrome randomized 78 patients to a diet that restricted raw fruits and vegetables and 75 patients to a diet that included those foods.⁸ The groups had similar rates of major infection (29% in the cooked group versus 35% in the raw group, P = .60) with no difference in mortality.⁷ In a randomized, multiinstitutional trial of 150 pediatric oncology patients, 77 patients received a neutropenic diet plus a diet based on the food safety guidelines approved by the Food and Drug Administration (FDA), while 73 children received a diet based on FDA-approved food safety guidelines.⁸ Infection rates between the groups were not significantly different (35% vs 33% respectively, P = .78).

Intensive conditioning regimens place hematopoietic stem-cell transplant (HSCT) recipients at an even greater risk of infectious complications than other patients and may increase gastrointestinal toxicity and prolong neutropenia. A study from a single academic US center included 726 HSCT recipients, 363 of whom received a neutropenic diet and 363 of whom received a general diet. Significantly fewer infections were observed in the general diet group than in the neutropenic diet group. Notably, this study was a retrospective trial, and approximately 75% of participants were autologous HSCT recipients, who traditionally have low risks of infection. A survey and analysis of nonpharmacologic anti-infective measures in 339 children with leukemia enrolled in the multicenter Acute Myeloid Leukemia Berlin-Frankfurt-Munster 2004 trial also did not show that the neutropenic diet has protective effects on infection rates.⁹ A metaanalysis that compiled data from the studies mentioned above found the hazard ratio for any infection (major or minor) and fever was actually higher in the neutropenic diet arm (relative risk 1.18, 95% confidence interval: 1.05-1.34, P = .007) relative to that in the unrestricted arm.¹⁰

The inefficacy of the neutropenic diet may be attributed to the fact that many of the organisms found on fresh fruits and vegetables are part of the normal flora in the gastrointestinal tract. A Dutch prospective randomized pilot study of 20 adult patients with acute myeloid leukemia undergoing chemotherapy compared the gut flora in patients on a low-bacteria diet versus that in patients on a normal hospital diet. Gut colonization by potential pathogens or infection rates were not significantly different between the 2 groups.¹¹

In addition to mucositis, the common gastrointestinal complications of chemotherapy include nausea, vomiting, diarrhea, food aversions, and changes in smells and taste, which limit oral intake.¹² Unnecessary dietary restrictions can place patients at further risk of inadequate intake and malnutrition.¹³ In the outpatient setting, compliance with the neutropenic diet is also problematic. In 1 study of 28 patients educated about the neutropenic diet, only 16 (57%) were compliant with the diet as revealed through telephone-based assessments at 6 and 12 weeks, and infection rates were not different between compliant versus noncompliant patients.¹⁴ Patients and family members reported that following the neutropenic diet requires considerably more effort than following a less restrictive diet.⁸ Maintaining nutrition in this patient population is already challenging, and the restriction of a wide variety of food items (fresh fruits, vegetables, dairy, certain meats, eggs) can cause malnutrition, low patient satisfaction, and poor quality of life.^13,14

Evidence shows no benefit of the neutropenic diet in any particular clinical scenario or patient population. However, despite the dearth of evidence to support neutropenic diets, the overall data regarding neutropenic diets are sparse. Randomized control trials to date have been limited by their small size with possible confounding by the type of malignancy and cancer therapy; use of prophylactic antibiotics, growth factors, and air-filtered rooms; variation in contents and adherence to the prescribed diet; and inpatient versus outpatient status. The study that included HSCT recipients was a retrospective trial, and a majority of patients were autologous HSCT recipients.¹⁵ Although no study has specifically investigated the neutropenic diet in preventing infection in patients with noncancer-related neutropenia, no reason exists to suspect that it is helpful. The FDA advises safe food-handling practices for other immunocompromised patients, such as transplant recipients and patients with human immunodeficiency virus/acquired immunodeficiency syndrome, and the same principles can likely be applied to patients with noncancer-related neutropenia.

WHAT WE SHOULD DO INSTEAD

Although the neutropenic diet has not been proven beneficial, the prevention of food-borne infection in this population remains important. FDA-published guidelines, which promote safe food handling to prevent food contamination in patients with cancer, should be followed in inpatient and outpatient settings.¹⁶ These guidelines allow for fresh fruits and vegetables as long as they have been adequately washed. Cleaning (eg, cleaning the lids of canned foods before opening, hand washing), separating raw meats from other foods, cooking to the right temperature (eg, cooking eggs until the yolk and white are firm), and chilling/refrigerating food appropriately are strongly emphasized. These guidelines are also recommended by the American Dietetic Association. Despite additional flexibility, patients following the FDA diet guidelines do not have increased risk of infection.⁸ At our hospitals, the neutropenic diet can no longer be ordered. Neutropenic patients are free to consume all items on the general hospital menu, including eggs, meat, soft cheeses, nuts, and washed raw fruits and vegetables. The National Comprehensive Cancer Network guidelines for the prevention and treatment of cancer-related infections do not specifically address diet.¹⁷ We call upon them to note the lack of benefit and potential harm of the neutropenic diet in the guidelines. Such an action may persuade more institutions to abandon this practice.

RECOMMENDATIONS

• Neutropenic diets, or low-bacteria diets, should not be prescribed to neutropenic patients.
• Properly handled and adequately washed fresh fruits and vegetables can safely be consumed by patients with neutropenia.
• Patients and hospitals should follow FDA-published safe food-handling guidelines to prevent food contamination.

A general diet can be safely ordered for our patient in the presented clinical scenario. Available data from individual studies and pooled data provide no evidence that neutropenic diets prevent infectious complications in patients with neutropenia.

Hospital kitchens must adhere to the food-handling guidelines issued by the FDA, and following these guidelines should provide adequate protection against food-borne infection, even in patients who are immunocompromised. Instead of restricting food groups, the FDA guidelines focus on safe food-handling practices. Less dietary restrictions provide patient’s additional opportunities for balanced nutrition and for food choices based on personal preferences or cultural practices.

Do you think this is a low-value practice? Is this truly a “Thing We Do for No Reason?” Share what you do in your practice and join in the conversation online by retweeting it on Twitter (#TWDFNR) and liking it on Facebook. We invite you to propose ideas for other “Things We Do for No Reason” topics by emailing TWDFNR@hospitalmedicine.org.Disclosures: There are no financial or other disclosures for any author.

Disclosures

There are no financial or other disclosures for any author.

The “Things We Do for No Reason” series reviews practices which have become common parts of hospital care but which may provide little value to our patients. Practices reviewed in the TWDFR series do not represent “black and white” conclusions or clinical practice standards, but are meant as a starting place for research and active discussions among hospitalists and patients. We invite you to be part of that discussion. https://www.choosingwisely.org/

A 67-year-old man with acute myeloid leukemia who has recently completed a cycle of consolidation chemotherapy presents to the emergency room with fatigue and bruising. He is found to have pancytopenia due to chemotherapy. His absolute neutrophil count (ANC) is 380/mm³,and he has no symptoms or signs of infection. He is admitted for transfusion support and asks for a dinner tray. The provider reflexively prescribes a neutropenic diet.

Although aggressive chemotherapy regimens have significantly improved survival rates in patients with cancer, these intensive regimens put patients at risk for a number of complications, including severe, prolonged neutropenia. Patients with neutropenia, particularly those with ANC< 500/mm³, are at a significantly increased risk for infection. Common sites of infection include the blood stream, skin, lungs, urinary tract, and, particularly, the gastrointestinal tract.¹ Oncologists and dieticians first designed neutropenic diets, or low-bacteria diets, to limit the introduction of pathogenic microbes to the gastrointestinal system. Neutropenic diets typically limit the intake of fresh fruits, fresh vegetables, raw or undercooked meats and fish, and soft cheese made from unpasteurized milk. Despite the widespread recommendation of the neutropenic diet, no standardized guidelines exist, and the utilization of the diet and its contents vary widely among and within institutions.²

The neutropenic diet is a national phenomenon. A survey of 156 United States members of the Association of Community Cancer Centers revealed that 120 (78%) of the members had placed patients with neutropenia on restricted diets.² The triggers for prescription (neutropenia, or starting chemotherapy), ANC threshold for prescription, and duration of prescription (throughout chemotherapy or just when neutropenic) were not uniform. A majority of centers restricted fresh fruits, fresh vegetables, and raw eggs, while some locations also restricted tap water, herbs and spices, and alcoholic beverages.² Similarly, a study of practices in 29 countries across 6 continents found that 88% of centers have some version of a neutropenic diet guideline with significant heterogeneity in their prescription and content. For example, dried fruits were unrestricted in 23% of centers but were forbidden in 43%.³

The rationale behind the neutropenic diet is to limit the bacterial load delivered to the gut. Studies have shown that organisms such as Enterobacter, Pseudomonas, and Klebsiella have been isolated from food, particularly fruits and vegetables.^4,5 The ingestion of contaminated food products may serve as a source of pathogenic bacteria, which may cause potentially life-threatening infections. Mucositis, a common complication among cancer patients receiving therapy, predisposes patients to infection by disrupting the mucosal barrier, allowing bacteria to translocate from the gut to the bloodstream. Given that neutropenia and mucositis often occur simultaneously, these patients are at an increased risk of infections.⁶ Cooking destroys bacteria if present, rendering cooked foods safe. Thus, the avoidance of fresh fruits and vegetables and other foods considered to have high bacterial loads should theoretically decrease the risk of infections in these patients.

Researchers have investigated the ability of the neutropenic diet to reduce infection in adult and pediatric neutropenic patients. A study involving 153 patients receiving chemotherapy for acute myeloid leukemia or myelodysplastic syndrome randomized 78 patients to a diet that restricted raw fruits and vegetables and 75 patients to a diet that included those foods.⁸ The groups had similar rates of major infection (29% in the cooked group versus 35% in the raw group, P = .60) with no difference in mortality.⁷ In a randomized, multiinstitutional trial of 150 pediatric oncology patients, 77 patients received a neutropenic diet plus a diet based on the food safety guidelines approved by the Food and Drug Administration (FDA), while 73 children received a diet based on FDA-approved food safety guidelines.⁸ Infection rates between the groups were not significantly different (35% vs 33% respectively, P = .78).

Intensive conditioning regimens place hematopoietic stem-cell transplant (HSCT) recipients at an even greater risk of infectious complications than other patients and may increase gastrointestinal toxicity and prolong neutropenia. A study from a single academic US center included 726 HSCT recipients, 363 of whom received a neutropenic diet and 363 of whom received a general diet. Significantly fewer infections were observed in the general diet group than in the neutropenic diet group. Notably, this study was a retrospective trial, and approximately 75% of participants were autologous HSCT recipients, who traditionally have low risks of infection. A survey and analysis of nonpharmacologic anti-infective measures in 339 children with leukemia enrolled in the multicenter Acute Myeloid Leukemia Berlin-Frankfurt-Munster 2004 trial also did not show that the neutropenic diet has protective effects on infection rates.⁹ A metaanalysis that compiled data from the studies mentioned above found the hazard ratio for any infection (major or minor) and fever was actually higher in the neutropenic diet arm (relative risk 1.18, 95% confidence interval: 1.05-1.34, P = .007) relative to that in the unrestricted arm.¹⁰

The inefficacy of the neutropenic diet may be attributed to the fact that many of the organisms found on fresh fruits and vegetables are part of the normal flora in the gastrointestinal tract. A Dutch prospective randomized pilot study of 20 adult patients with acute myeloid leukemia undergoing chemotherapy compared the gut flora in patients on a low-bacteria diet versus that in patients on a normal hospital diet. Gut colonization by potential pathogens or infection rates were not significantly different between the 2 groups.¹¹

In addition to mucositis, the common gastrointestinal complications of chemotherapy include nausea, vomiting, diarrhea, food aversions, and changes in smells and taste, which limit oral intake.¹² Unnecessary dietary restrictions can place patients at further risk of inadequate intake and malnutrition.¹³ In the outpatient setting, compliance with the neutropenic diet is also problematic. In 1 study of 28 patients educated about the neutropenic diet, only 16 (57%) were compliant with the diet as revealed through telephone-based assessments at 6 and 12 weeks, and infection rates were not different between compliant versus noncompliant patients.¹⁴ Patients and family members reported that following the neutropenic diet requires considerably more effort than following a less restrictive diet.⁸ Maintaining nutrition in this patient population is already challenging, and the restriction of a wide variety of food items (fresh fruits, vegetables, dairy, certain meats, eggs) can cause malnutrition, low patient satisfaction, and poor quality of life.^13,14

Evidence shows no benefit of the neutropenic diet in any particular clinical scenario or patient population. However, despite the dearth of evidence to support neutropenic diets, the overall data regarding neutropenic diets are sparse. Randomized control trials to date have been limited by their small size with possible confounding by the type of malignancy and cancer therapy; use of prophylactic antibiotics, growth factors, and air-filtered rooms; variation in contents and adherence to the prescribed diet; and inpatient versus outpatient status. The study that included HSCT recipients was a retrospective trial, and a majority of patients were autologous HSCT recipients.¹⁵ Although no study has specifically investigated the neutropenic diet in preventing infection in patients with noncancer-related neutropenia, no reason exists to suspect that it is helpful. The FDA advises safe food-handling practices for other immunocompromised patients, such as transplant recipients and patients with human immunodeficiency virus/acquired immunodeficiency syndrome, and the same principles can likely be applied to patients with noncancer-related neutropenia.

WHAT WE SHOULD DO INSTEAD

Although the neutropenic diet has not been proven beneficial, the prevention of food-borne infection in this population remains important. FDA-published guidelines, which promote safe food handling to prevent food contamination in patients with cancer, should be followed in inpatient and outpatient settings.¹⁶ These guidelines allow for fresh fruits and vegetables as long as they have been adequately washed. Cleaning (eg, cleaning the lids of canned foods before opening, hand washing), separating raw meats from other foods, cooking to the right temperature (eg, cooking eggs until the yolk and white are firm), and chilling/refrigerating food appropriately are strongly emphasized. These guidelines are also recommended by the American Dietetic Association. Despite additional flexibility, patients following the FDA diet guidelines do not have increased risk of infection.⁸ At our hospitals, the neutropenic diet can no longer be ordered. Neutropenic patients are free to consume all items on the general hospital menu, including eggs, meat, soft cheeses, nuts, and washed raw fruits and vegetables. The National Comprehensive Cancer Network guidelines for the prevention and treatment of cancer-related infections do not specifically address diet.¹⁷ We call upon them to note the lack of benefit and potential harm of the neutropenic diet in the guidelines. Such an action may persuade more institutions to abandon this practice.

RECOMMENDATIONS

• Neutropenic diets, or low-bacteria diets, should not be prescribed to neutropenic patients.
• Properly handled and adequately washed fresh fruits and vegetables can safely be consumed by patients with neutropenia.
• Patients and hospitals should follow FDA-published safe food-handling guidelines to prevent food contamination.

A general diet can be safely ordered for our patient in the presented clinical scenario. Available data from individual studies and pooled data provide no evidence that neutropenic diets prevent infectious complications in patients with neutropenia.

Hospital kitchens must adhere to the food-handling guidelines issued by the FDA, and following these guidelines should provide adequate protection against food-borne infection, even in patients who are immunocompromised. Instead of restricting food groups, the FDA guidelines focus on safe food-handling practices. Less dietary restrictions provide patient’s additional opportunities for balanced nutrition and for food choices based on personal preferences or cultural practices.

Do you think this is a low-value practice? Is this truly a “Thing We Do for No Reason?” Share what you do in your practice and join in the conversation online by retweeting it on Twitter (#TWDFNR) and liking it on Facebook. We invite you to propose ideas for other “Things We Do for No Reason” topics by emailing TWDFNR@hospitalmedicine.org.Disclosures: There are no financial or other disclosures for any author.

Disclosures

There are no financial or other disclosures for any author.

The “Things We Do for No Reason” series reviews practices which have become common parts of hospital care but which may provide little value to our patients. Practices reviewed in the TWDFR series do not represent “black and white” conclusions or clinical practice standards, but are meant as a starting place for research and active discussions among hospitalists and patients. We invite you to be part of that discussion. https://www.choosingwisely.org/

A 67-year-old man with acute myeloid leukemia who has recently completed a cycle of consolidation chemotherapy presents to the emergency room with fatigue and bruising. He is found to have pancytopenia due to chemotherapy. His absolute neutrophil count (ANC) is 380/mm³,and he has no symptoms or signs of infection. He is admitted for transfusion support and asks for a dinner tray. The provider reflexively prescribes a neutropenic diet.

Although aggressive chemotherapy regimens have significantly improved survival rates in patients with cancer, these intensive regimens put patients at risk for a number of complications, including severe, prolonged neutropenia. Patients with neutropenia, particularly those with ANC< 500/mm³, are at a significantly increased risk for infection. Common sites of infection include the blood stream, skin, lungs, urinary tract, and, particularly, the gastrointestinal tract.¹ Oncologists and dieticians first designed neutropenic diets, or low-bacteria diets, to limit the introduction of pathogenic microbes to the gastrointestinal system. Neutropenic diets typically limit the intake of fresh fruits, fresh vegetables, raw or undercooked meats and fish, and soft cheese made from unpasteurized milk. Despite the widespread recommendation of the neutropenic diet, no standardized guidelines exist, and the utilization of the diet and its contents vary widely among and within institutions.²

The neutropenic diet is a national phenomenon. A survey of 156 United States members of the Association of Community Cancer Centers revealed that 120 (78%) of the members had placed patients with neutropenia on restricted diets.² The triggers for prescription (neutropenia, or starting chemotherapy), ANC threshold for prescription, and duration of prescription (throughout chemotherapy or just when neutropenic) were not uniform. A majority of centers restricted fresh fruits, fresh vegetables, and raw eggs, while some locations also restricted tap water, herbs and spices, and alcoholic beverages.² Similarly, a study of practices in 29 countries across 6 continents found that 88% of centers have some version of a neutropenic diet guideline with significant heterogeneity in their prescription and content. For example, dried fruits were unrestricted in 23% of centers but were forbidden in 43%.³

The rationale behind the neutropenic diet is to limit the bacterial load delivered to the gut. Studies have shown that organisms such as Enterobacter, Pseudomonas, and Klebsiella have been isolated from food, particularly fruits and vegetables.^4,5 The ingestion of contaminated food products may serve as a source of pathogenic bacteria, which may cause potentially life-threatening infections. Mucositis, a common complication among cancer patients receiving therapy, predisposes patients to infection by disrupting the mucosal barrier, allowing bacteria to translocate from the gut to the bloodstream. Given that neutropenia and mucositis often occur simultaneously, these patients are at an increased risk of infections.⁶ Cooking destroys bacteria if present, rendering cooked foods safe. Thus, the avoidance of fresh fruits and vegetables and other foods considered to have high bacterial loads should theoretically decrease the risk of infections in these patients.

Researchers have investigated the ability of the neutropenic diet to reduce infection in adult and pediatric neutropenic patients. A study involving 153 patients receiving chemotherapy for acute myeloid leukemia or myelodysplastic syndrome randomized 78 patients to a diet that restricted raw fruits and vegetables and 75 patients to a diet that included those foods.⁸ The groups had similar rates of major infection (29% in the cooked group versus 35% in the raw group, P = .60) with no difference in mortality.⁷ In a randomized, multiinstitutional trial of 150 pediatric oncology patients, 77 patients received a neutropenic diet plus a diet based on the food safety guidelines approved by the Food and Drug Administration (FDA), while 73 children received a diet based on FDA-approved food safety guidelines.⁸ Infection rates between the groups were not significantly different (35% vs 33% respectively, P = .78).

Intensive conditioning regimens place hematopoietic stem-cell transplant (HSCT) recipients at an even greater risk of infectious complications than other patients and may increase gastrointestinal toxicity and prolong neutropenia. A study from a single academic US center included 726 HSCT recipients, 363 of whom received a neutropenic diet and 363 of whom received a general diet. Significantly fewer infections were observed in the general diet group than in the neutropenic diet group. Notably, this study was a retrospective trial, and approximately 75% of participants were autologous HSCT recipients, who traditionally have low risks of infection. A survey and analysis of nonpharmacologic anti-infective measures in 339 children with leukemia enrolled in the multicenter Acute Myeloid Leukemia Berlin-Frankfurt-Munster 2004 trial also did not show that the neutropenic diet has protective effects on infection rates.⁹ A metaanalysis that compiled data from the studies mentioned above found the hazard ratio for any infection (major or minor) and fever was actually higher in the neutropenic diet arm (relative risk 1.18, 95% confidence interval: 1.05-1.34, P = .007) relative to that in the unrestricted arm.¹⁰

The inefficacy of the neutropenic diet may be attributed to the fact that many of the organisms found on fresh fruits and vegetables are part of the normal flora in the gastrointestinal tract. A Dutch prospective randomized pilot study of 20 adult patients with acute myeloid leukemia undergoing chemotherapy compared the gut flora in patients on a low-bacteria diet versus that in patients on a normal hospital diet. Gut colonization by potential pathogens or infection rates were not significantly different between the 2 groups.¹¹

In addition to mucositis, the common gastrointestinal complications of chemotherapy include nausea, vomiting, diarrhea, food aversions, and changes in smells and taste, which limit oral intake.¹² Unnecessary dietary restrictions can place patients at further risk of inadequate intake and malnutrition.¹³ In the outpatient setting, compliance with the neutropenic diet is also problematic. In 1 study of 28 patients educated about the neutropenic diet, only 16 (57%) were compliant with the diet as revealed through telephone-based assessments at 6 and 12 weeks, and infection rates were not different between compliant versus noncompliant patients.¹⁴ Patients and family members reported that following the neutropenic diet requires considerably more effort than following a less restrictive diet.⁸ Maintaining nutrition in this patient population is already challenging, and the restriction of a wide variety of food items (fresh fruits, vegetables, dairy, certain meats, eggs) can cause malnutrition, low patient satisfaction, and poor quality of life.^13,14

Evidence shows no benefit of the neutropenic diet in any particular clinical scenario or patient population. However, despite the dearth of evidence to support neutropenic diets, the overall data regarding neutropenic diets are sparse. Randomized control trials to date have been limited by their small size with possible confounding by the type of malignancy and cancer therapy; use of prophylactic antibiotics, growth factors, and air-filtered rooms; variation in contents and adherence to the prescribed diet; and inpatient versus outpatient status. The study that included HSCT recipients was a retrospective trial, and a majority of patients were autologous HSCT recipients.¹⁵ Although no study has specifically investigated the neutropenic diet in preventing infection in patients with noncancer-related neutropenia, no reason exists to suspect that it is helpful. The FDA advises safe food-handling practices for other immunocompromised patients, such as transplant recipients and patients with human immunodeficiency virus/acquired immunodeficiency syndrome, and the same principles can likely be applied to patients with noncancer-related neutropenia.

WHAT WE SHOULD DO INSTEAD

Although the neutropenic diet has not been proven beneficial, the prevention of food-borne infection in this population remains important. FDA-published guidelines, which promote safe food handling to prevent food contamination in patients with cancer, should be followed in inpatient and outpatient settings.¹⁶ These guidelines allow for fresh fruits and vegetables as long as they have been adequately washed. Cleaning (eg, cleaning the lids of canned foods before opening, hand washing), separating raw meats from other foods, cooking to the right temperature (eg, cooking eggs until the yolk and white are firm), and chilling/refrigerating food appropriately are strongly emphasized. These guidelines are also recommended by the American Dietetic Association. Despite additional flexibility, patients following the FDA diet guidelines do not have increased risk of infection.⁸ At our hospitals, the neutropenic diet can no longer be ordered. Neutropenic patients are free to consume all items on the general hospital menu, including eggs, meat, soft cheeses, nuts, and washed raw fruits and vegetables. The National Comprehensive Cancer Network guidelines for the prevention and treatment of cancer-related infections do not specifically address diet.¹⁷ We call upon them to note the lack of benefit and potential harm of the neutropenic diet in the guidelines. Such an action may persuade more institutions to abandon this practice.

RECOMMENDATIONS

• Neutropenic diets, or low-bacteria diets, should not be prescribed to neutropenic patients.
• Properly handled and adequately washed fresh fruits and vegetables can safely be consumed by patients with neutropenia.
• Patients and hospitals should follow FDA-published safe food-handling guidelines to prevent food contamination.

A general diet can be safely ordered for our patient in the presented clinical scenario. Available data from individual studies and pooled data provide no evidence that neutropenic diets prevent infectious complications in patients with neutropenia.

Hospital kitchens must adhere to the food-handling guidelines issued by the FDA, and following these guidelines should provide adequate protection against food-borne infection, even in patients who are immunocompromised. Instead of restricting food groups, the FDA guidelines focus on safe food-handling practices. Less dietary restrictions provide patient’s additional opportunities for balanced nutrition and for food choices based on personal preferences or cultural practices.

Do you think this is a low-value practice? Is this truly a “Thing We Do for No Reason?” Share what you do in your practice and join in the conversation online by retweeting it on Twitter (#TWDFNR) and liking it on Facebook. We invite you to propose ideas for other “Things We Do for No Reason” topics by emailing TWDFNR@hospitalmedicine.org.Disclosures: There are no financial or other disclosures for any author.

Disclosures

There are no financial or other disclosures for any author.

The “Things We Do for No Reason” (TWDFNR) series reviews practices that have become common parts of hospital care but may provide little value to our patients. Practices reviewed in the TWDFNR series do not represent “black and white” conclusions or clinical practice standards but are meant as a starting place for research and active discussions among hospitalists and patients. We invite you to be part of that discussion. https://www.choosingwisely.org/

An 8-year-old previously healthy girl presented to the emergency department (ED) with 2 days of warmth, swelling, and pain over her right upper thigh. Three days prior before presentation, a “pimple” appeared on her leg and drained a small amount of pus. Over the next 24 hours, the lesion became swollen, red, and painful. Her pediatrician prescribed trimethoprim-sulfamethoxazole. The patient took 3 doses of this medication but still experienced worsening pain and swelling.

In the ED, she had normal vital signs for her age except for temperature of 100.8 °F. A 2 cm × 3 cm area of fluctuance, erythema, and warmth was noted, and bedside ultrasound demonstrated a simple fluid collection. Incision and drainage was performed with expression of several milliliters of pus. The patient was referred for admission due to worsening symptoms despite outpatient antibiotic therapy. The ED providers ordered a blood culture at the time of admission.

Skin and soft tissue infections (SSTIs) are common pediatric diagnoses, which account for an estimated 390,000 ED visits annually¹ and represent the 7^th most common reason for pediatric hospital admission in the United States.² The rates of SSTIs have increased over the past several decades partly due to the rise of methicillin-resistant Staphylococcus aureus (MRSA).³

Why You Might Think Blood Cultures are Helpful In Children with SSTIs?

Prior to the introduction of the Haemophilus influenzae vaccine, the rates of SSTI-associated bacteremia ranged from 8% to 20%.^4,5 Although the rate of bacteremia has declined significantly, blood cultures are still commonly performed as part of the evaluation of uncomplicated SSTIs in children; studies have shown that blood culture rates are 46% in the combined outpatient/inpatient setting,⁶ 34% in the ED setting,⁷ and 47%-94% in the inpatient setting.^7-11 Clinicians still feel that bacteremia detection is important to guide the selection of antibiotics and treatment duration. Providers may also underestimate the risk of obtaining a contaminant result and associated charges. Lastly, clinicians may perform blood cultures due to cultural norms at their institution.

Why Blood Cultures are Unnecessary in Children with Uncomplicated SSTIs

Several decades into the post vaccine era, the current guidelines from the Infectious Diseases Society of America (IDSA) do not recommend blood cultures as part of the routine evaluation of uncomplicated SSTIs.¹² Multiple single-center studies have failed to demonstrate the benefits of obtaining blood cultures in pediatric patients with uncomplicated SSTIs in the post-H. influenzae vaccine era.^6–11

Sadow et al¹¹ performed a retrospective case series of 381 children hospitalized with cellulitis to determine the rate and yield of blood cultures. Of the 266 (70%) patients who had a blood culture performed, 5 (1.9%) were true positives and 13 (5.4%) were contaminants. Notably, the true positive results included 3 children with active varicella infection and 2 children with septic joints; the latter would qualify as a complicated SSTI or as a separate infectious process altogether. No significant change in management resulted the positive blood cultures.

Harms Associated With Unnecessary Blood Cultures in SSTIs

Blood cultures necessitate venipunctures, which are painful for children and families. The inevitable false-positive contaminants also lead to repeat venipunctures and, potentially, unnecessary antibiotic exposure. From a high-value care perspective, Parikh et al reported hospital charges of $300 per blood culture and $250 for identification and sensitivity of positives.⁶ Assuming that these single-center charges are representative of national charges and using 0.5% true positivity and 1.5% false positivity rates, subjecting all children with uncomplicated SSTIs to blood culture would result in $60,250 charges to find one true positive blood culture, with no resultant changes in management. Additionally, among the 200 children cultured to find one true positive, there would be 3 false positives, necessitating another $1650 in charges for identification, sensitivity analysis, and repeat culture. These amounts do not factor in the significant expenditures associated with increased LOS. The potential savings associated with forgoing blood cultures in children with SSTIs should be an incentive for institutional change.

When Blood Cultures May Be Reasonable

The current IDSA guidelines recommend blood cultures for SSTIs in patients with immunodeficiency, animal bites, and immersion injuries (soft tissue injuries occurring in fresh or saltwater).¹² Previous studies also delineated criteria for “complicated” SSTIs, typically defined as surgical or traumatic wounds, infections requiring surgical intervention (not including simple incision and drainage), or infected ulcers or burns.^8,9 In the study of Malone et al, 10 (12.5%) positives were found among 80 patients with complicated SSTIs who had blood cultures performed.⁸ Although this work had a single-center study design with a relatively small sample size, no unusual organisms were found; the grown cultures included MRSA, MSSA, and S. pneumoniae. In addition to patients with complicated SSTIs, immunocompromised children, such as those receiving chemotherapy or other immunosuppressive agents, were excluded from the studies of blood culture yield in SSTIs and may warrant blood cultures given the risk of overwhelming infection and susceptibility to rare or invasive organisms.¹² In a study of 57 pediatric patients with leukemia and no central catheters who experienced skin or soft tissue complications, Demircioglu et al¹³ reported 6 positive blood cultures, including Klebsiella oxytoca, Pseudomonas aeruginosa, and Escherichia coli. These organisms would not be covered by typical SSTI antibiotic regimens, illustrating the value of blood cultures in this selected group of patients. Lastly, although the above studies included some infants, the data on utility of blood cultures in neonates are limited. Blood cultures may be reasonable in this group given the relative immunocompromised state of neonates compared with older children. Additionally, any infants aged

What You Should Do Instead Of Blood Cultures for Uncomplicated SSTIs

Gram stain and wound culture of any purulent material may assist with choice of empiric antibiotic therapy and appropriate narrowing of regimen for antibiotic stewardship. Wound cultures of purulent material can identify the causative organism in 58%-66% of the cases.^9,14 The rate of wound culture varies widely from 29% to 81% in studies across different healthcare systems.^9,10,15 The use of visually appealing posters advising clinicians to “culture pus, not blood” has been shown to significantly decreased the number of blood cultures performed at a single pediatric hospital.¹⁰

• Do not obtain blood cultures in pediatric patients with uncomplicated SSTIs.
• If purulent material is available spontaneously or after incision and drainage, then send it for Gram stain and bacterial culture.
• Blood cultures are reasonable in patients with complicated SSTIs and in immunocompromised patients with SSTIs.
• Despite limited data, blood cultures may be reasonable in neonates with SSTIs. Febrile infants with SSTIs aged less than 90 days should be managed under existing febrile infant guidelines.

Blood cultures in pediatric patients with uncomplicated SSTIs have no proven benefit and are associated with increased LOS, non-negligible false-positive rate, and associated increase in financial charges to the patient and healthcare system. The patient described in the clinical scenario would have an extremely low likelihood of having any meaningful clinical information provided by blood culture as part of her evaluation.

Do you think this is a low-value practice? Is this truly a “Thing We Do for No Reason?” Share what you do in your practice and join in the conversation online by retweeting it on Twitter (#TWDFNR) and liking it on Facebook. We invite you to propose ideas for other “Things We Do for No Reason” topics by emailing TWDFNR@hospitalmedicine.org.

DISCLOSURES

The authors have no conflicts of interest relevant to this article to disclose.

The “Things We Do for No Reason” (TWDFNR) series reviews practices that have become common parts of hospital care but may provide little value to our patients. Practices reviewed in the TWDFNR series do not represent “black and white” conclusions or clinical practice standards but are meant as a starting place for research and active discussions among hospitalists and patients. We invite you to be part of that discussion. https://www.choosingwisely.org/

An 8-year-old previously healthy girl presented to the emergency department (ED) with 2 days of warmth, swelling, and pain over her right upper thigh. Three days prior before presentation, a “pimple” appeared on her leg and drained a small amount of pus. Over the next 24 hours, the lesion became swollen, red, and painful. Her pediatrician prescribed trimethoprim-sulfamethoxazole. The patient took 3 doses of this medication but still experienced worsening pain and swelling.

In the ED, she had normal vital signs for her age except for temperature of 100.8 °F. A 2 cm × 3 cm area of fluctuance, erythema, and warmth was noted, and bedside ultrasound demonstrated a simple fluid collection. Incision and drainage was performed with expression of several milliliters of pus. The patient was referred for admission due to worsening symptoms despite outpatient antibiotic therapy. The ED providers ordered a blood culture at the time of admission.

Skin and soft tissue infections (SSTIs) are common pediatric diagnoses, which account for an estimated 390,000 ED visits annually¹ and represent the 7^th most common reason for pediatric hospital admission in the United States.² The rates of SSTIs have increased over the past several decades partly due to the rise of methicillin-resistant Staphylococcus aureus (MRSA).³

Why You Might Think Blood Cultures are Helpful In Children with SSTIs?

Prior to the introduction of the Haemophilus influenzae vaccine, the rates of SSTI-associated bacteremia ranged from 8% to 20%.^4,5 Although the rate of bacteremia has declined significantly, blood cultures are still commonly performed as part of the evaluation of uncomplicated SSTIs in children; studies have shown that blood culture rates are 46% in the combined outpatient/inpatient setting,⁶ 34% in the ED setting,⁷ and 47%-94% in the inpatient setting.^7-11 Clinicians still feel that bacteremia detection is important to guide the selection of antibiotics and treatment duration. Providers may also underestimate the risk of obtaining a contaminant result and associated charges. Lastly, clinicians may perform blood cultures due to cultural norms at their institution.

Why Blood Cultures are Unnecessary in Children with Uncomplicated SSTIs

Several decades into the post vaccine era, the current guidelines from the Infectious Diseases Society of America (IDSA) do not recommend blood cultures as part of the routine evaluation of uncomplicated SSTIs.¹² Multiple single-center studies have failed to demonstrate the benefits of obtaining blood cultures in pediatric patients with uncomplicated SSTIs in the post-H. influenzae vaccine era.^6–11

Sadow et al¹¹ performed a retrospective case series of 381 children hospitalized with cellulitis to determine the rate and yield of blood cultures. Of the 266 (70%) patients who had a blood culture performed, 5 (1.9%) were true positives and 13 (5.4%) were contaminants. Notably, the true positive results included 3 children with active varicella infection and 2 children with septic joints; the latter would qualify as a complicated SSTI or as a separate infectious process altogether. No significant change in management resulted the positive blood cultures.

Harms Associated With Unnecessary Blood Cultures in SSTIs

Blood cultures necessitate venipunctures, which are painful for children and families. The inevitable false-positive contaminants also lead to repeat venipunctures and, potentially, unnecessary antibiotic exposure. From a high-value care perspective, Parikh et al reported hospital charges of $300 per blood culture and $250 for identification and sensitivity of positives.⁶ Assuming that these single-center charges are representative of national charges and using 0.5% true positivity and 1.5% false positivity rates, subjecting all children with uncomplicated SSTIs to blood culture would result in $60,250 charges to find one true positive blood culture, with no resultant changes in management. Additionally, among the 200 children cultured to find one true positive, there would be 3 false positives, necessitating another $1650 in charges for identification, sensitivity analysis, and repeat culture. These amounts do not factor in the significant expenditures associated with increased LOS. The potential savings associated with forgoing blood cultures in children with SSTIs should be an incentive for institutional change.

When Blood Cultures May Be Reasonable

The current IDSA guidelines recommend blood cultures for SSTIs in patients with immunodeficiency, animal bites, and immersion injuries (soft tissue injuries occurring in fresh or saltwater).¹² Previous studies also delineated criteria for “complicated” SSTIs, typically defined as surgical or traumatic wounds, infections requiring surgical intervention (not including simple incision and drainage), or infected ulcers or burns.^8,9 In the study of Malone et al, 10 (12.5%) positives were found among 80 patients with complicated SSTIs who had blood cultures performed.⁸ Although this work had a single-center study design with a relatively small sample size, no unusual organisms were found; the grown cultures included MRSA, MSSA, and S. pneumoniae. In addition to patients with complicated SSTIs, immunocompromised children, such as those receiving chemotherapy or other immunosuppressive agents, were excluded from the studies of blood culture yield in SSTIs and may warrant blood cultures given the risk of overwhelming infection and susceptibility to rare or invasive organisms.¹² In a study of 57 pediatric patients with leukemia and no central catheters who experienced skin or soft tissue complications, Demircioglu et al¹³ reported 6 positive blood cultures, including Klebsiella oxytoca, Pseudomonas aeruginosa, and Escherichia coli. These organisms would not be covered by typical SSTI antibiotic regimens, illustrating the value of blood cultures in this selected group of patients. Lastly, although the above studies included some infants, the data on utility of blood cultures in neonates are limited. Blood cultures may be reasonable in this group given the relative immunocompromised state of neonates compared with older children. Additionally, any infants aged

What You Should Do Instead Of Blood Cultures for Uncomplicated SSTIs

Gram stain and wound culture of any purulent material may assist with choice of empiric antibiotic therapy and appropriate narrowing of regimen for antibiotic stewardship. Wound cultures of purulent material can identify the causative organism in 58%-66% of the cases.^9,14 The rate of wound culture varies widely from 29% to 81% in studies across different healthcare systems.^9,10,15 The use of visually appealing posters advising clinicians to “culture pus, not blood” has been shown to significantly decreased the number of blood cultures performed at a single pediatric hospital.¹⁰

• Do not obtain blood cultures in pediatric patients with uncomplicated SSTIs.
• If purulent material is available spontaneously or after incision and drainage, then send it for Gram stain and bacterial culture.
• Blood cultures are reasonable in patients with complicated SSTIs and in immunocompromised patients with SSTIs.
• Despite limited data, blood cultures may be reasonable in neonates with SSTIs. Febrile infants with SSTIs aged less than 90 days should be managed under existing febrile infant guidelines.

Blood cultures in pediatric patients with uncomplicated SSTIs have no proven benefit and are associated with increased LOS, non-negligible false-positive rate, and associated increase in financial charges to the patient and healthcare system. The patient described in the clinical scenario would have an extremely low likelihood of having any meaningful clinical information provided by blood culture as part of her evaluation.

Do you think this is a low-value practice? Is this truly a “Thing We Do for No Reason?” Share what you do in your practice and join in the conversation online by retweeting it on Twitter (#TWDFNR) and liking it on Facebook. We invite you to propose ideas for other “Things We Do for No Reason” topics by emailing TWDFNR@hospitalmedicine.org.

DISCLOSURES

The authors have no conflicts of interest relevant to this article to disclose.

The “Things We Do for No Reason” (TWDFNR) series reviews practices that have become common parts of hospital care but may provide little value to our patients. Practices reviewed in the TWDFNR series do not represent “black and white” conclusions or clinical practice standards but are meant as a starting place for research and active discussions among hospitalists and patients. We invite you to be part of that discussion. https://www.choosingwisely.org/

An 8-year-old previously healthy girl presented to the emergency department (ED) with 2 days of warmth, swelling, and pain over her right upper thigh. Three days prior before presentation, a “pimple” appeared on her leg and drained a small amount of pus. Over the next 24 hours, the lesion became swollen, red, and painful. Her pediatrician prescribed trimethoprim-sulfamethoxazole. The patient took 3 doses of this medication but still experienced worsening pain and swelling.

In the ED, she had normal vital signs for her age except for temperature of 100.8 °F. A 2 cm × 3 cm area of fluctuance, erythema, and warmth was noted, and bedside ultrasound demonstrated a simple fluid collection. Incision and drainage was performed with expression of several milliliters of pus. The patient was referred for admission due to worsening symptoms despite outpatient antibiotic therapy. The ED providers ordered a blood culture at the time of admission.

Skin and soft tissue infections (SSTIs) are common pediatric diagnoses, which account for an estimated 390,000 ED visits annually¹ and represent the 7^th most common reason for pediatric hospital admission in the United States.² The rates of SSTIs have increased over the past several decades partly due to the rise of methicillin-resistant Staphylococcus aureus (MRSA).³

Why You Might Think Blood Cultures are Helpful In Children with SSTIs?

Prior to the introduction of the Haemophilus influenzae vaccine, the rates of SSTI-associated bacteremia ranged from 8% to 20%.^4,5 Although the rate of bacteremia has declined significantly, blood cultures are still commonly performed as part of the evaluation of uncomplicated SSTIs in children; studies have shown that blood culture rates are 46% in the combined outpatient/inpatient setting,⁶ 34% in the ED setting,⁷ and 47%-94% in the inpatient setting.^7-11 Clinicians still feel that bacteremia detection is important to guide the selection of antibiotics and treatment duration. Providers may also underestimate the risk of obtaining a contaminant result and associated charges. Lastly, clinicians may perform blood cultures due to cultural norms at their institution.

Why Blood Cultures are Unnecessary in Children with Uncomplicated SSTIs

Several decades into the post vaccine era, the current guidelines from the Infectious Diseases Society of America (IDSA) do not recommend blood cultures as part of the routine evaluation of uncomplicated SSTIs.¹² Multiple single-center studies have failed to demonstrate the benefits of obtaining blood cultures in pediatric patients with uncomplicated SSTIs in the post-H. influenzae vaccine era.^6–11

Sadow et al¹¹ performed a retrospective case series of 381 children hospitalized with cellulitis to determine the rate and yield of blood cultures. Of the 266 (70%) patients who had a blood culture performed, 5 (1.9%) were true positives and 13 (5.4%) were contaminants. Notably, the true positive results included 3 children with active varicella infection and 2 children with septic joints; the latter would qualify as a complicated SSTI or as a separate infectious process altogether. No significant change in management resulted the positive blood cultures.

Harms Associated With Unnecessary Blood Cultures in SSTIs

Blood cultures necessitate venipunctures, which are painful for children and families. The inevitable false-positive contaminants also lead to repeat venipunctures and, potentially, unnecessary antibiotic exposure. From a high-value care perspective, Parikh et al reported hospital charges of $300 per blood culture and $250 for identification and sensitivity of positives.⁶ Assuming that these single-center charges are representative of national charges and using 0.5% true positivity and 1.5% false positivity rates, subjecting all children with uncomplicated SSTIs to blood culture would result in $60,250 charges to find one true positive blood culture, with no resultant changes in management. Additionally, among the 200 children cultured to find one true positive, there would be 3 false positives, necessitating another $1650 in charges for identification, sensitivity analysis, and repeat culture. These amounts do not factor in the significant expenditures associated with increased LOS. The potential savings associated with forgoing blood cultures in children with SSTIs should be an incentive for institutional change.

When Blood Cultures May Be Reasonable

The current IDSA guidelines recommend blood cultures for SSTIs in patients with immunodeficiency, animal bites, and immersion injuries (soft tissue injuries occurring in fresh or saltwater).¹² Previous studies also delineated criteria for “complicated” SSTIs, typically defined as surgical or traumatic wounds, infections requiring surgical intervention (not including simple incision and drainage), or infected ulcers or burns.^8,9 In the study of Malone et al, 10 (12.5%) positives were found among 80 patients with complicated SSTIs who had blood cultures performed.⁸ Although this work had a single-center study design with a relatively small sample size, no unusual organisms were found; the grown cultures included MRSA, MSSA, and S. pneumoniae. In addition to patients with complicated SSTIs, immunocompromised children, such as those receiving chemotherapy or other immunosuppressive agents, were excluded from the studies of blood culture yield in SSTIs and may warrant blood cultures given the risk of overwhelming infection and susceptibility to rare or invasive organisms.¹² In a study of 57 pediatric patients with leukemia and no central catheters who experienced skin or soft tissue complications, Demircioglu et al¹³ reported 6 positive blood cultures, including Klebsiella oxytoca, Pseudomonas aeruginosa, and Escherichia coli. These organisms would not be covered by typical SSTI antibiotic regimens, illustrating the value of blood cultures in this selected group of patients. Lastly, although the above studies included some infants, the data on utility of blood cultures in neonates are limited. Blood cultures may be reasonable in this group given the relative immunocompromised state of neonates compared with older children. Additionally, any infants aged

What You Should Do Instead Of Blood Cultures for Uncomplicated SSTIs

Gram stain and wound culture of any purulent material may assist with choice of empiric antibiotic therapy and appropriate narrowing of regimen for antibiotic stewardship. Wound cultures of purulent material can identify the causative organism in 58%-66% of the cases.^9,14 The rate of wound culture varies widely from 29% to 81% in studies across different healthcare systems.^9,10,15 The use of visually appealing posters advising clinicians to “culture pus, not blood” has been shown to significantly decreased the number of blood cultures performed at a single pediatric hospital.¹⁰

• Do not obtain blood cultures in pediatric patients with uncomplicated SSTIs.
• If purulent material is available spontaneously or after incision and drainage, then send it for Gram stain and bacterial culture.
• Blood cultures are reasonable in patients with complicated SSTIs and in immunocompromised patients with SSTIs.
• Despite limited data, blood cultures may be reasonable in neonates with SSTIs. Febrile infants with SSTIs aged less than 90 days should be managed under existing febrile infant guidelines.

Blood cultures in pediatric patients with uncomplicated SSTIs have no proven benefit and are associated with increased LOS, non-negligible false-positive rate, and associated increase in financial charges to the patient and healthcare system. The patient described in the clinical scenario would have an extremely low likelihood of having any meaningful clinical information provided by blood culture as part of her evaluation.

Do you think this is a low-value practice? Is this truly a “Thing We Do for No Reason?” Share what you do in your practice and join in the conversation online by retweeting it on Twitter (#TWDFNR) and liking it on Facebook. We invite you to propose ideas for other “Things We Do for No Reason” topics by emailing TWDFNR@hospitalmedicine.org.

DISCLOSURES

The authors have no conflicts of interest relevant to this article to disclose.

The “Things We Do for No Reason” (TWDFNR) series reviews practices that have become common parts of hospital care but may provide little value to our patients. Practices reviewed in the TWDFNR series do not represent “black and white” conclusions or clinical practice standards but are meant as a starting place for research and active discussions among hospitalists and patients. We invite you to be part of that discussion. https://www.choosingwisely.org/

A 3-week-old, full-term term male febrile infant was evaluated in the emergency department (ED). On the day of admission, he was noted to feel warm to the touch and was found to have a rectal temperature of 101.3°F (38.3°C) at home.

In the ED, the patient was well appearing and had normal physical exam findings. His workup in the ED included a normal chest radiograph, complete blood count (CBC) with differential count, cerebrospinal fluid (CSF) analysis (cell count, protein, and glucose), and urinalysis. Blood, CSF, and catheterized urine cultures were collected, and he was admitted to the hospital on parenteral antibiotics. His provider informed the parents that the infant would be observed in the hospital for 48 hours while monitoring the bacterial cultures. Is it necessary for the hospitalization of this child to last a full 48 hours?

Evaluation and management of fever (T ≥ 38°C) is a common cause of emergency department visits and accounts for up to 20% of pediatric emergency visits.²

Why You Might Think Hospitalization for at Least 48 Hours is Necessary

The evaluation and management of fever in infants aged less than 90 days is challenging due to concern for occult serious bacterial infections. In particular, providers may be concerned that the physical exam lacks sensitivity.⁹

There is also a perceived risk of poor outcomes in young infants if a serious bacterial infection is missed. For these reasons, the evaluation and management of febrile infants has been characterized by practice variability in both outpatient¹⁰ and ED³ settings.

Commonly used febrile infant management protocols vary in approach and do not provide clear guidelines on the recommended duration of hospitalization and empiric antimicrobial treatment.^11-14 Length of hospitalization was widely studied in infants between 1979 and 1999, and results showed that the majority of clinically important bacterial pathogens can be detected within 48 hours.^15-17 Many textbooks and online references, based on this literature, continue to support 48 to 72 hours of observation and empiric antimicrobial treatment for febrile infants.^18,19 A 2012 AAP Clinical Report advocated for limiting the antimicrobial treatment in low-risk infants suspected of early-onset sepsis to 48 hours.²⁰

Why Shorten the Period of In-Hospital Observation to a Maximum of 36 Hours of Culture Incubation

Discharge of low-risk infants with negative enhanced urinalysis and negative bacterial cultures at 36 hours or earlier can reduce costs²¹ and potentially preventable harm (eg, intravenous catheter complications, nosocomial infections) without negatively impacting patient outcomes.²² Early discharge is also patient-centered, given the stress and indirect costs associated with hospitalization, including potential separation of a breastfeeding infant and mother, lost wages from time off work, or childcare for well siblings.²³

Initial studies that evaluated the time-to-positivity (TTP) of bacterial cultures in febrile infants predate the use of continuous monitoring systems for blood cultures. Traditional bacterial culturing techniques require direct observation of broth turbidity and subsequent subculturing onto chocolate and sheep blood agar, typically occurring only once daily.²⁴ Current commercially available continuous monitoring bacterial culture systems decrease TTP by immediately alerting laboratory technicians to bacterial growth through the detection of ¹⁴CO2 released by organisms utilizing radiolabeled glucose in growth media.²⁴ In addition, many studies supporting the evaluation of febrile infants in the hospital for a 48-hour period include those in ICU settings,²⁵ with medically complex histories,²⁴ and aged < 28 days admitted in the NICU,¹⁵ where pathogens with longer incubation times are frequently seen.

When a Longer Observation Period May Be Warranted

What You Should Do Instead: Limit Hospitalization to a Maximum of 36 Hours

For well-appearing febrile infants between 0–90 days of age hospitalized for observation and awaiting bacterial culture results, providers should consider discharge at 36 hours or less, rather than 48 hours, if blood, urine, and CSF cultures do not show bacterial growth. In a large health system, researchers implemented an evidence-based care process model for febrile infants to provide specific guidelines for laboratory testing, criteria for admission, and recommendation for discontinuation of empiric antibiotics and discharge after 36 hours in infants with negative bacterial cultures. These changes led to a 27% reduction in the length of hospital stay and 23% reduction in inpatient costs without any cases of missed bacteremia.²¹ The reduction in the in-hospital observation duration to 24 hours of culture incubation for well-appearing febrile infants has been advocated ³² and is a common practice for infants with appropriate follow up and parental assurance. This recommendation is supported by the following:

• Recent data showing the overwhelming majority of pathogens will be identified by blood culture <24 hours in infants aged 0-90 days³² with blood culture TTP in infants aged 0-30 days being either no different²⁶ or potentially shorter³²
• Studies showing that infants meeting low-risk clinical and laboratory profiles further reduce the likelihood of identifying serious bacterial infection after 24 hours to 0.3%.¹

• Determine if febrile infants aged 0-90 days are at low risk for serious bacterial infection and obtain appropriate bacterial cultures.
• If hospitalized for observation, discharge low-risk febrile infants aged 0–90 days after 36 hours or less if bacterial cultures remain negative.
• If hospitalized for observation, consider reducing the length of inpatient observation for low-risk febrile infants aged 0–90 days with reliable follow-up to 24 hours or less when the culture results are negative.

Monitoring patients in the hospital for greater than 36 hours of bacterial culture incubation is unnecessary for patients similar to the 3 week-old full-term infant in the case presentation, who are at low risk for serious bacterial infection based on available scoring systems and have negative cultures. If patients are not deemed low risk, have an incomplete laboratory evaluation, or have had prior antibiotic treatment, longer observation in the hospital may be warranted. Close reassessment of the rare patients whose blood cultures return positive after 36 hours is necessary, but their outcomes are excellent, especially in well-appearing infants.^7,33

What do you do?

Do you think this is a low-value practice? Is this truly a “Thing We Do for No Reason”? Let us know what you do in your practice and propose ideas for other “Things We Do for No Reason” topics. Please join in the conversation online at Twitter (#TWDFNR)/Facebook and don’t forget to “Like It” on Facebook or retweet it on Twitter. We invite you to propose ideas for other “Things We Do for No Reason” topics by emailingTWDFNR@hospitalmedicine.org.

Disclosures

There are no conflicts of interest relevant to this work reported by any of the authors.

The “Things We Do for No Reason” (TWDFNR) series reviews practices that have become common parts of hospital care but may provide little value to our patients. Practices reviewed in the TWDFNR series do not represent “black and white” conclusions or clinical practice standards but are meant as a starting place for research and active discussions among hospitalists and patients. We invite you to be part of that discussion. https://www.choosingwisely.org/

A 3-week-old, full-term term male febrile infant was evaluated in the emergency department (ED). On the day of admission, he was noted to feel warm to the touch and was found to have a rectal temperature of 101.3°F (38.3°C) at home.

In the ED, the patient was well appearing and had normal physical exam findings. His workup in the ED included a normal chest radiograph, complete blood count (CBC) with differential count, cerebrospinal fluid (CSF) analysis (cell count, protein, and glucose), and urinalysis. Blood, CSF, and catheterized urine cultures were collected, and he was admitted to the hospital on parenteral antibiotics. His provider informed the parents that the infant would be observed in the hospital for 48 hours while monitoring the bacterial cultures. Is it necessary for the hospitalization of this child to last a full 48 hours?

Evaluation and management of fever (T ≥ 38°C) is a common cause of emergency department visits and accounts for up to 20% of pediatric emergency visits.²

Why You Might Think Hospitalization for at Least 48 Hours is Necessary

The evaluation and management of fever in infants aged less than 90 days is challenging due to concern for occult serious bacterial infections. In particular, providers may be concerned that the physical exam lacks sensitivity.⁹

There is also a perceived risk of poor outcomes in young infants if a serious bacterial infection is missed. For these reasons, the evaluation and management of febrile infants has been characterized by practice variability in both outpatient¹⁰ and ED³ settings.

Commonly used febrile infant management protocols vary in approach and do not provide clear guidelines on the recommended duration of hospitalization and empiric antimicrobial treatment.^11-14 Length of hospitalization was widely studied in infants between 1979 and 1999, and results showed that the majority of clinically important bacterial pathogens can be detected within 48 hours.^15-17 Many textbooks and online references, based on this literature, continue to support 48 to 72 hours of observation and empiric antimicrobial treatment for febrile infants.^18,19 A 2012 AAP Clinical Report advocated for limiting the antimicrobial treatment in low-risk infants suspected of early-onset sepsis to 48 hours.²⁰

Why Shorten the Period of In-Hospital Observation to a Maximum of 36 Hours of Culture Incubation

Discharge of low-risk infants with negative enhanced urinalysis and negative bacterial cultures at 36 hours or earlier can reduce costs²¹ and potentially preventable harm (eg, intravenous catheter complications, nosocomial infections) without negatively impacting patient outcomes.²² Early discharge is also patient-centered, given the stress and indirect costs associated with hospitalization, including potential separation of a breastfeeding infant and mother, lost wages from time off work, or childcare for well siblings.²³

Initial studies that evaluated the time-to-positivity (TTP) of bacterial cultures in febrile infants predate the use of continuous monitoring systems for blood cultures. Traditional bacterial culturing techniques require direct observation of broth turbidity and subsequent subculturing onto chocolate and sheep blood agar, typically occurring only once daily.²⁴ Current commercially available continuous monitoring bacterial culture systems decrease TTP by immediately alerting laboratory technicians to bacterial growth through the detection of ¹⁴CO2 released by organisms utilizing radiolabeled glucose in growth media.²⁴ In addition, many studies supporting the evaluation of febrile infants in the hospital for a 48-hour period include those in ICU settings,²⁵ with medically complex histories,²⁴ and aged < 28 days admitted in the NICU,¹⁵ where pathogens with longer incubation times are frequently seen.

When a Longer Observation Period May Be Warranted

What You Should Do Instead: Limit Hospitalization to a Maximum of 36 Hours

For well-appearing febrile infants between 0–90 days of age hospitalized for observation and awaiting bacterial culture results, providers should consider discharge at 36 hours or less, rather than 48 hours, if blood, urine, and CSF cultures do not show bacterial growth. In a large health system, researchers implemented an evidence-based care process model for febrile infants to provide specific guidelines for laboratory testing, criteria for admission, and recommendation for discontinuation of empiric antibiotics and discharge after 36 hours in infants with negative bacterial cultures. These changes led to a 27% reduction in the length of hospital stay and 23% reduction in inpatient costs without any cases of missed bacteremia.²¹ The reduction in the in-hospital observation duration to 24 hours of culture incubation for well-appearing febrile infants has been advocated ³² and is a common practice for infants with appropriate follow up and parental assurance. This recommendation is supported by the following:

• Recent data showing the overwhelming majority of pathogens will be identified by blood culture <24 hours in infants aged 0-90 days³² with blood culture TTP in infants aged 0-30 days being either no different²⁶ or potentially shorter³²
• Studies showing that infants meeting low-risk clinical and laboratory profiles further reduce the likelihood of identifying serious bacterial infection after 24 hours to 0.3%.¹

• Determine if febrile infants aged 0-90 days are at low risk for serious bacterial infection and obtain appropriate bacterial cultures.
• If hospitalized for observation, discharge low-risk febrile infants aged 0–90 days after 36 hours or less if bacterial cultures remain negative.
• If hospitalized for observation, consider reducing the length of inpatient observation for low-risk febrile infants aged 0–90 days with reliable follow-up to 24 hours or less when the culture results are negative.

Monitoring patients in the hospital for greater than 36 hours of bacterial culture incubation is unnecessary for patients similar to the 3 week-old full-term infant in the case presentation, who are at low risk for serious bacterial infection based on available scoring systems and have negative cultures. If patients are not deemed low risk, have an incomplete laboratory evaluation, or have had prior antibiotic treatment, longer observation in the hospital may be warranted. Close reassessment of the rare patients whose blood cultures return positive after 36 hours is necessary, but their outcomes are excellent, especially in well-appearing infants.^7,33

What do you do?

Do you think this is a low-value practice? Is this truly a “Thing We Do for No Reason”? Let us know what you do in your practice and propose ideas for other “Things We Do for No Reason” topics. Please join in the conversation online at Twitter (#TWDFNR)/Facebook and don’t forget to “Like It” on Facebook or retweet it on Twitter. We invite you to propose ideas for other “Things We Do for No Reason” topics by emailingTWDFNR@hospitalmedicine.org.

Disclosures

There are no conflicts of interest relevant to this work reported by any of the authors.

The “Things We Do for No Reason” (TWDFNR) series reviews practices that have become common parts of hospital care but may provide little value to our patients. Practices reviewed in the TWDFNR series do not represent “black and white” conclusions or clinical practice standards but are meant as a starting place for research and active discussions among hospitalists and patients. We invite you to be part of that discussion. https://www.choosingwisely.org/

A 3-week-old, full-term term male febrile infant was evaluated in the emergency department (ED). On the day of admission, he was noted to feel warm to the touch and was found to have a rectal temperature of 101.3°F (38.3°C) at home.

In the ED, the patient was well appearing and had normal physical exam findings. His workup in the ED included a normal chest radiograph, complete blood count (CBC) with differential count, cerebrospinal fluid (CSF) analysis (cell count, protein, and glucose), and urinalysis. Blood, CSF, and catheterized urine cultures were collected, and he was admitted to the hospital on parenteral antibiotics. His provider informed the parents that the infant would be observed in the hospital for 48 hours while monitoring the bacterial cultures. Is it necessary for the hospitalization of this child to last a full 48 hours?

Evaluation and management of fever (T ≥ 38°C) is a common cause of emergency department visits and accounts for up to 20% of pediatric emergency visits.²

Why You Might Think Hospitalization for at Least 48 Hours is Necessary

The evaluation and management of fever in infants aged less than 90 days is challenging due to concern for occult serious bacterial infections. In particular, providers may be concerned that the physical exam lacks sensitivity.⁹

There is also a perceived risk of poor outcomes in young infants if a serious bacterial infection is missed. For these reasons, the evaluation and management of febrile infants has been characterized by practice variability in both outpatient¹⁰ and ED³ settings.

Commonly used febrile infant management protocols vary in approach and do not provide clear guidelines on the recommended duration of hospitalization and empiric antimicrobial treatment.^11-14 Length of hospitalization was widely studied in infants between 1979 and 1999, and results showed that the majority of clinically important bacterial pathogens can be detected within 48 hours.^15-17 Many textbooks and online references, based on this literature, continue to support 48 to 72 hours of observation and empiric antimicrobial treatment for febrile infants.^18,19 A 2012 AAP Clinical Report advocated for limiting the antimicrobial treatment in low-risk infants suspected of early-onset sepsis to 48 hours.²⁰

Why Shorten the Period of In-Hospital Observation to a Maximum of 36 Hours of Culture Incubation

Discharge of low-risk infants with negative enhanced urinalysis and negative bacterial cultures at 36 hours or earlier can reduce costs²¹ and potentially preventable harm (eg, intravenous catheter complications, nosocomial infections) without negatively impacting patient outcomes.²² Early discharge is also patient-centered, given the stress and indirect costs associated with hospitalization, including potential separation of a breastfeeding infant and mother, lost wages from time off work, or childcare for well siblings.²³

Initial studies that evaluated the time-to-positivity (TTP) of bacterial cultures in febrile infants predate the use of continuous monitoring systems for blood cultures. Traditional bacterial culturing techniques require direct observation of broth turbidity and subsequent subculturing onto chocolate and sheep blood agar, typically occurring only once daily.²⁴ Current commercially available continuous monitoring bacterial culture systems decrease TTP by immediately alerting laboratory technicians to bacterial growth through the detection of ¹⁴CO2 released by organisms utilizing radiolabeled glucose in growth media.²⁴ In addition, many studies supporting the evaluation of febrile infants in the hospital for a 48-hour period include those in ICU settings,²⁵ with medically complex histories,²⁴ and aged < 28 days admitted in the NICU,¹⁵ where pathogens with longer incubation times are frequently seen.

When a Longer Observation Period May Be Warranted

What You Should Do Instead: Limit Hospitalization to a Maximum of 36 Hours

For well-appearing febrile infants between 0–90 days of age hospitalized for observation and awaiting bacterial culture results, providers should consider discharge at 36 hours or less, rather than 48 hours, if blood, urine, and CSF cultures do not show bacterial growth. In a large health system, researchers implemented an evidence-based care process model for febrile infants to provide specific guidelines for laboratory testing, criteria for admission, and recommendation for discontinuation of empiric antibiotics and discharge after 36 hours in infants with negative bacterial cultures. These changes led to a 27% reduction in the length of hospital stay and 23% reduction in inpatient costs without any cases of missed bacteremia.²¹ The reduction in the in-hospital observation duration to 24 hours of culture incubation for well-appearing febrile infants has been advocated ³² and is a common practice for infants with appropriate follow up and parental assurance. This recommendation is supported by the following:

• Recent data showing the overwhelming majority of pathogens will be identified by blood culture <24 hours in infants aged 0-90 days³² with blood culture TTP in infants aged 0-30 days being either no different²⁶ or potentially shorter³²
• Studies showing that infants meeting low-risk clinical and laboratory profiles further reduce the likelihood of identifying serious bacterial infection after 24 hours to 0.3%.¹

• Determine if febrile infants aged 0-90 days are at low risk for serious bacterial infection and obtain appropriate bacterial cultures.
• If hospitalized for observation, discharge low-risk febrile infants aged 0–90 days after 36 hours or less if bacterial cultures remain negative.
• If hospitalized for observation, consider reducing the length of inpatient observation for low-risk febrile infants aged 0–90 days with reliable follow-up to 24 hours or less when the culture results are negative.

Monitoring patients in the hospital for greater than 36 hours of bacterial culture incubation is unnecessary for patients similar to the 3 week-old full-term infant in the case presentation, who are at low risk for serious bacterial infection based on available scoring systems and have negative cultures. If patients are not deemed low risk, have an incomplete laboratory evaluation, or have had prior antibiotic treatment, longer observation in the hospital may be warranted. Close reassessment of the rare patients whose blood cultures return positive after 36 hours is necessary, but their outcomes are excellent, especially in well-appearing infants.^7,33

What do you do?

Do you think this is a low-value practice? Is this truly a “Thing We Do for No Reason”? Let us know what you do in your practice and propose ideas for other “Things We Do for No Reason” topics. Please join in the conversation online at Twitter (#TWDFNR)/Facebook and don’t forget to “Like It” on Facebook or retweet it on Twitter. We invite you to propose ideas for other “Things We Do for No Reason” topics by emailingTWDFNR@hospitalmedicine.org.

Disclosures

There are no conflicts of interest relevant to this work reported by any of the authors.

The “Things We Do for No Reason” (TWDFNR) series reviews practices that have become common parts of hospital care but may provide little value to our patients. Practices reviewed in the TWDFNR series do not represent “black and white” conclusions or clinical practice standards, but are meant as a starting place for research and active discussions among hospitalists and patients. We invite you to be part of that discussion. https://www.choosingwisely.org/

Chest pain is one of the most common complaints among patients presenting to the emergency department. Moreover, at least 30% of patients who present with chest pain are admitted for observation, and >70% of those admitted with chest pain undergo cardiac stress testing (CST) during hospitalization. Several clinical risk prediction models have validated evaluation processes for managing patients with chest pain, helping to identify those at a low risk of major adverse cardiac events. Among these, the Thrombolysis in Myocardial Infarction or HEART score can identify patients safe to be discharged with outpatient CST within 72 h. It is unnecessary to hospitalize all low-risk patients for cardiac testing because it may expose them to needless risk and avoidable care costs, with little additional benefit.

A 60-year-old man with a history of osteoarthritis and depression presented to our emergency department (ED) with a 1-month history of left-sided chest pain that was present both at rest and exertion. There were no aggravating or relieving factors for the pain and no associated shortness of breath, diaphoresis, nausea, or lightheadedness. He smoked a half pack of cigarettes daily for 5 years in his twenties. The patient was taking aspirin 81 mg daily and paroxetine 40 mg daily, which he had been taking for 10 years. There was a family history of coronary artery disease in his mother, father, and sister. On examination, he was afebrile, with a blood pressure of 138/78 mm Hg and a heart rate of 62 beats/min; he appeared well, with no abnormal cardiopulmonary findings. Investigation revealed a normal initial troponin I level (<0.034 mg/mL) and normal electrocardiogram (ECG) with normal sinus rhythm (75 beats/min), normal axis, no ST changes, and no Q waves. He was therefore admitted to the hospital for further evaluation.

Each year, >7 million patients visit ED for chest pain in the United States,¹ with approximately 13% diagnosed with acute coronary syndromes (ACSs).² Over 30% of patients who present to ED with chest pain are hospitalized for observation, symptom evaluation, and risk stratification.³ In 2012, the mean Medicare reimbursement cost was $1,741 for in-hospital observation,⁴ with up to 70% of admitted patients undergoing cardiac stress testing (CST) before discharge.⁵

A scientific statement by the American Heart Association in 2010 recommended that patients considered to be at low risk for ACS after initial evaluation (based on presenting symptoms, past history, ECG findings, and initial cardiac biomarkers) should undergo CST within 72 h (preferably within 24 h) of presentation to provoke ischemia or detect anatomic coronary artery disease.⁶ Early exercise treadmill testing as part of an accelerated diagnostic pathway can also reduce the length of stays (LOS) in hospital and lower the medical costs.⁷ Moreover, when there is noncompliance or poor accessibility, failure to pursue early exercise testing in a hospital could result in a loss of patients to follow-up. Hospitalization for testing through accelerated diagnostic pathways may improve access to care and reduce clinical and legal risks associated with a major adverse cardiac event (MACE).

Clinical Risk Prediction Models

When a patient initially presents with chest pain, it should be determined if the symptoms are related to ACS or some other diagnosis. Hospitalization is required for patients with ACS but may not be for those without ACS and those with a low risk of inducible ischemia. Clinical risk scores and risk prediction models, such as the Thrombolysis in Myocardial Infarction (TIMI) and HEART scores, have been used in accelerated diagnostic protocols to determine a patient’s likelihood of having ACS. Several large trials of these clinical risk prediction models have validated the processes for evaluating patients with chest pain.

The TIMI risk score, the most well-known model, assesses risk based on the presence or absence of 7 characteristics (Appendix 1). It should be noted that the patient population studied for initial validation of this model comprised high-risk patients with unstable angina or non-ST elevation myocardial infarction who would benefit from early or urgent invasive therapy.⁸ In this population, TIMI scores of 0-1 are associated with low risk, with a 4.7% risk of ACS at 14 days.⁸ In another study of patients presenting to ED with undifferentiated chest pain and a TIMI score of zero, the risk of MACE at 30 days was approximately 2%.⁹

The HEART score is also used for patients presenting to ED with undifferentiated chest pain and assesses 5 separate variables scored 0–2 (Appendix 2). The original research gave a score of 2 to a troponin I level greater than twice the upper limit of the normal level,¹⁰ whereas a subsequent validation study gave a score of 2 to a troponin I or T level greater than or equal to 3 times the upper limit of the normal level.¹¹ Patients are considered at low, intermediate, and high risk based on scores of 0–3, 4–6, and 7–10, respectively.^10,11 Backus et al. performed a prospective randomized trial of 2388 patients who presented to ED with chest pain to validate the HEART score and compare it to the TIMI risk score. The HEART score performed better than the TIMI risk score in low-risk patients, with TIMI scores of 0-1 and HEART scores of 0–3 having a 6-week MACE risk of 2.8% and 1.7%, respectively.¹¹

A HEART pathway was developed that combines the HEART score with serial troponin I assays assessed at the time of initial presentation and approximately 3 h later.¹² Mahler et al. randomized 282 patients presenting to ED with chest pain to either the HEART pathway or conventional care. Patients with low-risk HEART scores and an abnormal troponin I level were admitted for cardiology consultation, whereas discharge was recommended for those with low scores and a normal troponin I level. Despite nearly 20% of the study cohort having a history of myocardial infarction, percutaneous coronary intervention, or coronary artery bypass grafting, approximately 40% of patients in the HEART pathway were identified as low risk, increasing early discharge rates by 21.3% and decreasing the average LOS by 12 h. No low-risk patient suffered a MACE within 30 days, and the HEART pathway had a sensitivity and a negative predictive value of approximately 99%.

Costs and Harms of Hospitalization for Cardiac Testing

Hospitalization carries measurable risks.^13,14 Between 2008 and 2013, Weinstock et al. evaluated the outcomes of patients presenting with chest pain who were placed in an observation unit for suspected ACS.¹⁵ Low-risk patients were defined as those with normal ECGs (no ischemic changes), 2 negative troponin tests performed 60–420 min apart (no particular troponin assay specified), and stable vital signs. They identified 7266 patients who were considered to have low risk, among whom 4 (0.06%) had an adverse outcome in the hospital (eg, life-threatening arrhythmia, ST-segment elevation myocardial infarction, cardiac or respiratory arrest, or death); 3 among the 4 patients had a cardiac-related adverse outcome. The overall risk of adverse outcomes due to cardiac causes was 1 in 2422 admissions (0.04%). The authors compared their results with the reported risk of 1 in 164 admissions for preventable adverse events contributing to patient death during routine hospitalization (eg, medication or procedure errors).¹⁴

Outpatient CST can be reliably and safely performed for patients with chest pain.^16-18 There is no clear evidence that earlier CST leads to improved patient outcomes, and CST in the absence of acute ischemia (or ACS) increases the rates of angiography and revascularization without improvements in the rate of myocardial infarction.^19-21 Given the costs of in-hospital observation⁴ and the dubious benefits of providing CST for patients with low-risk chest pain, admitting all patients with low-risk chest pain exposes them to costs and harms with little potential benefit.

Patients presenting with chest pain with either dynamic ECG changes or an elevated troponin level require hospitalization for further ACS diagnosis and treatment. When ACS cannot be clearly diagnosed at the initial evaluation, healthcare providers should use clinical risk prediction models to stratify patients. Those deemed to be at an intermediate or high risk by these models should be hospitalized for further evaluation, as should those at low risk but for whom access to outpatient follow-up is difficult (eg, those without health insurance).

A complete history and physical examination, along with ECG and cardiac biomarker testing, are required for all patients presenting with chest pain. Validated clinical risk prediction models should then be used to determine the likelihood of a cardiac event. Fanaroff et al. reported that low-risk HEART scores of 0–3 and TIMI scores of 0-1 gave positive likelihood ratios of 0.2 and 0.31, respectively.²² Using a pre-test probability of 13%, as reported by Bhuiya et al.,² the likelihood of ACS or MACE within 6 weeks is 2.9% for patients with low-risk HEART scores and 4.4% for those with low-risk TIMI scores.²² These risk prediction models allow clinicians to provide a shared decision-making plan with the patient and discuss the risks and benefits of in-hospital versus outpatient cardiac testing, especially among patients with access to appropriate outpatient follow-up.²³ Low-risk patients can be referred for outpatient testing within 72 h, reducing hospitalization-associated costs and harms.

• Patients presenting with chest pain should undergo a complete history taking and physical examination, as well as ECG and cardiac biomarker testing (eg, troponin I level at presentation and approximately 3 h later).
• Clinical risk prediction models, such as TIMI or HEART scores, should then be used to determine the risk of MACE.
• Patients at a low risk may be safely discharged with outpatient CST performed within 72 h.
• Patients at an intermediate or high risk of MACE should be hospitalized for further evaluation, as should those with low-risk chest pain who are unable to attend follow-up for outpatient CST within 72 h.
• Clinicians should provide a shared decision-making plan with each patient, taking care to discuss the risks and benefits of in-hospital versus outpatient CST.

The risk of MACE should be assessed in all patients presenting to ED with low-risk chest pain to avoid unnecessary hospitalization that exposes them to potential costs and harms with few additional benefits. If the risk scoring system was applied to the patient described in our original clinical scenario, he would have had a HEART score of 3 (ie, 1 point for a moderately suspicious history, 1 point for the age of 60 years, and 1 point for a positive family history) and a TIMI score of 1 (ie, 1 point for aspirin use within past 7 days). Therefore, he could be stratified as having a low-risk presentation. With a second negative troponin I test at 3 h, discharge from ED with timely outpatient CST within 72 h would be an appropriate management strategy.

Do you think this is a low-value practice? Is this truly a “Thing We Do for No Reason”? Share what you do in your practice and join in the conversation online by retweeting it on Twitter (#TWDFNR) and liking it on Facebook. We invite you to propose ideas for other “Things We Do for No Reason” topics by emailing TWDFNR@hospitalmedicine.org.

Conflicts of Interest

 The authors have no conflicts of interest relevant to this article to disclose.

The “Things We Do for No Reason” (TWDFNR) series reviews practices that have become common parts of hospital care but may provide little value to our patients. Practices reviewed in the TWDFNR series do not represent “black and white” conclusions or clinical practice standards, but are meant as a starting place for research and active discussions among hospitalists and patients. We invite you to be part of that discussion. https://www.choosingwisely.org/

Chest pain is one of the most common complaints among patients presenting to the emergency department. Moreover, at least 30% of patients who present with chest pain are admitted for observation, and >70% of those admitted with chest pain undergo cardiac stress testing (CST) during hospitalization. Several clinical risk prediction models have validated evaluation processes for managing patients with chest pain, helping to identify those at a low risk of major adverse cardiac events. Among these, the Thrombolysis in Myocardial Infarction or HEART score can identify patients safe to be discharged with outpatient CST within 72 h. It is unnecessary to hospitalize all low-risk patients for cardiac testing because it may expose them to needless risk and avoidable care costs, with little additional benefit.

A 60-year-old man with a history of osteoarthritis and depression presented to our emergency department (ED) with a 1-month history of left-sided chest pain that was present both at rest and exertion. There were no aggravating or relieving factors for the pain and no associated shortness of breath, diaphoresis, nausea, or lightheadedness. He smoked a half pack of cigarettes daily for 5 years in his twenties. The patient was taking aspirin 81 mg daily and paroxetine 40 mg daily, which he had been taking for 10 years. There was a family history of coronary artery disease in his mother, father, and sister. On examination, he was afebrile, with a blood pressure of 138/78 mm Hg and a heart rate of 62 beats/min; he appeared well, with no abnormal cardiopulmonary findings. Investigation revealed a normal initial troponin I level (<0.034 mg/mL) and normal electrocardiogram (ECG) with normal sinus rhythm (75 beats/min), normal axis, no ST changes, and no Q waves. He was therefore admitted to the hospital for further evaluation.

Each year, >7 million patients visit ED for chest pain in the United States,¹ with approximately 13% diagnosed with acute coronary syndromes (ACSs).² Over 30% of patients who present to ED with chest pain are hospitalized for observation, symptom evaluation, and risk stratification.³ In 2012, the mean Medicare reimbursement cost was $1,741 for in-hospital observation,⁴ with up to 70% of admitted patients undergoing cardiac stress testing (CST) before discharge.⁵

A scientific statement by the American Heart Association in 2010 recommended that patients considered to be at low risk for ACS after initial evaluation (based on presenting symptoms, past history, ECG findings, and initial cardiac biomarkers) should undergo CST within 72 h (preferably within 24 h) of presentation to provoke ischemia or detect anatomic coronary artery disease.⁶ Early exercise treadmill testing as part of an accelerated diagnostic pathway can also reduce the length of stays (LOS) in hospital and lower the medical costs.⁷ Moreover, when there is noncompliance or poor accessibility, failure to pursue early exercise testing in a hospital could result in a loss of patients to follow-up. Hospitalization for testing through accelerated diagnostic pathways may improve access to care and reduce clinical and legal risks associated with a major adverse cardiac event (MACE).

Clinical Risk Prediction Models

When a patient initially presents with chest pain, it should be determined if the symptoms are related to ACS or some other diagnosis. Hospitalization is required for patients with ACS but may not be for those without ACS and those with a low risk of inducible ischemia. Clinical risk scores and risk prediction models, such as the Thrombolysis in Myocardial Infarction (TIMI) and HEART scores, have been used in accelerated diagnostic protocols to determine a patient’s likelihood of having ACS. Several large trials of these clinical risk prediction models have validated the processes for evaluating patients with chest pain.

The TIMI risk score, the most well-known model, assesses risk based on the presence or absence of 7 characteristics (Appendix 1). It should be noted that the patient population studied for initial validation of this model comprised high-risk patients with unstable angina or non-ST elevation myocardial infarction who would benefit from early or urgent invasive therapy.⁸ In this population, TIMI scores of 0-1 are associated with low risk, with a 4.7% risk of ACS at 14 days.⁸ In another study of patients presenting to ED with undifferentiated chest pain and a TIMI score of zero, the risk of MACE at 30 days was approximately 2%.⁹

The HEART score is also used for patients presenting to ED with undifferentiated chest pain and assesses 5 separate variables scored 0–2 (Appendix 2). The original research gave a score of 2 to a troponin I level greater than twice the upper limit of the normal level,¹⁰ whereas a subsequent validation study gave a score of 2 to a troponin I or T level greater than or equal to 3 times the upper limit of the normal level.¹¹ Patients are considered at low, intermediate, and high risk based on scores of 0–3, 4–6, and 7–10, respectively.^10,11 Backus et al. performed a prospective randomized trial of 2388 patients who presented to ED with chest pain to validate the HEART score and compare it to the TIMI risk score. The HEART score performed better than the TIMI risk score in low-risk patients, with TIMI scores of 0-1 and HEART scores of 0–3 having a 6-week MACE risk of 2.8% and 1.7%, respectively.¹¹

A HEART pathway was developed that combines the HEART score with serial troponin I assays assessed at the time of initial presentation and approximately 3 h later.¹² Mahler et al. randomized 282 patients presenting to ED with chest pain to either the HEART pathway or conventional care. Patients with low-risk HEART scores and an abnormal troponin I level were admitted for cardiology consultation, whereas discharge was recommended for those with low scores and a normal troponin I level. Despite nearly 20% of the study cohort having a history of myocardial infarction, percutaneous coronary intervention, or coronary artery bypass grafting, approximately 40% of patients in the HEART pathway were identified as low risk, increasing early discharge rates by 21.3% and decreasing the average LOS by 12 h. No low-risk patient suffered a MACE within 30 days, and the HEART pathway had a sensitivity and a negative predictive value of approximately 99%.

Costs and Harms of Hospitalization for Cardiac Testing

Hospitalization carries measurable risks.^13,14 Between 2008 and 2013, Weinstock et al. evaluated the outcomes of patients presenting with chest pain who were placed in an observation unit for suspected ACS.¹⁵ Low-risk patients were defined as those with normal ECGs (no ischemic changes), 2 negative troponin tests performed 60–420 min apart (no particular troponin assay specified), and stable vital signs. They identified 7266 patients who were considered to have low risk, among whom 4 (0.06%) had an adverse outcome in the hospital (eg, life-threatening arrhythmia, ST-segment elevation myocardial infarction, cardiac or respiratory arrest, or death); 3 among the 4 patients had a cardiac-related adverse outcome. The overall risk of adverse outcomes due to cardiac causes was 1 in 2422 admissions (0.04%). The authors compared their results with the reported risk of 1 in 164 admissions for preventable adverse events contributing to patient death during routine hospitalization (eg, medication or procedure errors).¹⁴

Outpatient CST can be reliably and safely performed for patients with chest pain.^16-18 There is no clear evidence that earlier CST leads to improved patient outcomes, and CST in the absence of acute ischemia (or ACS) increases the rates of angiography and revascularization without improvements in the rate of myocardial infarction.^19-21 Given the costs of in-hospital observation⁴ and the dubious benefits of providing CST for patients with low-risk chest pain, admitting all patients with low-risk chest pain exposes them to costs and harms with little potential benefit.

Patients presenting with chest pain with either dynamic ECG changes or an elevated troponin level require hospitalization for further ACS diagnosis and treatment. When ACS cannot be clearly diagnosed at the initial evaluation, healthcare providers should use clinical risk prediction models to stratify patients. Those deemed to be at an intermediate or high risk by these models should be hospitalized for further evaluation, as should those at low risk but for whom access to outpatient follow-up is difficult (eg, those without health insurance).

A complete history and physical examination, along with ECG and cardiac biomarker testing, are required for all patients presenting with chest pain. Validated clinical risk prediction models should then be used to determine the likelihood of a cardiac event. Fanaroff et al. reported that low-risk HEART scores of 0–3 and TIMI scores of 0-1 gave positive likelihood ratios of 0.2 and 0.31, respectively.²² Using a pre-test probability of 13%, as reported by Bhuiya et al.,² the likelihood of ACS or MACE within 6 weeks is 2.9% for patients with low-risk HEART scores and 4.4% for those with low-risk TIMI scores.²² These risk prediction models allow clinicians to provide a shared decision-making plan with the patient and discuss the risks and benefits of in-hospital versus outpatient cardiac testing, especially among patients with access to appropriate outpatient follow-up.²³ Low-risk patients can be referred for outpatient testing within 72 h, reducing hospitalization-associated costs and harms.

• Patients presenting with chest pain should undergo a complete history taking and physical examination, as well as ECG and cardiac biomarker testing (eg, troponin I level at presentation and approximately 3 h later).
• Clinical risk prediction models, such as TIMI or HEART scores, should then be used to determine the risk of MACE.
• Patients at a low risk may be safely discharged with outpatient CST performed within 72 h.
• Patients at an intermediate or high risk of MACE should be hospitalized for further evaluation, as should those with low-risk chest pain who are unable to attend follow-up for outpatient CST within 72 h.
• Clinicians should provide a shared decision-making plan with each patient, taking care to discuss the risks and benefits of in-hospital versus outpatient CST.

The risk of MACE should be assessed in all patients presenting to ED with low-risk chest pain to avoid unnecessary hospitalization that exposes them to potential costs and harms with few additional benefits. If the risk scoring system was applied to the patient described in our original clinical scenario, he would have had a HEART score of 3 (ie, 1 point for a moderately suspicious history, 1 point for the age of 60 years, and 1 point for a positive family history) and a TIMI score of 1 (ie, 1 point for aspirin use within past 7 days). Therefore, he could be stratified as having a low-risk presentation. With a second negative troponin I test at 3 h, discharge from ED with timely outpatient CST within 72 h would be an appropriate management strategy.

Do you think this is a low-value practice? Is this truly a “Thing We Do for No Reason”? Share what you do in your practice and join in the conversation online by retweeting it on Twitter (#TWDFNR) and liking it on Facebook. We invite you to propose ideas for other “Things We Do for No Reason” topics by emailing TWDFNR@hospitalmedicine.org.

Conflicts of Interest

 The authors have no conflicts of interest relevant to this article to disclose.

The “Things We Do for No Reason” (TWDFNR) series reviews practices that have become common parts of hospital care but may provide little value to our patients. Practices reviewed in the TWDFNR series do not represent “black and white” conclusions or clinical practice standards, but are meant as a starting place for research and active discussions among hospitalists and patients. We invite you to be part of that discussion. https://www.choosingwisely.org/

Chest pain is one of the most common complaints among patients presenting to the emergency department. Moreover, at least 30% of patients who present with chest pain are admitted for observation, and >70% of those admitted with chest pain undergo cardiac stress testing (CST) during hospitalization. Several clinical risk prediction models have validated evaluation processes for managing patients with chest pain, helping to identify those at a low risk of major adverse cardiac events. Among these, the Thrombolysis in Myocardial Infarction or HEART score can identify patients safe to be discharged with outpatient CST within 72 h. It is unnecessary to hospitalize all low-risk patients for cardiac testing because it may expose them to needless risk and avoidable care costs, with little additional benefit.

A 60-year-old man with a history of osteoarthritis and depression presented to our emergency department (ED) with a 1-month history of left-sided chest pain that was present both at rest and exertion. There were no aggravating or relieving factors for the pain and no associated shortness of breath, diaphoresis, nausea, or lightheadedness. He smoked a half pack of cigarettes daily for 5 years in his twenties. The patient was taking aspirin 81 mg daily and paroxetine 40 mg daily, which he had been taking for 10 years. There was a family history of coronary artery disease in his mother, father, and sister. On examination, he was afebrile, with a blood pressure of 138/78 mm Hg and a heart rate of 62 beats/min; he appeared well, with no abnormal cardiopulmonary findings. Investigation revealed a normal initial troponin I level (<0.034 mg/mL) and normal electrocardiogram (ECG) with normal sinus rhythm (75 beats/min), normal axis, no ST changes, and no Q waves. He was therefore admitted to the hospital for further evaluation.

Each year, >7 million patients visit ED for chest pain in the United States,¹ with approximately 13% diagnosed with acute coronary syndromes (ACSs).² Over 30% of patients who present to ED with chest pain are hospitalized for observation, symptom evaluation, and risk stratification.³ In 2012, the mean Medicare reimbursement cost was $1,741 for in-hospital observation,⁴ with up to 70% of admitted patients undergoing cardiac stress testing (CST) before discharge.⁵

A scientific statement by the American Heart Association in 2010 recommended that patients considered to be at low risk for ACS after initial evaluation (based on presenting symptoms, past history, ECG findings, and initial cardiac biomarkers) should undergo CST within 72 h (preferably within 24 h) of presentation to provoke ischemia or detect anatomic coronary artery disease.⁶ Early exercise treadmill testing as part of an accelerated diagnostic pathway can also reduce the length of stays (LOS) in hospital and lower the medical costs.⁷ Moreover, when there is noncompliance or poor accessibility, failure to pursue early exercise testing in a hospital could result in a loss of patients to follow-up. Hospitalization for testing through accelerated diagnostic pathways may improve access to care and reduce clinical and legal risks associated with a major adverse cardiac event (MACE).

Clinical Risk Prediction Models

When a patient initially presents with chest pain, it should be determined if the symptoms are related to ACS or some other diagnosis. Hospitalization is required for patients with ACS but may not be for those without ACS and those with a low risk of inducible ischemia. Clinical risk scores and risk prediction models, such as the Thrombolysis in Myocardial Infarction (TIMI) and HEART scores, have been used in accelerated diagnostic protocols to determine a patient’s likelihood of having ACS. Several large trials of these clinical risk prediction models have validated the processes for evaluating patients with chest pain.

The TIMI risk score, the most well-known model, assesses risk based on the presence or absence of 7 characteristics (Appendix 1). It should be noted that the patient population studied for initial validation of this model comprised high-risk patients with unstable angina or non-ST elevation myocardial infarction who would benefit from early or urgent invasive therapy.⁸ In this population, TIMI scores of 0-1 are associated with low risk, with a 4.7% risk of ACS at 14 days.⁸ In another study of patients presenting to ED with undifferentiated chest pain and a TIMI score of zero, the risk of MACE at 30 days was approximately 2%.⁹

The HEART score is also used for patients presenting to ED with undifferentiated chest pain and assesses 5 separate variables scored 0–2 (Appendix 2). The original research gave a score of 2 to a troponin I level greater than twice the upper limit of the normal level,¹⁰ whereas a subsequent validation study gave a score of 2 to a troponin I or T level greater than or equal to 3 times the upper limit of the normal level.¹¹ Patients are considered at low, intermediate, and high risk based on scores of 0–3, 4–6, and 7–10, respectively.^10,11 Backus et al. performed a prospective randomized trial of 2388 patients who presented to ED with chest pain to validate the HEART score and compare it to the TIMI risk score. The HEART score performed better than the TIMI risk score in low-risk patients, with TIMI scores of 0-1 and HEART scores of 0–3 having a 6-week MACE risk of 2.8% and 1.7%, respectively.¹¹

A HEART pathway was developed that combines the HEART score with serial troponin I assays assessed at the time of initial presentation and approximately 3 h later.¹² Mahler et al. randomized 282 patients presenting to ED with chest pain to either the HEART pathway or conventional care. Patients with low-risk HEART scores and an abnormal troponin I level were admitted for cardiology consultation, whereas discharge was recommended for those with low scores and a normal troponin I level. Despite nearly 20% of the study cohort having a history of myocardial infarction, percutaneous coronary intervention, or coronary artery bypass grafting, approximately 40% of patients in the HEART pathway were identified as low risk, increasing early discharge rates by 21.3% and decreasing the average LOS by 12 h. No low-risk patient suffered a MACE within 30 days, and the HEART pathway had a sensitivity and a negative predictive value of approximately 99%.

Costs and Harms of Hospitalization for Cardiac Testing

Hospitalization carries measurable risks.^13,14 Between 2008 and 2013, Weinstock et al. evaluated the outcomes of patients presenting with chest pain who were placed in an observation unit for suspected ACS.¹⁵ Low-risk patients were defined as those with normal ECGs (no ischemic changes), 2 negative troponin tests performed 60–420 min apart (no particular troponin assay specified), and stable vital signs. They identified 7266 patients who were considered to have low risk, among whom 4 (0.06%) had an adverse outcome in the hospital (eg, life-threatening arrhythmia, ST-segment elevation myocardial infarction, cardiac or respiratory arrest, or death); 3 among the 4 patients had a cardiac-related adverse outcome. The overall risk of adverse outcomes due to cardiac causes was 1 in 2422 admissions (0.04%). The authors compared their results with the reported risk of 1 in 164 admissions for preventable adverse events contributing to patient death during routine hospitalization (eg, medication or procedure errors).¹⁴

Outpatient CST can be reliably and safely performed for patients with chest pain.^16-18 There is no clear evidence that earlier CST leads to improved patient outcomes, and CST in the absence of acute ischemia (or ACS) increases the rates of angiography and revascularization without improvements in the rate of myocardial infarction.^19-21 Given the costs of in-hospital observation⁴ and the dubious benefits of providing CST for patients with low-risk chest pain, admitting all patients with low-risk chest pain exposes them to costs and harms with little potential benefit.

Patients presenting with chest pain with either dynamic ECG changes or an elevated troponin level require hospitalization for further ACS diagnosis and treatment. When ACS cannot be clearly diagnosed at the initial evaluation, healthcare providers should use clinical risk prediction models to stratify patients. Those deemed to be at an intermediate or high risk by these models should be hospitalized for further evaluation, as should those at low risk but for whom access to outpatient follow-up is difficult (eg, those without health insurance).

A complete history and physical examination, along with ECG and cardiac biomarker testing, are required for all patients presenting with chest pain. Validated clinical risk prediction models should then be used to determine the likelihood of a cardiac event. Fanaroff et al. reported that low-risk HEART scores of 0–3 and TIMI scores of 0-1 gave positive likelihood ratios of 0.2 and 0.31, respectively.²² Using a pre-test probability of 13%, as reported by Bhuiya et al.,² the likelihood of ACS or MACE within 6 weeks is 2.9% for patients with low-risk HEART scores and 4.4% for those with low-risk TIMI scores.²² These risk prediction models allow clinicians to provide a shared decision-making plan with the patient and discuss the risks and benefits of in-hospital versus outpatient cardiac testing, especially among patients with access to appropriate outpatient follow-up.²³ Low-risk patients can be referred for outpatient testing within 72 h, reducing hospitalization-associated costs and harms.

• Patients presenting with chest pain should undergo a complete history taking and physical examination, as well as ECG and cardiac biomarker testing (eg, troponin I level at presentation and approximately 3 h later).
• Clinical risk prediction models, such as TIMI or HEART scores, should then be used to determine the risk of MACE.
• Patients at a low risk may be safely discharged with outpatient CST performed within 72 h.
• Patients at an intermediate or high risk of MACE should be hospitalized for further evaluation, as should those with low-risk chest pain who are unable to attend follow-up for outpatient CST within 72 h.
• Clinicians should provide a shared decision-making plan with each patient, taking care to discuss the risks and benefits of in-hospital versus outpatient CST.

The risk of MACE should be assessed in all patients presenting to ED with low-risk chest pain to avoid unnecessary hospitalization that exposes them to potential costs and harms with few additional benefits. If the risk scoring system was applied to the patient described in our original clinical scenario, he would have had a HEART score of 3 (ie, 1 point for a moderately suspicious history, 1 point for the age of 60 years, and 1 point for a positive family history) and a TIMI score of 1 (ie, 1 point for aspirin use within past 7 days). Therefore, he could be stratified as having a low-risk presentation. With a second negative troponin I test at 3 h, discharge from ED with timely outpatient CST within 72 h would be an appropriate management strategy.

Do you think this is a low-value practice? Is this truly a “Thing We Do for No Reason”? Share what you do in your practice and join in the conversation online by retweeting it on Twitter (#TWDFNR) and liking it on Facebook. We invite you to propose ideas for other “Things We Do for No Reason” topics by emailing TWDFNR@hospitalmedicine.org.

Conflicts of Interest

 The authors have no conflicts of interest relevant to this article to disclose.

The “Things We Do for No Reason” (TWDFNR) series reviews practices that have become common parts of hospital care but that may provide little value to our patients. Practices reviewed in the TWDFNR series do not represent “black and white” conclusions or clinical practice standards, but are meant as a starting place for research and active discussions among hospitalists and patients. We invite you to be part of that discussion. https://www.choosingwisely.org/

Oral anticoagulation (OAC) is commonly prescribed to patients with atrial fibrillation, venous thromboembolism (VTE), and mechanical heart valves (MHVs) for primary and secondary thromboembolism prevention. When patients require surgery or an invasive procedure, “bridging” anticoagulants (eg, enoxaparin) are commonly administered during the period of OAC interruption to reduce thromboembolic risk. This practice stems from small observational studies and expert opinion, which influenced several clinical guidelines despite the lack of high-quality evidence. Although prospective randomized trials of periprocedural bridging in patients with VTE and MHVs are lacking, available evidence is consistent with findings from the BRIDGE trial, which guides the following general recommendations: (1) avoid unnecessary periprocedural interruptions of OAC, especially for low bleeding risk procedures; (2) avoid the administration of periprocedural bridging anticoagulation in patients with low to moderate thromboembolic risk; (3) in patients with high thromboembolic risk, individually assess the patient-specific and procedure-specific bleeding risks versus thromboembolic risks.

A 75-year-old man with a history of hypertension, diabetes mellitus, and atrial fibrillation is admitted for surgical repair of a comminuted intertrochanteric left hip fracture. He suffered a mechanical ground-level fall without loss of consciousness. At baseline, he denies any chest pain, dyspnea on exertion, or recent change in his exercise tolerance. A physical examination is notable for stable vital signs, irregular cardiac rhythm, and a shortened and externally rotated left lower extremity with exquisite tenderness to palpation and range of motion. The patient is taking warfarin for stroke prophylaxis based on a CHA₂DS₂VaSc score of 4 points. The international normalized ratio (INR) is 1.9 upon admission, and surgery is planned within 48 hours, once the patient is “medically cleared.” Will this patient benefit from periprocedural bridging anticoagulation?

OAC is commonly prescribed to patients with atrial fibrillation, venous thromboembolism (VTE), and mechanical heart valves (MHVs) for the primary or secondary prevention of thromboembolic events, with more than 35 million prescriptions written annually in the United States alone.¹ Many of these patients will require a temporary interruption of their OAC for surgery or an invasive procedure.² As a result, patients may be treated with short-acting, or “bridging,” anticoagulants, such as low-molecular-weight heparin (LMWH), to minimize the duration of anticoagulation interruption and theoretically reduce their thromboembolic risk. The rationale for bridging stemmed from small observational studies and expert opinion that perceived the estimated thromboembolic risk to be higher than the estimated bleeding risk.^3-5 One such example estimated that the VTE risk increased 100-fold postoperatively, whereas heparin administration only doubled the bleeding risk.³ Furthermore, clinical practice guidelines published from the American Heart Association, American College of Cardiology, European Heart Rhythm Society, and American College of Chest Physicians recommend when and how to initiate bridging anticoagulation. Clinicians have widely adopted these recommendations despite an acknowledged paucity of high-quality supporting evidence.^6,7

Periprocedural Anticoagulation Interruption is Often Not Indicated

Patients undergoing a surgical or invasive procedure may require an interruption of OAC to minimize the periprocedural bleeding risk. The decision to interrupt OAC should generally be based on the procedure-specific bleeding risk. Procedures with low bleeding risk such as cataract surgery, dermatologic biopsy (including Mohs), arthrocentesis, diagnostic gastrointestinal endoscopy, and cardiac pacemaker implantation can be performed safely without OAC interruption.^5,7 Despite evidence supporting the safety of periprocedural OAC continuation, unnecessary OAC interruptions remain commonplace and are associated with increased adverse outcomes.⁸ The BRUISE CONTROL trial compared uninterrupted OAC to interrupted OAC with periprocedural bridging for cardiac pacemaker or defibrillator implantation in a moderate to high thromboembolic risk population. The uninterrupted OAC group experienced significantly fewer pocket hematomas, hematoma evacuations, and prolonged hospitalizations (relative risk [RR] 0.19-0.24; P < .05) without significantly increased thromboembolic events, highlighting the potential benefits of this approach.⁹

Nevertheless, many surgical and invasive procedures do warrant OAC interruption due to the inherent bleeding risk of the procedure or other logistical considerations. Procedures associated with an increased bleeding risk include urologic surgery (except laser lithotripsy), surgery on highly vascular organs (eg, kidney, liver, spleen), bowel resection, cardiac surgery, and intracranial or spinal surgery.⁷ Alternatively, some procedures with acceptably low bleeding risk (eg, colonoscopy) are routinely performed during an OAC interruption due to the fact that a high bleeding risk intervention may be necessary during the procedure (eg, polypectomy). This approach may be preferable when a significant amount of preparation is required (eg, bowel preparation) and may be a more efficient use of healthcare resources by avoiding repeat procedures.

Bridging Anticoagulation Does Not Significantly Reduce Thromboembolic Events

BRIDGE was a randomized, double-blind, placebo-controlled trial among patients with atrial fibrillation (n = 1884) requiring OAC interruption for mostly low-risk, ambulatory surgeries or invasive procedures (eg, gastrointestinal endoscopy, cardiac catheterization). Notably, thromboembolism events were rare, and there was no significant difference in thromboembolism events between patients randomized to placebo or bridging with LMWH (0.4% vs 0.3%, respectively; P = .73).¹⁴ However, the proportion of patients enrolled with the highest thromboembolic risk (ie, CHADS₂ score 5-6 or prior transient ischemic attack and/or stroke) was low, potentially indicating an underestimated benefit in these patients. Major bleeding was significantly reduced in patients forgoing bridging anticoagulation (1.3% vs 3.2%; RR 0.41; 95% confidence interval, 0.20-0.78; P = .005), although bleeding occurred more frequently than thromboembolism in both groups.

Even though randomized trials assessing the safety and efficacy of bridging for VTE or MHVs have not been completed, evidence is not entirely lacking.^16,17 A rigorous observational study limited to a VTE cohort (deep vein thrombosis of upper or lower extremity and/or pulmonary embolism) analyzed the effects of bridging in patients with a surgical or invasive procedure-related OAC interruption. Patients were stratified according to the American College of Chest Physicians perioperative guideline risk-stratification schema, and most VTE events (≥93%) occurred more than 12 months prior to OAC interruption.⁷ Importantly, the study found a nonsignificant difference in thromboembolism events between patients who were bridged and those who were not (0.0% vs 0.2%, respectively; P = .56), a very low overall thromboembolism event rate (0.2%), and a lack of correlation between events and risk-stratification category.¹⁷ In other words, all thromboembolic events occurred in the low- and moderate-risk groups, which include patients who do not warrant bridging under current guidelines. Clinically relevant bleeding occurred in 17 (0.9%) of 1812 patients studied. Notably, 15 (2.7%) of 555 patients receiving bridging suffered clinically relevant bleeding as compared with 2 (0.2%) of 1257 patients forgoing bridging anticoagulation.

The Bleeding Risk of Bridging Anticoagulation Often Outweighs the Potential Benefits

The early observational studies on LMWH bridging demonstrated that thromboembolic events are infrequent (0.4%-0.9%), whereas major bleeding events occur up to 7 times more often (0.7%-6.7%).^10-12 The BRIDGE trial demonstrated comparably low thromboembolic events (0.3%). In the patients treated with bridging LMWH, major bleeding (3.2%) occurred 10 times more frequently than thromboembolism.¹⁴ Likewise, in a VTE cohort study, Clark et al.¹⁷ demonstrated “a 17-fold higher risk of bleeding without a significant difference in the rate of recurrent VTE” in patients bridged with heparin as compared with those who were not. Considering that recurrent VTE and major bleeding events have similar case-fatality rates,¹⁸ these increases in major bleeding events without reductions in thromboembolic events unmistakably tip the risk–benefit balance sharply towards an increased risk of harm.

When is bridging anticoagulation potentially helpful?

First, determine whether periprocedural OAC interruption is necessary for patients on chronic OAC due to atrial fibrillation, VTE, or MHVs. Avoid unwarranted OAC interruption by discussing the need for OAC interruptions with the surgeon or proceduralist, especially if the surgery is associated with a low bleeding risk and the patient has a high thromboembolic risk. When a periprocedural OAC interruption is justified, bridging should be avoided in the majority of patients, especially those with low to moderate thromboembolic risk or increased bleeding risk according to current risk-stratification schema.^7,15,19

Periprocedural management of direct oral anticoagulants (DOACs) is different than that of warfarin. The duration of DOAC interruption is determined by the procedural bleeding risk, drug half-life, and a patient’s creatinine clearance. Although the pharmacokinetics of DOACs generally allow for brief interruptions (eg, 24-48 hours), longer interruptions (eg, 96-120 hours) are warranted prior to high bleeding risk procedures, when drug half-life is prolonged (ie, dabigatran), and in patients with renal impairment. Parenteral bridging anticoagulation is not recommended during brief DOAC interruptions, and substituting a DOAC in place of LMWH for bridging is not advised. The 2017 American College of Cardiology Expert Consensus Decision Pathway provides periprocedural OAC interruption guidance for atrial fibrillation, with many principles applicable to other OAC indications.¹⁵We developed an institutional guideline that provides clinicians a structured approach to bridging OAC that steers them away from inappropriate bridging and helps them make decisions when evidence is lacking. Shared decision-making represents another effective method for well-informed patients and clinicians to arrive at a mutually agreed upon bridging decision.

• Avoid unnecessary periprocedural interruptions of OAC, especially for procedures with a low bleeding risk.
• Avoid the administration of bridging anticoagulation in patients with low to moderate thromboembolic risk during periprocedural OAC interruptions.
• In patients with a high thromboembolic risk, an individualized assessment of the patient-specific and procedure-specific bleeding risks versus the thromboembolic risks is necessary when considering bridging anticoagulation administration.

The “Things We Do for No Reason” (TWDFNR) series reviews practices that have become common parts of hospital care but that may provide little value to our patients. Practices reviewed in the TWDFNR series do not represent “black and white” conclusions or clinical practice standards, but are meant as a starting place for research and active discussions among hospitalists and patients. We invite you to be part of that discussion. https://www.choosingwisely.org/

Oral anticoagulation (OAC) is commonly prescribed to patients with atrial fibrillation, venous thromboembolism (VTE), and mechanical heart valves (MHVs) for primary and secondary thromboembolism prevention. When patients require surgery or an invasive procedure, “bridging” anticoagulants (eg, enoxaparin) are commonly administered during the period of OAC interruption to reduce thromboembolic risk. This practice stems from small observational studies and expert opinion, which influenced several clinical guidelines despite the lack of high-quality evidence. Although prospective randomized trials of periprocedural bridging in patients with VTE and MHVs are lacking, available evidence is consistent with findings from the BRIDGE trial, which guides the following general recommendations: (1) avoid unnecessary periprocedural interruptions of OAC, especially for low bleeding risk procedures; (2) avoid the administration of periprocedural bridging anticoagulation in patients with low to moderate thromboembolic risk; (3) in patients with high thromboembolic risk, individually assess the patient-specific and procedure-specific bleeding risks versus thromboembolic risks.

A 75-year-old man with a history of hypertension, diabetes mellitus, and atrial fibrillation is admitted for surgical repair of a comminuted intertrochanteric left hip fracture. He suffered a mechanical ground-level fall without loss of consciousness. At baseline, he denies any chest pain, dyspnea on exertion, or recent change in his exercise tolerance. A physical examination is notable for stable vital signs, irregular cardiac rhythm, and a shortened and externally rotated left lower extremity with exquisite tenderness to palpation and range of motion. The patient is taking warfarin for stroke prophylaxis based on a CHA₂DS₂VaSc score of 4 points. The international normalized ratio (INR) is 1.9 upon admission, and surgery is planned within 48 hours, once the patient is “medically cleared.” Will this patient benefit from periprocedural bridging anticoagulation?

OAC is commonly prescribed to patients with atrial fibrillation, venous thromboembolism (VTE), and mechanical heart valves (MHVs) for the primary or secondary prevention of thromboembolic events, with more than 35 million prescriptions written annually in the United States alone.¹ Many of these patients will require a temporary interruption of their OAC for surgery or an invasive procedure.² As a result, patients may be treated with short-acting, or “bridging,” anticoagulants, such as low-molecular-weight heparin (LMWH), to minimize the duration of anticoagulation interruption and theoretically reduce their thromboembolic risk. The rationale for bridging stemmed from small observational studies and expert opinion that perceived the estimated thromboembolic risk to be higher than the estimated bleeding risk.^3-5 One such example estimated that the VTE risk increased 100-fold postoperatively, whereas heparin administration only doubled the bleeding risk.³ Furthermore, clinical practice guidelines published from the American Heart Association, American College of Cardiology, European Heart Rhythm Society, and American College of Chest Physicians recommend when and how to initiate bridging anticoagulation. Clinicians have widely adopted these recommendations despite an acknowledged paucity of high-quality supporting evidence.^6,7

Periprocedural Anticoagulation Interruption is Often Not Indicated

Patients undergoing a surgical or invasive procedure may require an interruption of OAC to minimize the periprocedural bleeding risk. The decision to interrupt OAC should generally be based on the procedure-specific bleeding risk. Procedures with low bleeding risk such as cataract surgery, dermatologic biopsy (including Mohs), arthrocentesis, diagnostic gastrointestinal endoscopy, and cardiac pacemaker implantation can be performed safely without OAC interruption.^5,7 Despite evidence supporting the safety of periprocedural OAC continuation, unnecessary OAC interruptions remain commonplace and are associated with increased adverse outcomes.⁸ The BRUISE CONTROL trial compared uninterrupted OAC to interrupted OAC with periprocedural bridging for cardiac pacemaker or defibrillator implantation in a moderate to high thromboembolic risk population. The uninterrupted OAC group experienced significantly fewer pocket hematomas, hematoma evacuations, and prolonged hospitalizations (relative risk [RR] 0.19-0.24; P < .05) without significantly increased thromboembolic events, highlighting the potential benefits of this approach.⁹

Nevertheless, many surgical and invasive procedures do warrant OAC interruption due to the inherent bleeding risk of the procedure or other logistical considerations. Procedures associated with an increased bleeding risk include urologic surgery (except laser lithotripsy), surgery on highly vascular organs (eg, kidney, liver, spleen), bowel resection, cardiac surgery, and intracranial or spinal surgery.⁷ Alternatively, some procedures with acceptably low bleeding risk (eg, colonoscopy) are routinely performed during an OAC interruption due to the fact that a high bleeding risk intervention may be necessary during the procedure (eg, polypectomy). This approach may be preferable when a significant amount of preparation is required (eg, bowel preparation) and may be a more efficient use of healthcare resources by avoiding repeat procedures.

Bridging Anticoagulation Does Not Significantly Reduce Thromboembolic Events

BRIDGE was a randomized, double-blind, placebo-controlled trial among patients with atrial fibrillation (n = 1884) requiring OAC interruption for mostly low-risk, ambulatory surgeries or invasive procedures (eg, gastrointestinal endoscopy, cardiac catheterization). Notably, thromboembolism events were rare, and there was no significant difference in thromboembolism events between patients randomized to placebo or bridging with LMWH (0.4% vs 0.3%, respectively; P = .73).¹⁴ However, the proportion of patients enrolled with the highest thromboembolic risk (ie, CHADS₂ score 5-6 or prior transient ischemic attack and/or stroke) was low, potentially indicating an underestimated benefit in these patients. Major bleeding was significantly reduced in patients forgoing bridging anticoagulation (1.3% vs 3.2%; RR 0.41; 95% confidence interval, 0.20-0.78; P = .005), although bleeding occurred more frequently than thromboembolism in both groups.

Even though randomized trials assessing the safety and efficacy of bridging for VTE or MHVs have not been completed, evidence is not entirely lacking.^16,17 A rigorous observational study limited to a VTE cohort (deep vein thrombosis of upper or lower extremity and/or pulmonary embolism) analyzed the effects of bridging in patients with a surgical or invasive procedure-related OAC interruption. Patients were stratified according to the American College of Chest Physicians perioperative guideline risk-stratification schema, and most VTE events (≥93%) occurred more than 12 months prior to OAC interruption.⁷ Importantly, the study found a nonsignificant difference in thromboembolism events between patients who were bridged and those who were not (0.0% vs 0.2%, respectively; P = .56), a very low overall thromboembolism event rate (0.2%), and a lack of correlation between events and risk-stratification category.¹⁷ In other words, all thromboembolic events occurred in the low- and moderate-risk groups, which include patients who do not warrant bridging under current guidelines. Clinically relevant bleeding occurred in 17 (0.9%) of 1812 patients studied. Notably, 15 (2.7%) of 555 patients receiving bridging suffered clinically relevant bleeding as compared with 2 (0.2%) of 1257 patients forgoing bridging anticoagulation.

The Bleeding Risk of Bridging Anticoagulation Often Outweighs the Potential Benefits

The early observational studies on LMWH bridging demonstrated that thromboembolic events are infrequent (0.4%-0.9%), whereas major bleeding events occur up to 7 times more often (0.7%-6.7%).^10-12 The BRIDGE trial demonstrated comparably low thromboembolic events (0.3%). In the patients treated with bridging LMWH, major bleeding (3.2%) occurred 10 times more frequently than thromboembolism.¹⁴ Likewise, in a VTE cohort study, Clark et al.¹⁷ demonstrated “a 17-fold higher risk of bleeding without a significant difference in the rate of recurrent VTE” in patients bridged with heparin as compared with those who were not. Considering that recurrent VTE and major bleeding events have similar case-fatality rates,¹⁸ these increases in major bleeding events without reductions in thromboembolic events unmistakably tip the risk–benefit balance sharply towards an increased risk of harm.

When is bridging anticoagulation potentially helpful?

First, determine whether periprocedural OAC interruption is necessary for patients on chronic OAC due to atrial fibrillation, VTE, or MHVs. Avoid unwarranted OAC interruption by discussing the need for OAC interruptions with the surgeon or proceduralist, especially if the surgery is associated with a low bleeding risk and the patient has a high thromboembolic risk. When a periprocedural OAC interruption is justified, bridging should be avoided in the majority of patients, especially those with low to moderate thromboembolic risk or increased bleeding risk according to current risk-stratification schema.^7,15,19

Periprocedural management of direct oral anticoagulants (DOACs) is different than that of warfarin. The duration of DOAC interruption is determined by the procedural bleeding risk, drug half-life, and a patient’s creatinine clearance. Although the pharmacokinetics of DOACs generally allow for brief interruptions (eg, 24-48 hours), longer interruptions (eg, 96-120 hours) are warranted prior to high bleeding risk procedures, when drug half-life is prolonged (ie, dabigatran), and in patients with renal impairment. Parenteral bridging anticoagulation is not recommended during brief DOAC interruptions, and substituting a DOAC in place of LMWH for bridging is not advised. The 2017 American College of Cardiology Expert Consensus Decision Pathway provides periprocedural OAC interruption guidance for atrial fibrillation, with many principles applicable to other OAC indications.¹⁵We developed an institutional guideline that provides clinicians a structured approach to bridging OAC that steers them away from inappropriate bridging and helps them make decisions when evidence is lacking. Shared decision-making represents another effective method for well-informed patients and clinicians to arrive at a mutually agreed upon bridging decision.

• Avoid unnecessary periprocedural interruptions of OAC, especially for procedures with a low bleeding risk.
• Avoid the administration of bridging anticoagulation in patients with low to moderate thromboembolic risk during periprocedural OAC interruptions.
• In patients with a high thromboembolic risk, an individualized assessment of the patient-specific and procedure-specific bleeding risks versus the thromboembolic risks is necessary when considering bridging anticoagulation administration.

The “Things We Do for No Reason” (TWDFNR) series reviews practices that have become common parts of hospital care but that may provide little value to our patients. Practices reviewed in the TWDFNR series do not represent “black and white” conclusions or clinical practice standards, but are meant as a starting place for research and active discussions among hospitalists and patients. We invite you to be part of that discussion. https://www.choosingwisely.org/

Oral anticoagulation (OAC) is commonly prescribed to patients with atrial fibrillation, venous thromboembolism (VTE), and mechanical heart valves (MHVs) for primary and secondary thromboembolism prevention. When patients require surgery or an invasive procedure, “bridging” anticoagulants (eg, enoxaparin) are commonly administered during the period of OAC interruption to reduce thromboembolic risk. This practice stems from small observational studies and expert opinion, which influenced several clinical guidelines despite the lack of high-quality evidence. Although prospective randomized trials of periprocedural bridging in patients with VTE and MHVs are lacking, available evidence is consistent with findings from the BRIDGE trial, which guides the following general recommendations: (1) avoid unnecessary periprocedural interruptions of OAC, especially for low bleeding risk procedures; (2) avoid the administration of periprocedural bridging anticoagulation in patients with low to moderate thromboembolic risk; (3) in patients with high thromboembolic risk, individually assess the patient-specific and procedure-specific bleeding risks versus thromboembolic risks.

A 75-year-old man with a history of hypertension, diabetes mellitus, and atrial fibrillation is admitted for surgical repair of a comminuted intertrochanteric left hip fracture. He suffered a mechanical ground-level fall without loss of consciousness. At baseline, he denies any chest pain, dyspnea on exertion, or recent change in his exercise tolerance. A physical examination is notable for stable vital signs, irregular cardiac rhythm, and a shortened and externally rotated left lower extremity with exquisite tenderness to palpation and range of motion. The patient is taking warfarin for stroke prophylaxis based on a CHA₂DS₂VaSc score of 4 points. The international normalized ratio (INR) is 1.9 upon admission, and surgery is planned within 48 hours, once the patient is “medically cleared.” Will this patient benefit from periprocedural bridging anticoagulation?

OAC is commonly prescribed to patients with atrial fibrillation, venous thromboembolism (VTE), and mechanical heart valves (MHVs) for the primary or secondary prevention of thromboembolic events, with more than 35 million prescriptions written annually in the United States alone.¹ Many of these patients will require a temporary interruption of their OAC for surgery or an invasive procedure.² As a result, patients may be treated with short-acting, or “bridging,” anticoagulants, such as low-molecular-weight heparin (LMWH), to minimize the duration of anticoagulation interruption and theoretically reduce their thromboembolic risk. The rationale for bridging stemmed from small observational studies and expert opinion that perceived the estimated thromboembolic risk to be higher than the estimated bleeding risk.^3-5 One such example estimated that the VTE risk increased 100-fold postoperatively, whereas heparin administration only doubled the bleeding risk.³ Furthermore, clinical practice guidelines published from the American Heart Association, American College of Cardiology, European Heart Rhythm Society, and American College of Chest Physicians recommend when and how to initiate bridging anticoagulation. Clinicians have widely adopted these recommendations despite an acknowledged paucity of high-quality supporting evidence.^6,7

Periprocedural Anticoagulation Interruption is Often Not Indicated

Patients undergoing a surgical or invasive procedure may require an interruption of OAC to minimize the periprocedural bleeding risk. The decision to interrupt OAC should generally be based on the procedure-specific bleeding risk. Procedures with low bleeding risk such as cataract surgery, dermatologic biopsy (including Mohs), arthrocentesis, diagnostic gastrointestinal endoscopy, and cardiac pacemaker implantation can be performed safely without OAC interruption.^5,7 Despite evidence supporting the safety of periprocedural OAC continuation, unnecessary OAC interruptions remain commonplace and are associated with increased adverse outcomes.⁸ The BRUISE CONTROL trial compared uninterrupted OAC to interrupted OAC with periprocedural bridging for cardiac pacemaker or defibrillator implantation in a moderate to high thromboembolic risk population. The uninterrupted OAC group experienced significantly fewer pocket hematomas, hematoma evacuations, and prolonged hospitalizations (relative risk [RR] 0.19-0.24; P < .05) without significantly increased thromboembolic events, highlighting the potential benefits of this approach.⁹

Nevertheless, many surgical and invasive procedures do warrant OAC interruption due to the inherent bleeding risk of the procedure or other logistical considerations. Procedures associated with an increased bleeding risk include urologic surgery (except laser lithotripsy), surgery on highly vascular organs (eg, kidney, liver, spleen), bowel resection, cardiac surgery, and intracranial or spinal surgery.⁷ Alternatively, some procedures with acceptably low bleeding risk (eg, colonoscopy) are routinely performed during an OAC interruption due to the fact that a high bleeding risk intervention may be necessary during the procedure (eg, polypectomy). This approach may be preferable when a significant amount of preparation is required (eg, bowel preparation) and may be a more efficient use of healthcare resources by avoiding repeat procedures.

Bridging Anticoagulation Does Not Significantly Reduce Thromboembolic Events

BRIDGE was a randomized, double-blind, placebo-controlled trial among patients with atrial fibrillation (n = 1884) requiring OAC interruption for mostly low-risk, ambulatory surgeries or invasive procedures (eg, gastrointestinal endoscopy, cardiac catheterization). Notably, thromboembolism events were rare, and there was no significant difference in thromboembolism events between patients randomized to placebo or bridging with LMWH (0.4% vs 0.3%, respectively; P = .73).¹⁴ However, the proportion of patients enrolled with the highest thromboembolic risk (ie, CHADS₂ score 5-6 or prior transient ischemic attack and/or stroke) was low, potentially indicating an underestimated benefit in these patients. Major bleeding was significantly reduced in patients forgoing bridging anticoagulation (1.3% vs 3.2%; RR 0.41; 95% confidence interval, 0.20-0.78; P = .005), although bleeding occurred more frequently than thromboembolism in both groups.

Even though randomized trials assessing the safety and efficacy of bridging for VTE or MHVs have not been completed, evidence is not entirely lacking.^16,17 A rigorous observational study limited to a VTE cohort (deep vein thrombosis of upper or lower extremity and/or pulmonary embolism) analyzed the effects of bridging in patients with a surgical or invasive procedure-related OAC interruption. Patients were stratified according to the American College of Chest Physicians perioperative guideline risk-stratification schema, and most VTE events (≥93%) occurred more than 12 months prior to OAC interruption.⁷ Importantly, the study found a nonsignificant difference in thromboembolism events between patients who were bridged and those who were not (0.0% vs 0.2%, respectively; P = .56), a very low overall thromboembolism event rate (0.2%), and a lack of correlation between events and risk-stratification category.¹⁷ In other words, all thromboembolic events occurred in the low- and moderate-risk groups, which include patients who do not warrant bridging under current guidelines. Clinically relevant bleeding occurred in 17 (0.9%) of 1812 patients studied. Notably, 15 (2.7%) of 555 patients receiving bridging suffered clinically relevant bleeding as compared with 2 (0.2%) of 1257 patients forgoing bridging anticoagulation.

The Bleeding Risk of Bridging Anticoagulation Often Outweighs the Potential Benefits

The early observational studies on LMWH bridging demonstrated that thromboembolic events are infrequent (0.4%-0.9%), whereas major bleeding events occur up to 7 times more often (0.7%-6.7%).^10-12 The BRIDGE trial demonstrated comparably low thromboembolic events (0.3%). In the patients treated with bridging LMWH, major bleeding (3.2%) occurred 10 times more frequently than thromboembolism.¹⁴ Likewise, in a VTE cohort study, Clark et al.¹⁷ demonstrated “a 17-fold higher risk of bleeding without a significant difference in the rate of recurrent VTE” in patients bridged with heparin as compared with those who were not. Considering that recurrent VTE and major bleeding events have similar case-fatality rates,¹⁸ these increases in major bleeding events without reductions in thromboembolic events unmistakably tip the risk–benefit balance sharply towards an increased risk of harm.

When is bridging anticoagulation potentially helpful?

First, determine whether periprocedural OAC interruption is necessary for patients on chronic OAC due to atrial fibrillation, VTE, or MHVs. Avoid unwarranted OAC interruption by discussing the need for OAC interruptions with the surgeon or proceduralist, especially if the surgery is associated with a low bleeding risk and the patient has a high thromboembolic risk. When a periprocedural OAC interruption is justified, bridging should be avoided in the majority of patients, especially those with low to moderate thromboembolic risk or increased bleeding risk according to current risk-stratification schema.^7,15,19

Periprocedural management of direct oral anticoagulants (DOACs) is different than that of warfarin. The duration of DOAC interruption is determined by the procedural bleeding risk, drug half-life, and a patient’s creatinine clearance. Although the pharmacokinetics of DOACs generally allow for brief interruptions (eg, 24-48 hours), longer interruptions (eg, 96-120 hours) are warranted prior to high bleeding risk procedures, when drug half-life is prolonged (ie, dabigatran), and in patients with renal impairment. Parenteral bridging anticoagulation is not recommended during brief DOAC interruptions, and substituting a DOAC in place of LMWH for bridging is not advised. The 2017 American College of Cardiology Expert Consensus Decision Pathway provides periprocedural OAC interruption guidance for atrial fibrillation, with many principles applicable to other OAC indications.¹⁵We developed an institutional guideline that provides clinicians a structured approach to bridging OAC that steers them away from inappropriate bridging and helps them make decisions when evidence is lacking. Shared decision-making represents another effective method for well-informed patients and clinicians to arrive at a mutually agreed upon bridging decision.

• Avoid unnecessary periprocedural interruptions of OAC, especially for procedures with a low bleeding risk.
• Avoid the administration of bridging anticoagulation in patients with low to moderate thromboembolic risk during periprocedural OAC interruptions.
• In patients with a high thromboembolic risk, an individualized assessment of the patient-specific and procedure-specific bleeding risks versus the thromboembolic risks is necessary when considering bridging anticoagulation administration.

The “Things We Do for No Reason” (TWDFNR) series reviews practices that have become common parts of hospital care but that may provide little value to our patients. Practices reviewed in the TWDFNR series do not represent “black and white” conclusions or clinical practice standards, but are meant as a starting place for research and active discussions among hospitalists and patients. We invite you to be part of that discussion. https://www.choosingwisely.org/

Acute gastroenteritis (AGE) remains a substantial cause of childhood illness and is 1 of the top 10 reasons for pediatric hospitalization nationwide. In the United States, AGE is responsible for 10% of hospital admissions and approximately 300 deaths annually.¹ The American Academy of Pediatrics (AAP) and other organizations have emphasized supportive care in the management of AGE. Routine diagnostic testing has been discouraged in national guidelines except in cases of severe dehydration or an otherwise complicated course. Despite AGE guidelines, diagnostic laboratory tests are still widely used even though they have been shown to be poor predictors of dehydration. Studies have shown that high test utilization in various pediatric disease processes often influences the decision for hospitalization without improvement in patient outcome. In children with AGE, the initial and follow-up laboratory tests may not only be something that we do for no reason, but something that is associated with more risk than benefit.

An 18-month-old healthy male is brought to the emergency department (ED) with a chief complaint of 2 days of nonbloody, nonbilious emesis and watery diarrhea. He has decreased energy but smiles and plays for a few minutes. He has had decreased wet diapers. His exam is notable for mild tachycardia, mildly dry lips, and capillary refill of 3 seconds. A serum electrolyte panel is normal except for a sodium of 134 mEq/L, a bicarbonate of 16 mEq/L, and an anion gap of 18, which are flagged as abnormal by the electronic medical record. These results prompt intravenous (IV) access, a normal saline bolus, and admission on maintenance fluids overnight. The next morning, his electrolyte panel is repeated, and his sodium is 140 mEq/L and bicarbonate is 15 mEq/L. He is now drinking well with no further episodes of emesis, so he is discharged home.

Many physicians across the United States continue to order electrolytes in AGE as a way to avoid missing severe dehydration, severe electrolyte abnormalities, or rare diagnoses, such as adrenal insufficiency or new-onset diabetes, in a child. Previous studies have revealed that bicarbonate and blood urea nitrogen (BUN) may be helpful predictors of severe dehydration. A retrospective study of 168 patients by Yilmaz et al.² showed that BUN and bicarbonate strongly correlated with dehydration severity (P < 0.00001 and P = 0.01, respectively). A 97-patient prospective study by Vega and Avner³ showed that bicarbonate <17 can help in predicting percent body weight loss (PBWL) (sensitivity of 77% for PBWL 6-10 and 94% for PBWL >10).

In AGE, obtaining laboratory data is often considered to be the more conservative approach. Some attribute this to the medical education and legal system rewarding the uncovering of rare diagnoses,⁴ while others believe physicians obtain laboratory data to avoid missing severe electrolyte disorders. One author notes, “physicians who are anxious about a patient’s problem may be tempted to do something—anything—decisive in order to diminish their own anxiety.”⁵ Severe electrolyte derangements are common in developing countries⁶ but less so in the United States. A prospective pediatric dehydration study over 1 year in the United States demonstrated rates of 6% and 3% of hypo- and hypernatremia, respectively (n = 182). Only 1 patient had a sodium level >160, and this patient had an underlying genetic syndrome, and none had hyponatremia <130. Hypoglycemia was the most common electrolyte abnormality, which was present in 9.8% of patients. Electrolyte results changed management in 10.4% of patients.⁷

In AGE with or without dehydration, guidelines from the AAP and other international organizations emphasize supportive care in the management of AGE and discourage routine diagnostic testing.^8-10 Yet, there continues to be wide variation in AGE management.^11-13 Most AGE cases presenting to an outpatient setting or ED are uncomplicated: age >6 months, nontoxic appearance, no comorbidities, no hematochezia, diarrhea <7 days, and mild-to-moderate dehydration.

Steiner et al.¹⁴ performed a systematic meta-analysis of the precision and accuracy of symptoms, signs, and laboratory tests for evaluating dehydration in children. They concluded that a standardized clinical assessment based on physical exam (PE) findings more accurately classifies the degree of dehydration than laboratory testing. Steiner et al¹⁴ specifically analyzed the works by Yilmaz et al.² and Vega and Avner,³ and determined that the positive likelihood ratios for >5% dehydration resulting from a BUN >45 or bicarbonate <17 were too small or had confidence intervals that were too wide to be clinically helpful alone. Therefore, Steiner et al.¹⁴ recommended that laboratory testing should not be considered definitive for dehydration.

Vega and Avner³ found that electrolyte testing is less helpful in distinguishing between <5% (mild) and 5% to 10% (moderate) dehydration compared to PBWL. Because both mild and moderate dehydration respond equally well to oral rehydration therapy (ORT),⁸ electrolyte testing is not helpful in managing these categories. Many studies have excluded children with hypernatremia, but generally, severe hypernatremia is uncommon in healthy patients with AGE. In most cases of mild hypernatremia, ORT is the preferred resuscitation method and is possibly safer than IV rehydration because ORT may induce less rapid shifts in intracellular water.¹⁵

Tieder et al.¹⁶ demonstrated that better hospital adherence to national recommendations to avoid diagnostic testing in children with AGE resulted in lower charges and equivalent outcomes. In this large, multicenter study among 27 children’s hospitals in the Pediatric Hospital Information System (PHIS) database, only 70% of the 188,000 patients received guideline-adherent care. Nonrecommended laboratory testing was common, especially in the admitted population. Electrolytes were measured in 22.1% of the ED and observation patients compared with 85% of admitted patients. Hospitals that were most guideline adherent in the ED demonstrated 50% lower charges. The authors estimate that standardizing AGE care and eliminating nonrecommended laboratory testing would decrease admissions by 45% and save more than $1 billion per year in direct medical costs.¹⁶ In a similar PHIS study, laboratory testing was strongly correlated with the percentage of children hospitalized for AGE at each hospital (r = 0.73, P < 0.001). Results were unchanged when excluding children <1 year of age (r = 0.75, P < 0.001). In contrast, the mean testing count was not correlated with return visits within 3 days for children discharged from the ED (r = 0.21, P = 0.235), nor was it correlated with hospital length of stay (r = −0.04, P = 0.804) or return visits within 7 days (r = 0.03, P = 0.862) for hospitalized children.¹² In addition, Freedman et al.¹⁷ revealed that the clinical dehydration score is independently associated with successful ED discharge without revisits, and laboratory testing does not prevent missed cases of severe dehydration.

Nonrecommended and often unnecessary laboratory testing in AGE results in IV procedures that are sometimes repeated because of abnormal values. “Shotgun testing,” or ordering a panel of labs, can result in abnormal laboratory values in healthy patients. Deyo et al.¹⁸ cite that for a panel of 12 laboratory values, there is a 46% chance of having at least 1 abnormal lab, even in healthy patients. These false-positive results can then drive further testing. In AGE, an abnormal bicarbonate may drive repeat testing to confirm normalization, but the bicarbonate may actually decrease once IV fluid therapy is initiated due to excessive chloride in isotonic fluids. Coon et al.¹⁹ have shown that seemingly innocuous testing or screening can lead to overdiagnosis, which can cause physical and psychological harm to the patient and has financial implications for the family and healthcare system. While this has not been directly investigated in pediatric AGE, it has been studied in common pediatric illnesses, including pneumonia and urinary tract infections.^20,21 For children, venipuncture and IV placements are often the most distressful components of a hospital visit and can affect future healthcare encounters, making children anxious and distrustful of the healthcare system.^22,23

Electrolyte panels may be useful in assessing children with severe dehydration (scores of 5-8 on the Clinical Dehydration Scale (CDS) or more than 10% weight loss) or in complicated cases of AGE (those that do not meet the criteria of age >6 months, nontoxic appearance, no comorbidities, no hematochezia, and diarrhea <7 days) to guide IV fluid management and correct markedly abnormal electrolytes.¹⁴

Electrolyte panels may also rarely uncover disease processes, such as new-onset diabetes, hemolytic uremic syndrome, adrenal insufficiency, or inborn errors of metabolism, allowing for early diagnosis and preventing adverse outcomes. Suspicion to investigate such entities should arise during a thorough history and PE instead of routinely screening all children with symptoms of AGE. One should also have a higher level of concern for other disease processes when clinical recovery does not occur within the expected amount of time; symptoms usually resolve within 2 to 3 days but sometimes will last up to a week.

• Perform a thorough history and PE to diagnose AGE.⁸
• Clinical assessment of dehydration should be performed upon initial presentation and repeatedly with vital signs throughout the stay using a validated CDS to classify the patient’s initial dehydration severity and monitor improvement. Obtain a current patient weight and compare with previously recorded weights, if available.^25,26
• Laboratory testing in patients with AGE should not be performed unless a patient is classified as severely dehydrated, is toxic appearing, has a comorbidity that increases the likelihood of complications, or is not improving as expected.
• Rehydration via ORT is preferred to an IV in mild and moderate dehydration.¹⁵
• If initial testing is performed and indicates an expected value indicative of dehydration, do not repeat testing to demonstrate normalization as long as the child is clinically improving as expected.

Children presenting with mild-to-moderate dehydration should be treated with supportive measures in accordance with current guidelines. Electrolyte panels very rarely provide clinical information that cannot be garnered through a thorough history and PE. As in our clinical scenario, the laboratory values obtained may have led to potential harm, including overdiagnosis, painful procedures, and psychological distress. Without testing, the patient likely could have been appropriately treated with ORT and discharged from the ED.

Do you think this is a low-value practice? Is this truly a “Thing We Do for No Reason?” Share what you do in your practice and join in the conversation online by retweeting it on Twitter (#TWDFNR) and liking it on Facebook. We invite you to propose ideas for other “Things We Do for No Reason” topics by emailing TWDFNR@hospitalmedicine.org.

Disclosure

The authors have nothing to disclose.

The “Things We Do for No Reason” (TWDFNR) series reviews practices that have become common parts of hospital care but that may provide little value to our patients. Practices reviewed in the TWDFNR series do not represent “black and white” conclusions or clinical practice standards, but are meant as a starting place for research and active discussions among hospitalists and patients. We invite you to be part of that discussion. https://www.choosingwisely.org/

Acute gastroenteritis (AGE) remains a substantial cause of childhood illness and is 1 of the top 10 reasons for pediatric hospitalization nationwide. In the United States, AGE is responsible for 10% of hospital admissions and approximately 300 deaths annually.¹ The American Academy of Pediatrics (AAP) and other organizations have emphasized supportive care in the management of AGE. Routine diagnostic testing has been discouraged in national guidelines except in cases of severe dehydration or an otherwise complicated course. Despite AGE guidelines, diagnostic laboratory tests are still widely used even though they have been shown to be poor predictors of dehydration. Studies have shown that high test utilization in various pediatric disease processes often influences the decision for hospitalization without improvement in patient outcome. In children with AGE, the initial and follow-up laboratory tests may not only be something that we do for no reason, but something that is associated with more risk than benefit.

An 18-month-old healthy male is brought to the emergency department (ED) with a chief complaint of 2 days of nonbloody, nonbilious emesis and watery diarrhea. He has decreased energy but smiles and plays for a few minutes. He has had decreased wet diapers. His exam is notable for mild tachycardia, mildly dry lips, and capillary refill of 3 seconds. A serum electrolyte panel is normal except for a sodium of 134 mEq/L, a bicarbonate of 16 mEq/L, and an anion gap of 18, which are flagged as abnormal by the electronic medical record. These results prompt intravenous (IV) access, a normal saline bolus, and admission on maintenance fluids overnight. The next morning, his electrolyte panel is repeated, and his sodium is 140 mEq/L and bicarbonate is 15 mEq/L. He is now drinking well with no further episodes of emesis, so he is discharged home.

Many physicians across the United States continue to order electrolytes in AGE as a way to avoid missing severe dehydration, severe electrolyte abnormalities, or rare diagnoses, such as adrenal insufficiency or new-onset diabetes, in a child. Previous studies have revealed that bicarbonate and blood urea nitrogen (BUN) may be helpful predictors of severe dehydration. A retrospective study of 168 patients by Yilmaz et al.² showed that BUN and bicarbonate strongly correlated with dehydration severity (P < 0.00001 and P = 0.01, respectively). A 97-patient prospective study by Vega and Avner³ showed that bicarbonate <17 can help in predicting percent body weight loss (PBWL) (sensitivity of 77% for PBWL 6-10 and 94% for PBWL >10).

In AGE, obtaining laboratory data is often considered to be the more conservative approach. Some attribute this to the medical education and legal system rewarding the uncovering of rare diagnoses,⁴ while others believe physicians obtain laboratory data to avoid missing severe electrolyte disorders. One author notes, “physicians who are anxious about a patient’s problem may be tempted to do something—anything—decisive in order to diminish their own anxiety.”⁵ Severe electrolyte derangements are common in developing countries⁶ but less so in the United States. A prospective pediatric dehydration study over 1 year in the United States demonstrated rates of 6% and 3% of hypo- and hypernatremia, respectively (n = 182). Only 1 patient had a sodium level >160, and this patient had an underlying genetic syndrome, and none had hyponatremia <130. Hypoglycemia was the most common electrolyte abnormality, which was present in 9.8% of patients. Electrolyte results changed management in 10.4% of patients.⁷

In AGE with or without dehydration, guidelines from the AAP and other international organizations emphasize supportive care in the management of AGE and discourage routine diagnostic testing.^8-10 Yet, there continues to be wide variation in AGE management.^11-13 Most AGE cases presenting to an outpatient setting or ED are uncomplicated: age >6 months, nontoxic appearance, no comorbidities, no hematochezia, diarrhea <7 days, and mild-to-moderate dehydration.

Steiner et al.¹⁴ performed a systematic meta-analysis of the precision and accuracy of symptoms, signs, and laboratory tests for evaluating dehydration in children. They concluded that a standardized clinical assessment based on physical exam (PE) findings more accurately classifies the degree of dehydration than laboratory testing. Steiner et al¹⁴ specifically analyzed the works by Yilmaz et al.² and Vega and Avner,³ and determined that the positive likelihood ratios for >5% dehydration resulting from a BUN >45 or bicarbonate <17 were too small or had confidence intervals that were too wide to be clinically helpful alone. Therefore, Steiner et al.¹⁴ recommended that laboratory testing should not be considered definitive for dehydration.

Vega and Avner³ found that electrolyte testing is less helpful in distinguishing between <5% (mild) and 5% to 10% (moderate) dehydration compared to PBWL. Because both mild and moderate dehydration respond equally well to oral rehydration therapy (ORT),⁸ electrolyte testing is not helpful in managing these categories. Many studies have excluded children with hypernatremia, but generally, severe hypernatremia is uncommon in healthy patients with AGE. In most cases of mild hypernatremia, ORT is the preferred resuscitation method and is possibly safer than IV rehydration because ORT may induce less rapid shifts in intracellular water.¹⁵

Tieder et al.¹⁶ demonstrated that better hospital adherence to national recommendations to avoid diagnostic testing in children with AGE resulted in lower charges and equivalent outcomes. In this large, multicenter study among 27 children’s hospitals in the Pediatric Hospital Information System (PHIS) database, only 70% of the 188,000 patients received guideline-adherent care. Nonrecommended laboratory testing was common, especially in the admitted population. Electrolytes were measured in 22.1% of the ED and observation patients compared with 85% of admitted patients. Hospitals that were most guideline adherent in the ED demonstrated 50% lower charges. The authors estimate that standardizing AGE care and eliminating nonrecommended laboratory testing would decrease admissions by 45% and save more than $1 billion per year in direct medical costs.¹⁶ In a similar PHIS study, laboratory testing was strongly correlated with the percentage of children hospitalized for AGE at each hospital (r = 0.73, P < 0.001). Results were unchanged when excluding children <1 year of age (r = 0.75, P < 0.001). In contrast, the mean testing count was not correlated with return visits within 3 days for children discharged from the ED (r = 0.21, P = 0.235), nor was it correlated with hospital length of stay (r = −0.04, P = 0.804) or return visits within 7 days (r = 0.03, P = 0.862) for hospitalized children.¹² In addition, Freedman et al.¹⁷ revealed that the clinical dehydration score is independently associated with successful ED discharge without revisits, and laboratory testing does not prevent missed cases of severe dehydration.

Nonrecommended and often unnecessary laboratory testing in AGE results in IV procedures that are sometimes repeated because of abnormal values. “Shotgun testing,” or ordering a panel of labs, can result in abnormal laboratory values in healthy patients. Deyo et al.¹⁸ cite that for a panel of 12 laboratory values, there is a 46% chance of having at least 1 abnormal lab, even in healthy patients. These false-positive results can then drive further testing. In AGE, an abnormal bicarbonate may drive repeat testing to confirm normalization, but the bicarbonate may actually decrease once IV fluid therapy is initiated due to excessive chloride in isotonic fluids. Coon et al.¹⁹ have shown that seemingly innocuous testing or screening can lead to overdiagnosis, which can cause physical and psychological harm to the patient and has financial implications for the family and healthcare system. While this has not been directly investigated in pediatric AGE, it has been studied in common pediatric illnesses, including pneumonia and urinary tract infections.^20,21 For children, venipuncture and IV placements are often the most distressful components of a hospital visit and can affect future healthcare encounters, making children anxious and distrustful of the healthcare system.^22,23

Electrolyte panels may be useful in assessing children with severe dehydration (scores of 5-8 on the Clinical Dehydration Scale (CDS) or more than 10% weight loss) or in complicated cases of AGE (those that do not meet the criteria of age >6 months, nontoxic appearance, no comorbidities, no hematochezia, and diarrhea <7 days) to guide IV fluid management and correct markedly abnormal electrolytes.¹⁴

Electrolyte panels may also rarely uncover disease processes, such as new-onset diabetes, hemolytic uremic syndrome, adrenal insufficiency, or inborn errors of metabolism, allowing for early diagnosis and preventing adverse outcomes. Suspicion to investigate such entities should arise during a thorough history and PE instead of routinely screening all children with symptoms of AGE. One should also have a higher level of concern for other disease processes when clinical recovery does not occur within the expected amount of time; symptoms usually resolve within 2 to 3 days but sometimes will last up to a week.

• Perform a thorough history and PE to diagnose AGE.⁸
• Clinical assessment of dehydration should be performed upon initial presentation and repeatedly with vital signs throughout the stay using a validated CDS to classify the patient’s initial dehydration severity and monitor improvement. Obtain a current patient weight and compare with previously recorded weights, if available.^25,26
• Laboratory testing in patients with AGE should not be performed unless a patient is classified as severely dehydrated, is toxic appearing, has a comorbidity that increases the likelihood of complications, or is not improving as expected.
• Rehydration via ORT is preferred to an IV in mild and moderate dehydration.¹⁵
• If initial testing is performed and indicates an expected value indicative of dehydration, do not repeat testing to demonstrate normalization as long as the child is clinically improving as expected.

Children presenting with mild-to-moderate dehydration should be treated with supportive measures in accordance with current guidelines. Electrolyte panels very rarely provide clinical information that cannot be garnered through a thorough history and PE. As in our clinical scenario, the laboratory values obtained may have led to potential harm, including overdiagnosis, painful procedures, and psychological distress. Without testing, the patient likely could have been appropriately treated with ORT and discharged from the ED.

Do you think this is a low-value practice? Is this truly a “Thing We Do for No Reason?” Share what you do in your practice and join in the conversation online by retweeting it on Twitter (#TWDFNR) and liking it on Facebook. We invite you to propose ideas for other “Things We Do for No Reason” topics by emailing TWDFNR@hospitalmedicine.org.

Disclosure

The authors have nothing to disclose.

The “Things We Do for No Reason” (TWDFNR) series reviews practices that have become common parts of hospital care but that may provide little value to our patients. Practices reviewed in the TWDFNR series do not represent “black and white” conclusions or clinical practice standards, but are meant as a starting place for research and active discussions among hospitalists and patients. We invite you to be part of that discussion. https://www.choosingwisely.org/

Acute gastroenteritis (AGE) remains a substantial cause of childhood illness and is 1 of the top 10 reasons for pediatric hospitalization nationwide. In the United States, AGE is responsible for 10% of hospital admissions and approximately 300 deaths annually.¹ The American Academy of Pediatrics (AAP) and other organizations have emphasized supportive care in the management of AGE. Routine diagnostic testing has been discouraged in national guidelines except in cases of severe dehydration or an otherwise complicated course. Despite AGE guidelines, diagnostic laboratory tests are still widely used even though they have been shown to be poor predictors of dehydration. Studies have shown that high test utilization in various pediatric disease processes often influences the decision for hospitalization without improvement in patient outcome. In children with AGE, the initial and follow-up laboratory tests may not only be something that we do for no reason, but something that is associated with more risk than benefit.

An 18-month-old healthy male is brought to the emergency department (ED) with a chief complaint of 2 days of nonbloody, nonbilious emesis and watery diarrhea. He has decreased energy but smiles and plays for a few minutes. He has had decreased wet diapers. His exam is notable for mild tachycardia, mildly dry lips, and capillary refill of 3 seconds. A serum electrolyte panel is normal except for a sodium of 134 mEq/L, a bicarbonate of 16 mEq/L, and an anion gap of 18, which are flagged as abnormal by the electronic medical record. These results prompt intravenous (IV) access, a normal saline bolus, and admission on maintenance fluids overnight. The next morning, his electrolyte panel is repeated, and his sodium is 140 mEq/L and bicarbonate is 15 mEq/L. He is now drinking well with no further episodes of emesis, so he is discharged home.

Many physicians across the United States continue to order electrolytes in AGE as a way to avoid missing severe dehydration, severe electrolyte abnormalities, or rare diagnoses, such as adrenal insufficiency or new-onset diabetes, in a child. Previous studies have revealed that bicarbonate and blood urea nitrogen (BUN) may be helpful predictors of severe dehydration. A retrospective study of 168 patients by Yilmaz et al.² showed that BUN and bicarbonate strongly correlated with dehydration severity (P < 0.00001 and P = 0.01, respectively). A 97-patient prospective study by Vega and Avner³ showed that bicarbonate <17 can help in predicting percent body weight loss (PBWL) (sensitivity of 77% for PBWL 6-10 and 94% for PBWL >10).

In AGE, obtaining laboratory data is often considered to be the more conservative approach. Some attribute this to the medical education and legal system rewarding the uncovering of rare diagnoses,⁴ while others believe physicians obtain laboratory data to avoid missing severe electrolyte disorders. One author notes, “physicians who are anxious about a patient’s problem may be tempted to do something—anything—decisive in order to diminish their own anxiety.”⁵ Severe electrolyte derangements are common in developing countries⁶ but less so in the United States. A prospective pediatric dehydration study over 1 year in the United States demonstrated rates of 6% and 3% of hypo- and hypernatremia, respectively (n = 182). Only 1 patient had a sodium level >160, and this patient had an underlying genetic syndrome, and none had hyponatremia <130. Hypoglycemia was the most common electrolyte abnormality, which was present in 9.8% of patients. Electrolyte results changed management in 10.4% of patients.⁷

In AGE with or without dehydration, guidelines from the AAP and other international organizations emphasize supportive care in the management of AGE and discourage routine diagnostic testing.^8-10 Yet, there continues to be wide variation in AGE management.^11-13 Most AGE cases presenting to an outpatient setting or ED are uncomplicated: age >6 months, nontoxic appearance, no comorbidities, no hematochezia, diarrhea <7 days, and mild-to-moderate dehydration.

Steiner et al.¹⁴ performed a systematic meta-analysis of the precision and accuracy of symptoms, signs, and laboratory tests for evaluating dehydration in children. They concluded that a standardized clinical assessment based on physical exam (PE) findings more accurately classifies the degree of dehydration than laboratory testing. Steiner et al¹⁴ specifically analyzed the works by Yilmaz et al.² and Vega and Avner,³ and determined that the positive likelihood ratios for >5% dehydration resulting from a BUN >45 or bicarbonate <17 were too small or had confidence intervals that were too wide to be clinically helpful alone. Therefore, Steiner et al.¹⁴ recommended that laboratory testing should not be considered definitive for dehydration.

Vega and Avner³ found that electrolyte testing is less helpful in distinguishing between <5% (mild) and 5% to 10% (moderate) dehydration compared to PBWL. Because both mild and moderate dehydration respond equally well to oral rehydration therapy (ORT),⁸ electrolyte testing is not helpful in managing these categories. Many studies have excluded children with hypernatremia, but generally, severe hypernatremia is uncommon in healthy patients with AGE. In most cases of mild hypernatremia, ORT is the preferred resuscitation method and is possibly safer than IV rehydration because ORT may induce less rapid shifts in intracellular water.¹⁵

Tieder et al.¹⁶ demonstrated that better hospital adherence to national recommendations to avoid diagnostic testing in children with AGE resulted in lower charges and equivalent outcomes. In this large, multicenter study among 27 children’s hospitals in the Pediatric Hospital Information System (PHIS) database, only 70% of the 188,000 patients received guideline-adherent care. Nonrecommended laboratory testing was common, especially in the admitted population. Electrolytes were measured in 22.1% of the ED and observation patients compared with 85% of admitted patients. Hospitals that were most guideline adherent in the ED demonstrated 50% lower charges. The authors estimate that standardizing AGE care and eliminating nonrecommended laboratory testing would decrease admissions by 45% and save more than $1 billion per year in direct medical costs.¹⁶ In a similar PHIS study, laboratory testing was strongly correlated with the percentage of children hospitalized for AGE at each hospital (r = 0.73, P < 0.001). Results were unchanged when excluding children <1 year of age (r = 0.75, P < 0.001). In contrast, the mean testing count was not correlated with return visits within 3 days for children discharged from the ED (r = 0.21, P = 0.235), nor was it correlated with hospital length of stay (r = −0.04, P = 0.804) or return visits within 7 days (r = 0.03, P = 0.862) for hospitalized children.¹² In addition, Freedman et al.¹⁷ revealed that the clinical dehydration score is independently associated with successful ED discharge without revisits, and laboratory testing does not prevent missed cases of severe dehydration.

Nonrecommended and often unnecessary laboratory testing in AGE results in IV procedures that are sometimes repeated because of abnormal values. “Shotgun testing,” or ordering a panel of labs, can result in abnormal laboratory values in healthy patients. Deyo et al.¹⁸ cite that for a panel of 12 laboratory values, there is a 46% chance of having at least 1 abnormal lab, even in healthy patients. These false-positive results can then drive further testing. In AGE, an abnormal bicarbonate may drive repeat testing to confirm normalization, but the bicarbonate may actually decrease once IV fluid therapy is initiated due to excessive chloride in isotonic fluids. Coon et al.¹⁹ have shown that seemingly innocuous testing or screening can lead to overdiagnosis, which can cause physical and psychological harm to the patient and has financial implications for the family and healthcare system. While this has not been directly investigated in pediatric AGE, it has been studied in common pediatric illnesses, including pneumonia and urinary tract infections.^20,21 For children, venipuncture and IV placements are often the most distressful components of a hospital visit and can affect future healthcare encounters, making children anxious and distrustful of the healthcare system.^22,23

Electrolyte panels may be useful in assessing children with severe dehydration (scores of 5-8 on the Clinical Dehydration Scale (CDS) or more than 10% weight loss) or in complicated cases of AGE (those that do not meet the criteria of age >6 months, nontoxic appearance, no comorbidities, no hematochezia, and diarrhea <7 days) to guide IV fluid management and correct markedly abnormal electrolytes.¹⁴

Electrolyte panels may also rarely uncover disease processes, such as new-onset diabetes, hemolytic uremic syndrome, adrenal insufficiency, or inborn errors of metabolism, allowing for early diagnosis and preventing adverse outcomes. Suspicion to investigate such entities should arise during a thorough history and PE instead of routinely screening all children with symptoms of AGE. One should also have a higher level of concern for other disease processes when clinical recovery does not occur within the expected amount of time; symptoms usually resolve within 2 to 3 days but sometimes will last up to a week.

• Perform a thorough history and PE to diagnose AGE.⁸
• Clinical assessment of dehydration should be performed upon initial presentation and repeatedly with vital signs throughout the stay using a validated CDS to classify the patient’s initial dehydration severity and monitor improvement. Obtain a current patient weight and compare with previously recorded weights, if available.^25,26
• Laboratory testing in patients with AGE should not be performed unless a patient is classified as severely dehydrated, is toxic appearing, has a comorbidity that increases the likelihood of complications, or is not improving as expected.
• Rehydration via ORT is preferred to an IV in mild and moderate dehydration.¹⁵
• If initial testing is performed and indicates an expected value indicative of dehydration, do not repeat testing to demonstrate normalization as long as the child is clinically improving as expected.

Children presenting with mild-to-moderate dehydration should be treated with supportive measures in accordance with current guidelines. Electrolyte panels very rarely provide clinical information that cannot be garnered through a thorough history and PE. As in our clinical scenario, the laboratory values obtained may have led to potential harm, including overdiagnosis, painful procedures, and psychological distress. Without testing, the patient likely could have been appropriately treated with ORT and discharged from the ED.

Do you think this is a low-value practice? Is this truly a “Thing We Do for No Reason?” Share what you do in your practice and join in the conversation online by retweeting it on Twitter (#TWDFNR) and liking it on Facebook. We invite you to propose ideas for other “Things We Do for No Reason” topics by emailing TWDFNR@hospitalmedicine.org.

Disclosure

The authors have nothing to disclose.