Matthew W McCarthy, MD, FACP

Community-acquired pneumonia (CAP) accounts for more than 1.5 million adult hospitalizations and 100,000 deaths each year in the United States.¹ Antibiotic overuse in the hospital setting is an important contributor to the rise of antibiotic resistance, prompting increased efforts to limit inappropriate antibiotic use in hospitals.² Procalcitonin, a precursor of the hormone calcitonin, is upregulated in bacterial infections and downregulated in viral infections. The US Food and Drug Administration has approved it as a serum biomarker to assist clinicians with decisions about using antibiotics.³

There is no consensus on how to best use procalcitonin in the management of CAP. We provide a practical update that includes a review of recent literature, added secondary analysis, and expert opinion surrounding the use of procalcitonin in the diagnosis and management of CAP in hospitalized adults.

Initial procalcitonin levels do not sufficiently exclude bacterial etiologies of CAP to withhold antibiotic prescription safely. The largest diagnostic accuracy study of procalcitonin in the diagnosis of CAP was a subanalysis of the Etiology of Pneumonia in the Community Study.⁴ A total of 1,735 adults hospitalized with CAP received procalcitonin testing along with systematic pathogen testing. The area under the receiver operating characteristic curve for procalcitonin in discriminating bacterial pathogens from viral pathogens was 0.73 (95% CI, 0.69-0.77). A procalcitonin cut-off of 0.1 ng/mL resulted in 80.9% (95% CI, 75.3%-85.7%) sensitivity and 51.6% (95% CI, 46.6%-56.5%) specificity for identification of any bacterial pathogen.

In a secondary analysis of this study, we calculated multilevel likelihood ratios (LRs) for ranges of procalcitonin values to determine the diagnostic accuracy of procalcitonin in distinguishing bacterial from viral etiologies of CAP (Table). Multilevel LRs offer more useful diagnostic information than dichotomizing at specified cut-points.⁵ A procalcitonin result less than 0.1 ng/mL has a negative LR of 0.4 (95% CI, 0.3-0.5), which is not low enough to rule out bacterial CAP effectively when starting with intermediate or high pretest probability. For a low result (<0.1 ng/mL) to be useful in ruling out bacterial CAP, for example having less than a 10% posttest probability of bacterial CAP, the pretest probability would have to be no greater than 22%. Even then, a 10% posttest probability of bacterial CAP may still be too high for clinicians to withhold initial antibiotics. For procalcitonin values between 0.1 ng/mL and 1.0 ng/mL, the probability of bacterial CAP does not change significantly, with an LR of 1.0 (95% CI, 0.8-1.3). Procalcitonin values up to 5 ng/mL reach a modest positive LR of 2.3 (95% CI, 0.8-4.3). Very high values, such as those >10 ng/mL, yield a positive LR of 5.5 (95% CI, 3.2-9.7), are potentially useful in decisions to initiate antibiotics in situations of very low pretest probability of bacterial CAP. For example, a 9% pretest probability of bacterial CAP is likely below many physicians’ threshold for starting antibiotics. A procalcitonin of 12 ng/mL in this patient would increase the posttest probability to 35%, a value that would prompt many physicians to initiate antibiotics.

Overall, there is insufficient evidence to support the use of procalcitonin as a stand-alone test for ruling out bacterial CAP, limiting its use in withholding antibiotics in patients with suspected bacterial CAP.

While initial procalcitonin measurements may not affect the initial antibiotic treatment decision, procalcitonin levels thereafter can guide the duration of therapy. A meta-analysis of procalcitonin-guided treatment in patients with upper or lower respiratory tract infection (LRTI) showed that procalcitonin guidance reduces antibiotic exposure and antibiotic-related adverse effects and improves survival, albeit a small absolute mortality difference of 1.4 percentage points, primarily observed in the intensive care unit setting.⁶ Most patients included in this meta-analysis were diagnosed with LRTI (91%), and CAP was the predominant subtype of LRTI (43%). The main effect of procalcitonin guidance for patients with CAP was earlier discontinuation of antibiotic treatment. Procalcitonin-guided algorithms in these trials discouraged, or strongly discouraged, antibiotics if procalcitonin was <0.25 ng/mL or <0.1 ng/mL, respectively. In addition, serial procalcitonin measurements were used to guide discontinuation of antibiotics if procalcitonin dropped below 0.25 ng/mL, or by 80% to 90% from the peak value. This approach safely shortened the duration of therapy in patients with CAP.

There are several limitations in the interpretation and generalizability of this meta-analysis. There is large heterogeneity across the included clinical trials in design, procalcitonin protocols, clinical setting, and respiratory infection type, including bronchitis, acute exacerbation of chronic obstructive pulmonary disease (AECOPD), and CAP. Results were consistent only in one moderate- to high-quality randomized trial specifically studying CAP in the inpatient setting.⁷ Additionally, most of these trials were conducted in Europe. Antibiotic prescribing practices may be different in the US, and prescribing practices on both continents may have changed over the years with greater awareness and appreciation of antibiotic stewardship.

The ProACT trial, the largest randomized, US multicenter trial to evaluate a procalcitonin-based algorithm to assist with antibiotic decision making, included over 1,600 emergency department patients at 14 academic medical centers.⁸ Procalcitonin guidance in this trial did not reduce antibiotic exposure compared with usual care for patients with suspected LRTI. However, its applicability to the practice of hospitalists and the inpatient setting is limited. First, only 48% of the study participants required hospitalization. Second, this study included all LRTIs, with CAP comprising just 20% of all final diagnoses. Third, the average number of antibiotic days during hospitalization for CAP was short in both groups (3.9 days in the procalcitonin group and 4.1 days in the usual care group). This relatively short antibiotic duration makes it difficult for any intervention to decrease antibiotic days meaningfully.

In a prepost controlled intervention study for inpatients at a single US tertiary care hospital, procalcitonin guidance in hospitalized patients safely reduced antibiotic use in LRTI, specifically for the discontinuation of antibiotics.⁹ The greatest benefit of procalcitonin guidance in antibiotic discontinuation was found in patients with AECOPD and patients with an admitting diagnosis of CAP, but with mild illness and a low procalcitonin. Although this prepost study suggested a safe reduction of antibiotic use due to implementation of procalcitonin guidance, the lack of randomization and the absence of a contemporaneous control group are important limitations. Given the mixed findings on the effectiveness of procalcitonin guidance for hospitalized CAP patients in the US, further investigation will be needed with large clinical trials in the inpatient setting for CAP.

There is insufficient evidence to support the use of serum procalcitonin to withhold initial antibiotics in patients with a clinical syndrome consistent with bacterial CAP. However, the literature supports the use of procalcitonin for the early discontinuation of antibiotics for cases in which the probability of bacterial CAP is low, and procalcitonin remains below 0.1 ng/mL (Figure).

Serial measurements of procalcitonin every one to two days may also be used when clinical uncertainty remains regarding the need for antibiotics. Very low or significantly decreasing procalcitonin levels in patients with CAP and no identified bacterial pathogen likely indicate the infection was not bacterial or was bacterial, but has now been adequately treated with antibiotics. For cases of proven bacterial etiology or high clinical suspicion of bacterial CAP, there is insufficient evidence to recommend the early discontinuation of antibiotics based on procalcitonin levels short of the recommended five-day course according to current guidelines.¹⁰ Future clinical trials are needed to determine if procalcitonin guidance can safely decrease the duration of antibiotic therapy for confirmed bacterial CAP to less than five days.

There are discrepancies between the apparent test characteristics of procalcitonin and the recommended antibiotic decisions in many procalcitonin algorithms. For example, algorithms discourage antibiotics when procalcitonin values are 0.1-0.24 ng/mL, and encourage (or even strongly encourage) antibiotic use for higher procalcitonin values of 0.25-1.0 ng/mL. However, the LRs for these ranges are identical and are approximately 1.0 (Table), suggesting that decision-making should be similar across the entire procalcitonin range of 0.1 to 1.0. Future clinical trials should study revised algorithms with different cut-points, including the thresholds found in our secondary analysis of multilevel LRs. Until then, we believe there is insufficient evidence to deviate from current antibiotic decision recommendations at the traditional cut-points.

While procalcitonin is an imperfect biomarker for discriminating bacterial and nonbacterial etiologies of CAP, it may still provide helpful information for the hospitalist in antibiotic decision-making in the same way we apply other commonly used clinical variables such as fever, white blood cell count, band count, and the pattern of infiltrate in chest imaging.

Procalcitonin should be interpreted cautiously in certain populations in which it has not been extensively studied (eg, immunocompromised) or in noninfectious conditions that may elevate procalcitonin, such as major physiologic stress (eg, surgery, trauma, burns) and end-stage renal disease.^12-14 Further investigation is needed to determine the efficacy and safety of procalcitonin-guided antibiotic therapy in these populations.

• Based on currently available data, a low procalcitonin value should not be used as a stand-alone test to withhold antibiotics in a patient with CAP.
• Serum procalcitonin measurements may help guide the early discontinuation of antibiotics for patients who the treating clinician judges the risks of bacterial etiology and clinical deterioration to be low.
• Interpret procalcitonin cautiously in immunocompromised patients, undergoing severe physiologic stress, or have underlying end-stage renal disease.
• Serum procalcitonin serves as an adjunct to, rather than a substitute for, clinical judgment.

Disclosures

Dr Choi, Dr Evans, and Dr Glesby have nothing to disclose. Dr Self reports receiving prior research funding from BRAHMS/Thermo-Fisher and BioMerieux for studies on procalcitonin. Dr Self reports personal fees from Inflammatix, grants from Axis Shield, Rapid Pathogen Screening, and BioMerieux, all outside the submitted work. Dr McCarthy reports receiving research funding from Allergan outside the submitted work. Dr Simon reports receiving consulting fees from Roche Diagnostics.

Community-acquired pneumonia (CAP) accounts for more than 1.5 million adult hospitalizations and 100,000 deaths each year in the United States.¹ Antibiotic overuse in the hospital setting is an important contributor to the rise of antibiotic resistance, prompting increased efforts to limit inappropriate antibiotic use in hospitals.² Procalcitonin, a precursor of the hormone calcitonin, is upregulated in bacterial infections and downregulated in viral infections. The US Food and Drug Administration has approved it as a serum biomarker to assist clinicians with decisions about using antibiotics.³

There is no consensus on how to best use procalcitonin in the management of CAP. We provide a practical update that includes a review of recent literature, added secondary analysis, and expert opinion surrounding the use of procalcitonin in the diagnosis and management of CAP in hospitalized adults.

Initial procalcitonin levels do not sufficiently exclude bacterial etiologies of CAP to withhold antibiotic prescription safely. The largest diagnostic accuracy study of procalcitonin in the diagnosis of CAP was a subanalysis of the Etiology of Pneumonia in the Community Study.⁴ A total of 1,735 adults hospitalized with CAP received procalcitonin testing along with systematic pathogen testing. The area under the receiver operating characteristic curve for procalcitonin in discriminating bacterial pathogens from viral pathogens was 0.73 (95% CI, 0.69-0.77). A procalcitonin cut-off of 0.1 ng/mL resulted in 80.9% (95% CI, 75.3%-85.7%) sensitivity and 51.6% (95% CI, 46.6%-56.5%) specificity for identification of any bacterial pathogen.

In a secondary analysis of this study, we calculated multilevel likelihood ratios (LRs) for ranges of procalcitonin values to determine the diagnostic accuracy of procalcitonin in distinguishing bacterial from viral etiologies of CAP (Table). Multilevel LRs offer more useful diagnostic information than dichotomizing at specified cut-points.⁵ A procalcitonin result less than 0.1 ng/mL has a negative LR of 0.4 (95% CI, 0.3-0.5), which is not low enough to rule out bacterial CAP effectively when starting with intermediate or high pretest probability. For a low result (<0.1 ng/mL) to be useful in ruling out bacterial CAP, for example having less than a 10% posttest probability of bacterial CAP, the pretest probability would have to be no greater than 22%. Even then, a 10% posttest probability of bacterial CAP may still be too high for clinicians to withhold initial antibiotics. For procalcitonin values between 0.1 ng/mL and 1.0 ng/mL, the probability of bacterial CAP does not change significantly, with an LR of 1.0 (95% CI, 0.8-1.3). Procalcitonin values up to 5 ng/mL reach a modest positive LR of 2.3 (95% CI, 0.8-4.3). Very high values, such as those >10 ng/mL, yield a positive LR of 5.5 (95% CI, 3.2-9.7), are potentially useful in decisions to initiate antibiotics in situations of very low pretest probability of bacterial CAP. For example, a 9% pretest probability of bacterial CAP is likely below many physicians’ threshold for starting antibiotics. A procalcitonin of 12 ng/mL in this patient would increase the posttest probability to 35%, a value that would prompt many physicians to initiate antibiotics.

Overall, there is insufficient evidence to support the use of procalcitonin as a stand-alone test for ruling out bacterial CAP, limiting its use in withholding antibiotics in patients with suspected bacterial CAP.

While initial procalcitonin measurements may not affect the initial antibiotic treatment decision, procalcitonin levels thereafter can guide the duration of therapy. A meta-analysis of procalcitonin-guided treatment in patients with upper or lower respiratory tract infection (LRTI) showed that procalcitonin guidance reduces antibiotic exposure and antibiotic-related adverse effects and improves survival, albeit a small absolute mortality difference of 1.4 percentage points, primarily observed in the intensive care unit setting.⁶ Most patients included in this meta-analysis were diagnosed with LRTI (91%), and CAP was the predominant subtype of LRTI (43%). The main effect of procalcitonin guidance for patients with CAP was earlier discontinuation of antibiotic treatment. Procalcitonin-guided algorithms in these trials discouraged, or strongly discouraged, antibiotics if procalcitonin was <0.25 ng/mL or <0.1 ng/mL, respectively. In addition, serial procalcitonin measurements were used to guide discontinuation of antibiotics if procalcitonin dropped below 0.25 ng/mL, or by 80% to 90% from the peak value. This approach safely shortened the duration of therapy in patients with CAP.

There are several limitations in the interpretation and generalizability of this meta-analysis. There is large heterogeneity across the included clinical trials in design, procalcitonin protocols, clinical setting, and respiratory infection type, including bronchitis, acute exacerbation of chronic obstructive pulmonary disease (AECOPD), and CAP. Results were consistent only in one moderate- to high-quality randomized trial specifically studying CAP in the inpatient setting.⁷ Additionally, most of these trials were conducted in Europe. Antibiotic prescribing practices may be different in the US, and prescribing practices on both continents may have changed over the years with greater awareness and appreciation of antibiotic stewardship.

The ProACT trial, the largest randomized, US multicenter trial to evaluate a procalcitonin-based algorithm to assist with antibiotic decision making, included over 1,600 emergency department patients at 14 academic medical centers.⁸ Procalcitonin guidance in this trial did not reduce antibiotic exposure compared with usual care for patients with suspected LRTI. However, its applicability to the practice of hospitalists and the inpatient setting is limited. First, only 48% of the study participants required hospitalization. Second, this study included all LRTIs, with CAP comprising just 20% of all final diagnoses. Third, the average number of antibiotic days during hospitalization for CAP was short in both groups (3.9 days in the procalcitonin group and 4.1 days in the usual care group). This relatively short antibiotic duration makes it difficult for any intervention to decrease antibiotic days meaningfully.

In a prepost controlled intervention study for inpatients at a single US tertiary care hospital, procalcitonin guidance in hospitalized patients safely reduced antibiotic use in LRTI, specifically for the discontinuation of antibiotics.⁹ The greatest benefit of procalcitonin guidance in antibiotic discontinuation was found in patients with AECOPD and patients with an admitting diagnosis of CAP, but with mild illness and a low procalcitonin. Although this prepost study suggested a safe reduction of antibiotic use due to implementation of procalcitonin guidance, the lack of randomization and the absence of a contemporaneous control group are important limitations. Given the mixed findings on the effectiveness of procalcitonin guidance for hospitalized CAP patients in the US, further investigation will be needed with large clinical trials in the inpatient setting for CAP.

There is insufficient evidence to support the use of serum procalcitonin to withhold initial antibiotics in patients with a clinical syndrome consistent with bacterial CAP. However, the literature supports the use of procalcitonin for the early discontinuation of antibiotics for cases in which the probability of bacterial CAP is low, and procalcitonin remains below 0.1 ng/mL (Figure).

Serial measurements of procalcitonin every one to two days may also be used when clinical uncertainty remains regarding the need for antibiotics. Very low or significantly decreasing procalcitonin levels in patients with CAP and no identified bacterial pathogen likely indicate the infection was not bacterial or was bacterial, but has now been adequately treated with antibiotics. For cases of proven bacterial etiology or high clinical suspicion of bacterial CAP, there is insufficient evidence to recommend the early discontinuation of antibiotics based on procalcitonin levels short of the recommended five-day course according to current guidelines.¹⁰ Future clinical trials are needed to determine if procalcitonin guidance can safely decrease the duration of antibiotic therapy for confirmed bacterial CAP to less than five days.

There are discrepancies between the apparent test characteristics of procalcitonin and the recommended antibiotic decisions in many procalcitonin algorithms. For example, algorithms discourage antibiotics when procalcitonin values are 0.1-0.24 ng/mL, and encourage (or even strongly encourage) antibiotic use for higher procalcitonin values of 0.25-1.0 ng/mL. However, the LRs for these ranges are identical and are approximately 1.0 (Table), suggesting that decision-making should be similar across the entire procalcitonin range of 0.1 to 1.0. Future clinical trials should study revised algorithms with different cut-points, including the thresholds found in our secondary analysis of multilevel LRs. Until then, we believe there is insufficient evidence to deviate from current antibiotic decision recommendations at the traditional cut-points.

While procalcitonin is an imperfect biomarker for discriminating bacterial and nonbacterial etiologies of CAP, it may still provide helpful information for the hospitalist in antibiotic decision-making in the same way we apply other commonly used clinical variables such as fever, white blood cell count, band count, and the pattern of infiltrate in chest imaging.

Procalcitonin should be interpreted cautiously in certain populations in which it has not been extensively studied (eg, immunocompromised) or in noninfectious conditions that may elevate procalcitonin, such as major physiologic stress (eg, surgery, trauma, burns) and end-stage renal disease.^12-14 Further investigation is needed to determine the efficacy and safety of procalcitonin-guided antibiotic therapy in these populations.

• Based on currently available data, a low procalcitonin value should not be used as a stand-alone test to withhold antibiotics in a patient with CAP.
• Serum procalcitonin measurements may help guide the early discontinuation of antibiotics for patients who the treating clinician judges the risks of bacterial etiology and clinical deterioration to be low.
• Interpret procalcitonin cautiously in immunocompromised patients, undergoing severe physiologic stress, or have underlying end-stage renal disease.
• Serum procalcitonin serves as an adjunct to, rather than a substitute for, clinical judgment.

Disclosures

Dr Choi, Dr Evans, and Dr Glesby have nothing to disclose. Dr Self reports receiving prior research funding from BRAHMS/Thermo-Fisher and BioMerieux for studies on procalcitonin. Dr Self reports personal fees from Inflammatix, grants from Axis Shield, Rapid Pathogen Screening, and BioMerieux, all outside the submitted work. Dr McCarthy reports receiving research funding from Allergan outside the submitted work. Dr Simon reports receiving consulting fees from Roche Diagnostics.

Community-acquired pneumonia (CAP) accounts for more than 1.5 million adult hospitalizations and 100,000 deaths each year in the United States.¹ Antibiotic overuse in the hospital setting is an important contributor to the rise of antibiotic resistance, prompting increased efforts to limit inappropriate antibiotic use in hospitals.² Procalcitonin, a precursor of the hormone calcitonin, is upregulated in bacterial infections and downregulated in viral infections. The US Food and Drug Administration has approved it as a serum biomarker to assist clinicians with decisions about using antibiotics.³

There is no consensus on how to best use procalcitonin in the management of CAP. We provide a practical update that includes a review of recent literature, added secondary analysis, and expert opinion surrounding the use of procalcitonin in the diagnosis and management of CAP in hospitalized adults.

Initial procalcitonin levels do not sufficiently exclude bacterial etiologies of CAP to withhold antibiotic prescription safely. The largest diagnostic accuracy study of procalcitonin in the diagnosis of CAP was a subanalysis of the Etiology of Pneumonia in the Community Study.⁴ A total of 1,735 adults hospitalized with CAP received procalcitonin testing along with systematic pathogen testing. The area under the receiver operating characteristic curve for procalcitonin in discriminating bacterial pathogens from viral pathogens was 0.73 (95% CI, 0.69-0.77). A procalcitonin cut-off of 0.1 ng/mL resulted in 80.9% (95% CI, 75.3%-85.7%) sensitivity and 51.6% (95% CI, 46.6%-56.5%) specificity for identification of any bacterial pathogen.

In a secondary analysis of this study, we calculated multilevel likelihood ratios (LRs) for ranges of procalcitonin values to determine the diagnostic accuracy of procalcitonin in distinguishing bacterial from viral etiologies of CAP (Table). Multilevel LRs offer more useful diagnostic information than dichotomizing at specified cut-points.⁵ A procalcitonin result less than 0.1 ng/mL has a negative LR of 0.4 (95% CI, 0.3-0.5), which is not low enough to rule out bacterial CAP effectively when starting with intermediate or high pretest probability. For a low result (<0.1 ng/mL) to be useful in ruling out bacterial CAP, for example having less than a 10% posttest probability of bacterial CAP, the pretest probability would have to be no greater than 22%. Even then, a 10% posttest probability of bacterial CAP may still be too high for clinicians to withhold initial antibiotics. For procalcitonin values between 0.1 ng/mL and 1.0 ng/mL, the probability of bacterial CAP does not change significantly, with an LR of 1.0 (95% CI, 0.8-1.3). Procalcitonin values up to 5 ng/mL reach a modest positive LR of 2.3 (95% CI, 0.8-4.3). Very high values, such as those >10 ng/mL, yield a positive LR of 5.5 (95% CI, 3.2-9.7), are potentially useful in decisions to initiate antibiotics in situations of very low pretest probability of bacterial CAP. For example, a 9% pretest probability of bacterial CAP is likely below many physicians’ threshold for starting antibiotics. A procalcitonin of 12 ng/mL in this patient would increase the posttest probability to 35%, a value that would prompt many physicians to initiate antibiotics.

Overall, there is insufficient evidence to support the use of procalcitonin as a stand-alone test for ruling out bacterial CAP, limiting its use in withholding antibiotics in patients with suspected bacterial CAP.

While initial procalcitonin measurements may not affect the initial antibiotic treatment decision, procalcitonin levels thereafter can guide the duration of therapy. A meta-analysis of procalcitonin-guided treatment in patients with upper or lower respiratory tract infection (LRTI) showed that procalcitonin guidance reduces antibiotic exposure and antibiotic-related adverse effects and improves survival, albeit a small absolute mortality difference of 1.4 percentage points, primarily observed in the intensive care unit setting.⁶ Most patients included in this meta-analysis were diagnosed with LRTI (91%), and CAP was the predominant subtype of LRTI (43%). The main effect of procalcitonin guidance for patients with CAP was earlier discontinuation of antibiotic treatment. Procalcitonin-guided algorithms in these trials discouraged, or strongly discouraged, antibiotics if procalcitonin was <0.25 ng/mL or <0.1 ng/mL, respectively. In addition, serial procalcitonin measurements were used to guide discontinuation of antibiotics if procalcitonin dropped below 0.25 ng/mL, or by 80% to 90% from the peak value. This approach safely shortened the duration of therapy in patients with CAP.

There are several limitations in the interpretation and generalizability of this meta-analysis. There is large heterogeneity across the included clinical trials in design, procalcitonin protocols, clinical setting, and respiratory infection type, including bronchitis, acute exacerbation of chronic obstructive pulmonary disease (AECOPD), and CAP. Results were consistent only in one moderate- to high-quality randomized trial specifically studying CAP in the inpatient setting.⁷ Additionally, most of these trials were conducted in Europe. Antibiotic prescribing practices may be different in the US, and prescribing practices on both continents may have changed over the years with greater awareness and appreciation of antibiotic stewardship.

The ProACT trial, the largest randomized, US multicenter trial to evaluate a procalcitonin-based algorithm to assist with antibiotic decision making, included over 1,600 emergency department patients at 14 academic medical centers.⁸ Procalcitonin guidance in this trial did not reduce antibiotic exposure compared with usual care for patients with suspected LRTI. However, its applicability to the practice of hospitalists and the inpatient setting is limited. First, only 48% of the study participants required hospitalization. Second, this study included all LRTIs, with CAP comprising just 20% of all final diagnoses. Third, the average number of antibiotic days during hospitalization for CAP was short in both groups (3.9 days in the procalcitonin group and 4.1 days in the usual care group). This relatively short antibiotic duration makes it difficult for any intervention to decrease antibiotic days meaningfully.

In a prepost controlled intervention study for inpatients at a single US tertiary care hospital, procalcitonin guidance in hospitalized patients safely reduced antibiotic use in LRTI, specifically for the discontinuation of antibiotics.⁹ The greatest benefit of procalcitonin guidance in antibiotic discontinuation was found in patients with AECOPD and patients with an admitting diagnosis of CAP, but with mild illness and a low procalcitonin. Although this prepost study suggested a safe reduction of antibiotic use due to implementation of procalcitonin guidance, the lack of randomization and the absence of a contemporaneous control group are important limitations. Given the mixed findings on the effectiveness of procalcitonin guidance for hospitalized CAP patients in the US, further investigation will be needed with large clinical trials in the inpatient setting for CAP.

There is insufficient evidence to support the use of serum procalcitonin to withhold initial antibiotics in patients with a clinical syndrome consistent with bacterial CAP. However, the literature supports the use of procalcitonin for the early discontinuation of antibiotics for cases in which the probability of bacterial CAP is low, and procalcitonin remains below 0.1 ng/mL (Figure).

Serial measurements of procalcitonin every one to two days may also be used when clinical uncertainty remains regarding the need for antibiotics. Very low or significantly decreasing procalcitonin levels in patients with CAP and no identified bacterial pathogen likely indicate the infection was not bacterial or was bacterial, but has now been adequately treated with antibiotics. For cases of proven bacterial etiology or high clinical suspicion of bacterial CAP, there is insufficient evidence to recommend the early discontinuation of antibiotics based on procalcitonin levels short of the recommended five-day course according to current guidelines.¹⁰ Future clinical trials are needed to determine if procalcitonin guidance can safely decrease the duration of antibiotic therapy for confirmed bacterial CAP to less than five days.

There are discrepancies between the apparent test characteristics of procalcitonin and the recommended antibiotic decisions in many procalcitonin algorithms. For example, algorithms discourage antibiotics when procalcitonin values are 0.1-0.24 ng/mL, and encourage (or even strongly encourage) antibiotic use for higher procalcitonin values of 0.25-1.0 ng/mL. However, the LRs for these ranges are identical and are approximately 1.0 (Table), suggesting that decision-making should be similar across the entire procalcitonin range of 0.1 to 1.0. Future clinical trials should study revised algorithms with different cut-points, including the thresholds found in our secondary analysis of multilevel LRs. Until then, we believe there is insufficient evidence to deviate from current antibiotic decision recommendations at the traditional cut-points.

While procalcitonin is an imperfect biomarker for discriminating bacterial and nonbacterial etiologies of CAP, it may still provide helpful information for the hospitalist in antibiotic decision-making in the same way we apply other commonly used clinical variables such as fever, white blood cell count, band count, and the pattern of infiltrate in chest imaging.

Procalcitonin should be interpreted cautiously in certain populations in which it has not been extensively studied (eg, immunocompromised) or in noninfectious conditions that may elevate procalcitonin, such as major physiologic stress (eg, surgery, trauma, burns) and end-stage renal disease.^12-14 Further investigation is needed to determine the efficacy and safety of procalcitonin-guided antibiotic therapy in these populations.

• Based on currently available data, a low procalcitonin value should not be used as a stand-alone test to withhold antibiotics in a patient with CAP.
• Serum procalcitonin measurements may help guide the early discontinuation of antibiotics for patients who the treating clinician judges the risks of bacterial etiology and clinical deterioration to be low.
• Interpret procalcitonin cautiously in immunocompromised patients, undergoing severe physiologic stress, or have underlying end-stage renal disease.
• Serum procalcitonin serves as an adjunct to, rather than a substitute for, clinical judgment.

Disclosures

Dr Choi, Dr Evans, and Dr Glesby have nothing to disclose. Dr Self reports receiving prior research funding from BRAHMS/Thermo-Fisher and BioMerieux for studies on procalcitonin. Dr Self reports personal fees from Inflammatix, grants from Axis Shield, Rapid Pathogen Screening, and BioMerieux, all outside the submitted work. Dr McCarthy reports receiving research funding from Allergan outside the submitted work. Dr Simon reports receiving consulting fees from Roche Diagnostics.

Much has been written about the sources of the hidden curriculum in clerkships and postgraduate medical education.^1-3 However, these descriptions do not adequately account for the critical role that hospitalists play in the development of trainees when they encounter ethical challenges on teaching services.⁴ As a role model, teacher, and the attending of record, a hospitalist’s response to ethical issues in practice can have a pivotal influence on the life and work of trainees, either instilling positive virtues or perpetuating the negative impact of the hidden curriculum.^5-8 Understanding the epidemiology of ethical issues arising on academic hospitalist services has important implications for medical education, clinical ethics, and professionalism, as well as for patient care.

Study Setting and Design

We conducted a mixed-method observational study at NewYork–Presbyterian–Weill Cornell Medical Center, an 862-bed, tertiary-care, academic institution located in New York, New York. We performed a prospective description of the frequency of all consecutively identified ethical and contextual issues pertinent to clinical decision-making by observing morning rounds with housestaff hospitalist services. Ethical issues were categorized using a comprehensive standardized instrument previously developed and published by the Division of Medical Ethics.⁹

The Division of Hospital Medicine employs 79 physicians, 30 of whom are dedicated full-time to daytime care on house-staff (or teaching) or physician assistant services. Of these 30 physicians, two (7%) were coinvestigators in this project and were excluded from participation to avoid bias. Between September 2017 and May 2018, the attending physicians of record of all available housestaff services were invited to participate with their teams in our research study on a weekly basis. We observed 10 different Hospital Medicine attending physicians (10/28, 36% of the available physician sample) over 19 sessions. Before rounds, a brief introduction to the nature of the study was provided to each team. It was explicitly stated that the observers were present to identify and document possible ethical issues that may arise while discussing the patients on rounds, and that the purpose of the study was neither an evaluation of the team members or their decisions nor a critique or quality improvement exercise. Observing researchers were not allowed to participate in the discussion of any case.

To avoid potential case duplication, we allowed for a minimum two-week interval before rounding twice on any particular team. To control for interobserver variability, we observed in pairs during these sessions. Discrepancies between observers were resolved by post hoc discussion and application of the definitions of the standardized instrument used to identify and catalog ethical and contextual issues.

Study Variables and Definitions

The following variables were collected in all cases: observation date, name of reviewers, demographic characteristics of the patient (age, gender, race, ethnicity, marital status, religion, preferred language, insurance type, and living situation before the admission), patient’s location during the admission (emergency room, regular nursing floor, step-down unit, or other), and ethical and contextual issues. “Ethical issues” were defined as those situations involving a conflict of values or preferences among different stakeholders, including, but not limited to, providers, patients, and/or families. Explicit definitions of each issue were generated, and additional standard rules for completion were provided.

Statistical Analysis

Results are presented as n (%) or mean ± standard deviation. Percentages were rounded to the closest integer. Interobserver variability between the observers in relation to evaluating the presence or absence of ethical or contextual issues was assessed by the kappa statistic. All P values are two-sided, with statistical significance evaluated at the 0.05 alpha level. A 95% confidence interval (95% CI) for the kappa statistic (ie, for assessing interobserver variability) was calculated to assess the precision of the obtained kappa estimate. All analyses were performed in SAS Version 9.4 (SAS Institute, Inc., Cary, NC) and Stata Version 14.0 (StataCorp, College Station, TX).

General Characteristics of the Study Sample

In total, 270 patients were evaluated from the teaching hospitalist services during the observation period. Ethical issues were identified in 86 of these patients (31.8%). Observer ethicists disagreed in their initial evaluation of 17 cases (6.3%). After review of and adjudication, both observers agreed that nine of these 17 cases (3.3%) should be excluded from the final analysis, as none reached the necessary threshold to be considered as a true ethical issue. Hence, we report the results of 77 patients (28.5%). These cases comprised the Hospitalist group and involved 113 ethical issues (1.48 ± 0.5 ethical issues/case). Only five patients in the Hospitalist group had a formal clinical ethics consult before our observation (5/270 patients [1.9%] vs 77/270 patients [28.5%] with an ethical issue, respectively, P < .001). Although the majority of ethical issues were noted by members of the primary team (84%), 12 of the 77 cases in the Hospitalist group (16%) were identified only by the observing ethicists. The kappa statistic for interobserver variability between the observing ethicists was 0.85 (95% CI = 0.76-0.92). The major demographic characteristics are summarized in Table 1.

Ethical Challenges

The most common ethical issues hospitalists encountered involved discussions about goals of care (including decisions to pursue aggressive treatment versus hospice care, or debates about the team’s ambivalence about the benefits and risks of pursuing investigational chemotherapy), treatment refusals (including the decision to forgo biopsy of a suspected malignancy), or decision-making capacity (Table 2). Less common were issues pertaining to resource allocation (specially related to pressures to discharge patients), pain management (some patients were suspected of drug-seeking behavior), or surrogate decision-making (when alternative decision-makers were suspected to lack decision-making capacity). Discussions about forgoing life-sustaining treatments occurred only in four cases (5%). These involved considerations of withdrawing Bilevel Positive Airway Pressure (BiPAP), artificial nutrition and hydration, and/or stopping antibiotic treatment.

DISCUSSION

Our data are the first prospective description of ethical issues arising on an academic hospitalist teaching service. These results indicate that there is an ethics epidemiology in the routine practice of Hospital Medicine that has heretofore not been characterized. By this, we mean a discreet incidence and prevalence of ethical challenges in Hospital Medicine that is distinct from that which is encountered by clinical ethics consultation (CEC) services. Although most practitioners recognize the utility of a traditional ethics consultation, there is a surprising paucity of data about the sources of ethical conflict encountered by academic hospitalists at the bedside, particularly those addressed without CEC. This suggests that the criteria for requesting a formal ethics consult could be limited and restrictive, which is both undersensitive and overspecific.¹⁰ Because of these limitations, viewing traditional ethics consultation as a proxy for ethical issues arising in daily hospitalist practice would lead to an underestimation of the true prevalence, as our data indicate.

More than one-fourth of the patients admitted to hospitalist teaching services pose ethical conflicts. Some of these are addressed on rounds, some are not, and only a handful of these cases will ever be referred to an ethicist. CEC services are made aware of the “tip of the iceberg,” which accounts for a vanishingly small percentage of ethical issues that arise on daily rounds. Some hospitalists may not involve CEC simply because they believe that the services are not helpful. However, the failure to obtain consultation may also reflect an inability to recognize a “problematic situation” and formulate a referral that might benefit from the assistance of an ethics consultation.¹¹

Our study faces several potential limitations. We are presenting a single-center experience that focuses on the perspective of physicians and trainees. Some ethical issues might have been underestimated because the perspectives of patients, families, nurses, social workers, or other ancillary staff were not directly included. Furthermore, since any ethical challenge could have been discussed on any moment other than on morning rounds, our results may underestimate the prevalence of ethical issues arising from the hospital floors. Moreover, medical teams participating in the study could have been subject to the Hawthorne effect and could have tried to identify a greater number of ethical issues on rounds, which would not reflect actual practice.

Almost two decades ago, Coulehan and Williams wrote about the positive impact that ethics and humanities could have if these disciplines could be embedded in the daily practice of medicine, which is as follows:

…ethics and humanities curricula are irrelevant unless they can produce a substantive and continuing impact on hospital culture (…) The idea, of course, is to infiltrate the culture by coopting residents and attending physicians(…) If an ethics program can somehow achieve a critical mass of ‘‘value-sensitive’’ clinical faculty, it may begin to influence the institution’s ethos.¹²

Coulehan and Williams wrote of a need to bring ethics to the bedside. Our data suggest that an ethics epidemiology is deeply embedded in hospitalist services and is waiting to be fully characterized to better inform the care of patients and guide the professional formation and education of students and trainees. Hospitalists frequently confront ethical problems in daily practice that do not come to the attention of the CEC services or the institutional ethics committee. Understanding this emerging epidemiology presents an unrealized opportunity to improve bedside teaching, reinforce normative reasoning, and enhance patient care.

Acknowledgments

The authors want to acknowledge Drs. Augustine I. Choi, Michael G. Stewart, Laura L. Forese, and Anthony Hollenberg for their support of the fellowship in medical ethics and thank Drs. Arthur T. Evans and Monika M. Safford for their guidance.

Disclosures

The authors report no conflicts of interest.

Funding

This work was supported by a Weill Cornell General Internal Medicine Primary Care Innovations Initiative seed grant. Dr. Paul Christos was partially supported by the following grant: Clinical and Translational Science Center at Weill Cornell Medical College (1-UL1-TR002384-01).

Much has been written about the sources of the hidden curriculum in clerkships and postgraduate medical education.^1-3 However, these descriptions do not adequately account for the critical role that hospitalists play in the development of trainees when they encounter ethical challenges on teaching services.⁴ As a role model, teacher, and the attending of record, a hospitalist’s response to ethical issues in practice can have a pivotal influence on the life and work of trainees, either instilling positive virtues or perpetuating the negative impact of the hidden curriculum.^5-8 Understanding the epidemiology of ethical issues arising on academic hospitalist services has important implications for medical education, clinical ethics, and professionalism, as well as for patient care.

Study Setting and Design

We conducted a mixed-method observational study at NewYork–Presbyterian–Weill Cornell Medical Center, an 862-bed, tertiary-care, academic institution located in New York, New York. We performed a prospective description of the frequency of all consecutively identified ethical and contextual issues pertinent to clinical decision-making by observing morning rounds with housestaff hospitalist services. Ethical issues were categorized using a comprehensive standardized instrument previously developed and published by the Division of Medical Ethics.⁹

The Division of Hospital Medicine employs 79 physicians, 30 of whom are dedicated full-time to daytime care on house-staff (or teaching) or physician assistant services. Of these 30 physicians, two (7%) were coinvestigators in this project and were excluded from participation to avoid bias. Between September 2017 and May 2018, the attending physicians of record of all available housestaff services were invited to participate with their teams in our research study on a weekly basis. We observed 10 different Hospital Medicine attending physicians (10/28, 36% of the available physician sample) over 19 sessions. Before rounds, a brief introduction to the nature of the study was provided to each team. It was explicitly stated that the observers were present to identify and document possible ethical issues that may arise while discussing the patients on rounds, and that the purpose of the study was neither an evaluation of the team members or their decisions nor a critique or quality improvement exercise. Observing researchers were not allowed to participate in the discussion of any case.

To avoid potential case duplication, we allowed for a minimum two-week interval before rounding twice on any particular team. To control for interobserver variability, we observed in pairs during these sessions. Discrepancies between observers were resolved by post hoc discussion and application of the definitions of the standardized instrument used to identify and catalog ethical and contextual issues.

Study Variables and Definitions

The following variables were collected in all cases: observation date, name of reviewers, demographic characteristics of the patient (age, gender, race, ethnicity, marital status, religion, preferred language, insurance type, and living situation before the admission), patient’s location during the admission (emergency room, regular nursing floor, step-down unit, or other), and ethical and contextual issues. “Ethical issues” were defined as those situations involving a conflict of values or preferences among different stakeholders, including, but not limited to, providers, patients, and/or families. Explicit definitions of each issue were generated, and additional standard rules for completion were provided.

Statistical Analysis

Results are presented as n (%) or mean ± standard deviation. Percentages were rounded to the closest integer. Interobserver variability between the observers in relation to evaluating the presence or absence of ethical or contextual issues was assessed by the kappa statistic. All P values are two-sided, with statistical significance evaluated at the 0.05 alpha level. A 95% confidence interval (95% CI) for the kappa statistic (ie, for assessing interobserver variability) was calculated to assess the precision of the obtained kappa estimate. All analyses were performed in SAS Version 9.4 (SAS Institute, Inc., Cary, NC) and Stata Version 14.0 (StataCorp, College Station, TX).

General Characteristics of the Study Sample

In total, 270 patients were evaluated from the teaching hospitalist services during the observation period. Ethical issues were identified in 86 of these patients (31.8%). Observer ethicists disagreed in their initial evaluation of 17 cases (6.3%). After review of and adjudication, both observers agreed that nine of these 17 cases (3.3%) should be excluded from the final analysis, as none reached the necessary threshold to be considered as a true ethical issue. Hence, we report the results of 77 patients (28.5%). These cases comprised the Hospitalist group and involved 113 ethical issues (1.48 ± 0.5 ethical issues/case). Only five patients in the Hospitalist group had a formal clinical ethics consult before our observation (5/270 patients [1.9%] vs 77/270 patients [28.5%] with an ethical issue, respectively, P < .001). Although the majority of ethical issues were noted by members of the primary team (84%), 12 of the 77 cases in the Hospitalist group (16%) were identified only by the observing ethicists. The kappa statistic for interobserver variability between the observing ethicists was 0.85 (95% CI = 0.76-0.92). The major demographic characteristics are summarized in Table 1.

Ethical Challenges

The most common ethical issues hospitalists encountered involved discussions about goals of care (including decisions to pursue aggressive treatment versus hospice care, or debates about the team’s ambivalence about the benefits and risks of pursuing investigational chemotherapy), treatment refusals (including the decision to forgo biopsy of a suspected malignancy), or decision-making capacity (Table 2). Less common were issues pertaining to resource allocation (specially related to pressures to discharge patients), pain management (some patients were suspected of drug-seeking behavior), or surrogate decision-making (when alternative decision-makers were suspected to lack decision-making capacity). Discussions about forgoing life-sustaining treatments occurred only in four cases (5%). These involved considerations of withdrawing Bilevel Positive Airway Pressure (BiPAP), artificial nutrition and hydration, and/or stopping antibiotic treatment.

DISCUSSION

Our data are the first prospective description of ethical issues arising on an academic hospitalist teaching service. These results indicate that there is an ethics epidemiology in the routine practice of Hospital Medicine that has heretofore not been characterized. By this, we mean a discreet incidence and prevalence of ethical challenges in Hospital Medicine that is distinct from that which is encountered by clinical ethics consultation (CEC) services. Although most practitioners recognize the utility of a traditional ethics consultation, there is a surprising paucity of data about the sources of ethical conflict encountered by academic hospitalists at the bedside, particularly those addressed without CEC. This suggests that the criteria for requesting a formal ethics consult could be limited and restrictive, which is both undersensitive and overspecific.¹⁰ Because of these limitations, viewing traditional ethics consultation as a proxy for ethical issues arising in daily hospitalist practice would lead to an underestimation of the true prevalence, as our data indicate.

More than one-fourth of the patients admitted to hospitalist teaching services pose ethical conflicts. Some of these are addressed on rounds, some are not, and only a handful of these cases will ever be referred to an ethicist. CEC services are made aware of the “tip of the iceberg,” which accounts for a vanishingly small percentage of ethical issues that arise on daily rounds. Some hospitalists may not involve CEC simply because they believe that the services are not helpful. However, the failure to obtain consultation may also reflect an inability to recognize a “problematic situation” and formulate a referral that might benefit from the assistance of an ethics consultation.¹¹

Our study faces several potential limitations. We are presenting a single-center experience that focuses on the perspective of physicians and trainees. Some ethical issues might have been underestimated because the perspectives of patients, families, nurses, social workers, or other ancillary staff were not directly included. Furthermore, since any ethical challenge could have been discussed on any moment other than on morning rounds, our results may underestimate the prevalence of ethical issues arising from the hospital floors. Moreover, medical teams participating in the study could have been subject to the Hawthorne effect and could have tried to identify a greater number of ethical issues on rounds, which would not reflect actual practice.

Almost two decades ago, Coulehan and Williams wrote about the positive impact that ethics and humanities could have if these disciplines could be embedded in the daily practice of medicine, which is as follows:

…ethics and humanities curricula are irrelevant unless they can produce a substantive and continuing impact on hospital culture (…) The idea, of course, is to infiltrate the culture by coopting residents and attending physicians(…) If an ethics program can somehow achieve a critical mass of ‘‘value-sensitive’’ clinical faculty, it may begin to influence the institution’s ethos.¹²

Coulehan and Williams wrote of a need to bring ethics to the bedside. Our data suggest that an ethics epidemiology is deeply embedded in hospitalist services and is waiting to be fully characterized to better inform the care of patients and guide the professional formation and education of students and trainees. Hospitalists frequently confront ethical problems in daily practice that do not come to the attention of the CEC services or the institutional ethics committee. Understanding this emerging epidemiology presents an unrealized opportunity to improve bedside teaching, reinforce normative reasoning, and enhance patient care.

Acknowledgments

The authors want to acknowledge Drs. Augustine I. Choi, Michael G. Stewart, Laura L. Forese, and Anthony Hollenberg for their support of the fellowship in medical ethics and thank Drs. Arthur T. Evans and Monika M. Safford for their guidance.

Disclosures

The authors report no conflicts of interest.

Funding

This work was supported by a Weill Cornell General Internal Medicine Primary Care Innovations Initiative seed grant. Dr. Paul Christos was partially supported by the following grant: Clinical and Translational Science Center at Weill Cornell Medical College (1-UL1-TR002384-01).

Much has been written about the sources of the hidden curriculum in clerkships and postgraduate medical education.^1-3 However, these descriptions do not adequately account for the critical role that hospitalists play in the development of trainees when they encounter ethical challenges on teaching services.⁴ As a role model, teacher, and the attending of record, a hospitalist’s response to ethical issues in practice can have a pivotal influence on the life and work of trainees, either instilling positive virtues or perpetuating the negative impact of the hidden curriculum.^5-8 Understanding the epidemiology of ethical issues arising on academic hospitalist services has important implications for medical education, clinical ethics, and professionalism, as well as for patient care.

Study Setting and Design

We conducted a mixed-method observational study at NewYork–Presbyterian–Weill Cornell Medical Center, an 862-bed, tertiary-care, academic institution located in New York, New York. We performed a prospective description of the frequency of all consecutively identified ethical and contextual issues pertinent to clinical decision-making by observing morning rounds with housestaff hospitalist services. Ethical issues were categorized using a comprehensive standardized instrument previously developed and published by the Division of Medical Ethics.⁹

The Division of Hospital Medicine employs 79 physicians, 30 of whom are dedicated full-time to daytime care on house-staff (or teaching) or physician assistant services. Of these 30 physicians, two (7%) were coinvestigators in this project and were excluded from participation to avoid bias. Between September 2017 and May 2018, the attending physicians of record of all available housestaff services were invited to participate with their teams in our research study on a weekly basis. We observed 10 different Hospital Medicine attending physicians (10/28, 36% of the available physician sample) over 19 sessions. Before rounds, a brief introduction to the nature of the study was provided to each team. It was explicitly stated that the observers were present to identify and document possible ethical issues that may arise while discussing the patients on rounds, and that the purpose of the study was neither an evaluation of the team members or their decisions nor a critique or quality improvement exercise. Observing researchers were not allowed to participate in the discussion of any case.

To avoid potential case duplication, we allowed for a minimum two-week interval before rounding twice on any particular team. To control for interobserver variability, we observed in pairs during these sessions. Discrepancies between observers were resolved by post hoc discussion and application of the definitions of the standardized instrument used to identify and catalog ethical and contextual issues.

Study Variables and Definitions

The following variables were collected in all cases: observation date, name of reviewers, demographic characteristics of the patient (age, gender, race, ethnicity, marital status, religion, preferred language, insurance type, and living situation before the admission), patient’s location during the admission (emergency room, regular nursing floor, step-down unit, or other), and ethical and contextual issues. “Ethical issues” were defined as those situations involving a conflict of values or preferences among different stakeholders, including, but not limited to, providers, patients, and/or families. Explicit definitions of each issue were generated, and additional standard rules for completion were provided.

Statistical Analysis

Results are presented as n (%) or mean ± standard deviation. Percentages were rounded to the closest integer. Interobserver variability between the observers in relation to evaluating the presence or absence of ethical or contextual issues was assessed by the kappa statistic. All P values are two-sided, with statistical significance evaluated at the 0.05 alpha level. A 95% confidence interval (95% CI) for the kappa statistic (ie, for assessing interobserver variability) was calculated to assess the precision of the obtained kappa estimate. All analyses were performed in SAS Version 9.4 (SAS Institute, Inc., Cary, NC) and Stata Version 14.0 (StataCorp, College Station, TX).

General Characteristics of the Study Sample

In total, 270 patients were evaluated from the teaching hospitalist services during the observation period. Ethical issues were identified in 86 of these patients (31.8%). Observer ethicists disagreed in their initial evaluation of 17 cases (6.3%). After review of and adjudication, both observers agreed that nine of these 17 cases (3.3%) should be excluded from the final analysis, as none reached the necessary threshold to be considered as a true ethical issue. Hence, we report the results of 77 patients (28.5%). These cases comprised the Hospitalist group and involved 113 ethical issues (1.48 ± 0.5 ethical issues/case). Only five patients in the Hospitalist group had a formal clinical ethics consult before our observation (5/270 patients [1.9%] vs 77/270 patients [28.5%] with an ethical issue, respectively, P < .001). Although the majority of ethical issues were noted by members of the primary team (84%), 12 of the 77 cases in the Hospitalist group (16%) were identified only by the observing ethicists. The kappa statistic for interobserver variability between the observing ethicists was 0.85 (95% CI = 0.76-0.92). The major demographic characteristics are summarized in Table 1.

Ethical Challenges

The most common ethical issues hospitalists encountered involved discussions about goals of care (including decisions to pursue aggressive treatment versus hospice care, or debates about the team’s ambivalence about the benefits and risks of pursuing investigational chemotherapy), treatment refusals (including the decision to forgo biopsy of a suspected malignancy), or decision-making capacity (Table 2). Less common were issues pertaining to resource allocation (specially related to pressures to discharge patients), pain management (some patients were suspected of drug-seeking behavior), or surrogate decision-making (when alternative decision-makers were suspected to lack decision-making capacity). Discussions about forgoing life-sustaining treatments occurred only in four cases (5%). These involved considerations of withdrawing Bilevel Positive Airway Pressure (BiPAP), artificial nutrition and hydration, and/or stopping antibiotic treatment.

DISCUSSION

Our data are the first prospective description of ethical issues arising on an academic hospitalist teaching service. These results indicate that there is an ethics epidemiology in the routine practice of Hospital Medicine that has heretofore not been characterized. By this, we mean a discreet incidence and prevalence of ethical challenges in Hospital Medicine that is distinct from that which is encountered by clinical ethics consultation (CEC) services. Although most practitioners recognize the utility of a traditional ethics consultation, there is a surprising paucity of data about the sources of ethical conflict encountered by academic hospitalists at the bedside, particularly those addressed without CEC. This suggests that the criteria for requesting a formal ethics consult could be limited and restrictive, which is both undersensitive and overspecific.¹⁰ Because of these limitations, viewing traditional ethics consultation as a proxy for ethical issues arising in daily hospitalist practice would lead to an underestimation of the true prevalence, as our data indicate.

More than one-fourth of the patients admitted to hospitalist teaching services pose ethical conflicts. Some of these are addressed on rounds, some are not, and only a handful of these cases will ever be referred to an ethicist. CEC services are made aware of the “tip of the iceberg,” which accounts for a vanishingly small percentage of ethical issues that arise on daily rounds. Some hospitalists may not involve CEC simply because they believe that the services are not helpful. However, the failure to obtain consultation may also reflect an inability to recognize a “problematic situation” and formulate a referral that might benefit from the assistance of an ethics consultation.¹¹

Our study faces several potential limitations. We are presenting a single-center experience that focuses on the perspective of physicians and trainees. Some ethical issues might have been underestimated because the perspectives of patients, families, nurses, social workers, or other ancillary staff were not directly included. Furthermore, since any ethical challenge could have been discussed on any moment other than on morning rounds, our results may underestimate the prevalence of ethical issues arising from the hospital floors. Moreover, medical teams participating in the study could have been subject to the Hawthorne effect and could have tried to identify a greater number of ethical issues on rounds, which would not reflect actual practice.

Almost two decades ago, Coulehan and Williams wrote about the positive impact that ethics and humanities could have if these disciplines could be embedded in the daily practice of medicine, which is as follows:

…ethics and humanities curricula are irrelevant unless they can produce a substantive and continuing impact on hospital culture (…) The idea, of course, is to infiltrate the culture by coopting residents and attending physicians(…) If an ethics program can somehow achieve a critical mass of ‘‘value-sensitive’’ clinical faculty, it may begin to influence the institution’s ethos.¹²

Coulehan and Williams wrote of a need to bring ethics to the bedside. Our data suggest that an ethics epidemiology is deeply embedded in hospitalist services and is waiting to be fully characterized to better inform the care of patients and guide the professional formation and education of students and trainees. Hospitalists frequently confront ethical problems in daily practice that do not come to the attention of the CEC services or the institutional ethics committee. Understanding this emerging epidemiology presents an unrealized opportunity to improve bedside teaching, reinforce normative reasoning, and enhance patient care.

Acknowledgments

The authors want to acknowledge Drs. Augustine I. Choi, Michael G. Stewart, Laura L. Forese, and Anthony Hollenberg for their support of the fellowship in medical ethics and thank Drs. Arthur T. Evans and Monika M. Safford for their guidance.

Disclosures

The authors report no conflicts of interest.

Funding

This work was supported by a Weill Cornell General Internal Medicine Primary Care Innovations Initiative seed grant. Dr. Paul Christos was partially supported by the following grant: Clinical and Translational Science Center at Weill Cornell Medical College (1-UL1-TR002384-01).