Inpatient falls are the most common type of inpatient adverse event,1 persist as a significant problem nationally, and result in patient injury, increased length of stay, healthcare costs, and litigation.27 Inpatient falls remain a main focus of patient safety and a measure of quality in this era of healthcare reform and quality improvement.8 Inpatient fall rates per 1000 patient‐days range from 1.4 to 18.2.4, 9 The absolute percentage of inpatients that fall ranges from 1.3% to 7%.4, 5, 9, 10 Of inpatient falls, almost all data suggest that roughly one‐third result in some type of injury while 3%‐8% result in serious injury or death.9, 1113

Fall prevention interventions have largely been aimed at modifiable risk factors such as getting out of bed with bed alarms, toileting needs with bedside commodes, and reducing delirium through reorientation techniques. There have been several attempts at decreasing fall rates in hospitals surrounding a multidisciplinary, team‐based approach. Two Cochrane reviews and 2 meta‐analyses have partially examined this issue with mixed results.1417 However, none of these reviews focused on the acute care inpatient population. In fact, the majority of the data analyzed for inpatients was from rehabilitation wards and long‐term care wards. Additionally, there exists almost no data examining fall prevention with single interventions in the acute inpatient population, likely due to the belief that falls are multifactorial in etiology and require more comprehensive interventions.

The aim of this article is to determine the impact of team‐based, multidisciplinary quality improvement efforts to reduce inpatient falls in acute care inpatient hospitals and identify key features that determine their effectiveness.

METHODS

RESULTS

Fall Rates

Dykes et al22 and Williams et al24 found a statistically significant reduction in fall rate with falls reduced by 1.16 per 1000‐patient days and 1.5 per 1000‐patient days, respectively. Mitchell and Jones25 demonstrated a large fall reduction but had an extremely small sample size. Brandis21 found an extremely small reduction in fall rates and failed to report a P‐value. Krauss et al23 showed a trend towards reducing falls, and even showed a statistically significant reduction over the first 5 months of the study, but lost significance in the final 4 months. Similarly, Schwendimann et al9 saw more impressive fall reductions in the first year of the study that dissipated in the final 3 years of data collection.

Results from the meta‐analysis of the 6 studies comparing odds ratios are displayed quantitatively and as a forest plot in Figure 1. The figure shows results with 95% CI for each individual study and overall. There was no statistical evidence of heterogeneity between the studies or study designs. Although, due to the small number of studies included, there is poor power to detect true heterogeneity among studies. The magnitude of boxes shown is a relative sample size indicator. Using the random‐effects model, the summary odds ratio is 0.90 (95% CI, 0.83 to 0.99) (P = 0.02) (I² = 0%).27

Figure 1

DISCUSSION

The frequency and morbidity associated with inpatient falls is well established, based on reproduced epidemiologic data. Reducing these adverse events could reduce morbidity, mortality, and healthcare costs, and has become the focus of most hospitals quality and patient safety initiatives. The focus of this review was to examine multidisciplinary efforts to reduce falls in acute care inpatient hospitals. Despite the importance and scope of the problem, there is a paucity of research available on this topic, with a wide literature search yielding only 6 primary research studies.

Our major finding is that multidisciplinary fall prevention strategies have a statistically significant impact on fall rates with a combined OR of 0.90. While this review demonstrates a significant benefit to multidisciplinary fall prevention strategies in the acute inpatient population, the clinical impact of these efforts may be limited. Based on rates ranging from 1.7 to 9.5 falls per 1000‐patient days, multidisciplinary interventions would reduce falls by 1 to 10 falls per 10,000‐patient days using the combined OR calculated of 0.9. Using other available incidence data regarding inpatient falls,4, 9 a reasonable baseline frequency to consider would be 8 falls per 1000 patient‐days. Assuming that prevalence, the number needed to treat (NNT) to prevent a single inpatient fall is 1250 patient days. Furthermore, based on available data, only approximately one‐third of these falls result in injury and only a minor fraction of these results in serious injury.9, 1113 The magnitude of this apparent benefit in the context of fall incidence rates raises some concerns about cost‐effectiveness given the high staffing and systems needs that multidisciplinary prevention programs require. This also suggests that there are limitations when using inpatient falls as a measure of healthcare quality given the absence of high‐quality evidence demonstrating a viable solution to the problem. At present, the Center for Medicare and Medicaid services limit reimbursement for fall‐related injuries if they occur during an acute inpatient hospitalization.28

The complexity of the interventions used may help explain the limited impact. Krauss et al23 examined compliance to their interventions and found less than ideal results. They found only 36.4% of intervention floor patients had maintained a toileting schedule compared to 24.6% on control floors. Additionally, a greater proportion of patients on control floors had a physical or occupational therapy consult, and only 1.8% more patients on intervention floors had walking aids provided. These were all strategies emphasized on the intervention floors. Similarly, Schwendimann et al9 questioned their staff's adherence to protocol after fall prevention committee audits. This may help explain why a potential benefit lost statistical significance with time, based on a natural tendency towards more participation at the beginning of a new policy. Williams et al24 reported only a 64% compliance rate with fall care plan forms and 77% rate of missing information on fall care plans. A multidisciplinary fall prevention study that did not meet inclusion criteria (based on study population) yielded strongly positive results for which the authors commented mostly on changing of the hospital culture surrounding fall prevention as a key to their success.29 Adoptability of a multidisciplinary intervention will clearly impact adherence and the intervention's ultimate effectiveness.

Single intervention strategies, not analyzed in this review, are simpler to execute and adhere to. While these types of interventions may be superior, there is extremely limited data supporting or refuting patient fall benefits in the acute care inpatient population when using simple single interventions. However, some data generated partially on acute care geriatrics wards targeting patient education only showed benefit.30

Dykes et al22 was able to improve compliance rates by removing steps in the process of executing interventions with the FPTK built into the EMR. Importantly, the FPTK was compared against very similar fall prevention strategies, the difference being that patients randomized to the FPTK arm had the assessment and interventions automatically prompted on admission in the EMR. Adherence was measured through Morse Fall Scale completion rates (81% in control units versus 94% in intervention units).22 In many ways, the utility of this study was displaying a fall risk reduction by simply enhancing compliance using health information technology with automated alerts. Additionally, both arms of the study reported low fall rates compared to previously reported data, and there may have been larger benefit seen if the FPTK was compared against no fall prevention strategy. This diminishing of effect size may have been present in all studies reviewed, as usual hospital care commonly includes basic patient safety measures.

Another potential problem with the multidisciplinary fall prevention programs included in the meta‐analysis is the inability to target interventions. Each study employed a fall risk score in an attempt to focus resources on a select group of high‐risk patients. This method is problematic given that countless risk factors for inpatient falls have been identified in the literature. Factors that have been described range from clinical characteristics to laboratory tests.31 The most consistently reproducible patient‐related risks are altered mental status (including cognitive impairment and depression), altered mobility (particularly lower limb weakness), a history of falls, and toileting needs.13, 3236 Less consistency is seen with other traditional risk factors such as age, sedating medication, and length of stay.5, 13, 32, 3638 Attempting to risk‐stratify patients using simple and accurate assessment tools developed from these risk factors has proven to be very difficult. Many tools have been developed based on identified risk factors, but perform very poorly when trying to identify patients who will fall with reasonable specificity and positive predictive value.34, 3944 In fact, it has been demonstrated that using a nurse's judgment, a physician's opinion based on a patient's likelihood to wander or a simple 2‐question tool have all performed better than sophisticated risk calculators.33, 45, 46 Therefore, it is possible that interventions could benefit from including all patients, with de‐emphasis on unproven risk stratification tools.

In contrast to our findings, a modest risk reduction has been demonstrated in several primary articles and meta‐analyses in the subacute, rehabilitation, and long‐term care populations.15, 16, 4750 Additionally, a recent study has described a 63.9% risk reduction in a population that included medical, surgical, psychiatric, and rehabilitation wards.29 One important difference between these settings and the acute inpatient populations may be the amount of time and energy that can be dedicated to fall prevention and overall care planning. Another likely factor is the added challenge of preventing falls in patients with more active medical illnesses. In the acute care setting, a patient's chief complaint may not be completely addressed at the time of first mobilization and ambulation. This may be most relevant in patients who are admitted with syncope, seizure, vertigo, and dehydration.

Our study has several limitations; most notably, the available evidence is limited in quality and quantity. Furthermore, omission of unpublished data may also lead to effect bias, though this would likely be in the direction of ineffective interventions supporting a conclusion that multidisciplinary efforts have had only a small impact on fall rates. Ideally, future studies can limit confounding variables through randomization. However, it is difficult to adequately blind when studying a multidisciplinary fall intervention that depends on patient and provider participation. As a result, none of the papers reviewed met criteria for high quality. However, almost all available data examined in this review came from large sample sizes in which thoughtful interventions were used. Since an inpatient fall will not affect the majority of patients, it was crucial for these studies to recruit a large sample size to have adequate power to detect a difference in fall rates. However, each study used risk assessment tools, which are poor indicators of who will and will not fall in the hospital.34, 39, 42 This may suggest a need for improved risk assessment tools, or be further evidence to include all patients in fall prevention regardless of risk. Quantitative synthesis of multidisciplinary fall interventions has the added limitation of comparing complex, multifaceted treatments that are not perfectly uniform. It is our opinion that interventions are semi‐standardized using the grouping methods employed in Table 3.

Preventing inpatient falls remains a difficult issue to address while convincing data is lacking. Based on current evidence, multidisciplinary fall prevention efforts on acutely ill inpatients show a possible small benefit and should be explored from a cost‐effectiveness standpoint to ensure they garner appropriate investment. Many resources are required to run such teams including nursing staff, equipment, physical and occupational therapy staff, pharmacists, and specialized staff training. We are unaware of any such cost‐effectiveness data available. Effective interventions may be those that maximize compliance through health information technology, maintain staff dedication, increase staff availability, improve risk assessment, or include all patients regardless of calculated fall risk, and take the patient's chief complaint into account in the fall prevention strategy. Where resources are limited, it appears most reasonable to focus on major risk factors for inpatient falls that have independently been shown to be detrimental to outcomes, such as delirium.51 Additionally, using inpatient fall rates as a hospital quality measure may be premature, given the lack of proven efforts to lower fall rates. Multidisciplinary fall prevention efforts on acutely ill inpatients should be further studied using high‐quality, randomized trials. It remains to be seen whether these large programs are cost‐effective, or on balance clinically effective.

Inpatient falls are the most common type of inpatient adverse event,1 persist as a significant problem nationally, and result in patient injury, increased length of stay, healthcare costs, and litigation.27 Inpatient falls remain a main focus of patient safety and a measure of quality in this era of healthcare reform and quality improvement.8 Inpatient fall rates per 1000 patient‐days range from 1.4 to 18.2.4, 9 The absolute percentage of inpatients that fall ranges from 1.3% to 7%.4, 5, 9, 10 Of inpatient falls, almost all data suggest that roughly one‐third result in some type of injury while 3%‐8% result in serious injury or death.9, 1113

Fall prevention interventions have largely been aimed at modifiable risk factors such as getting out of bed with bed alarms, toileting needs with bedside commodes, and reducing delirium through reorientation techniques. There have been several attempts at decreasing fall rates in hospitals surrounding a multidisciplinary, team‐based approach. Two Cochrane reviews and 2 meta‐analyses have partially examined this issue with mixed results.1417 However, none of these reviews focused on the acute care inpatient population. In fact, the majority of the data analyzed for inpatients was from rehabilitation wards and long‐term care wards. Additionally, there exists almost no data examining fall prevention with single interventions in the acute inpatient population, likely due to the belief that falls are multifactorial in etiology and require more comprehensive interventions.

The aim of this article is to determine the impact of team‐based, multidisciplinary quality improvement efforts to reduce inpatient falls in acute care inpatient hospitals and identify key features that determine their effectiveness.

METHODS

RESULTS

Fall Rates

Dykes et al22 and Williams et al24 found a statistically significant reduction in fall rate with falls reduced by 1.16 per 1000‐patient days and 1.5 per 1000‐patient days, respectively. Mitchell and Jones25 demonstrated a large fall reduction but had an extremely small sample size. Brandis21 found an extremely small reduction in fall rates and failed to report a P‐value. Krauss et al23 showed a trend towards reducing falls, and even showed a statistically significant reduction over the first 5 months of the study, but lost significance in the final 4 months. Similarly, Schwendimann et al9 saw more impressive fall reductions in the first year of the study that dissipated in the final 3 years of data collection.

Results from the meta‐analysis of the 6 studies comparing odds ratios are displayed quantitatively and as a forest plot in Figure 1. The figure shows results with 95% CI for each individual study and overall. There was no statistical evidence of heterogeneity between the studies or study designs. Although, due to the small number of studies included, there is poor power to detect true heterogeneity among studies. The magnitude of boxes shown is a relative sample size indicator. Using the random‐effects model, the summary odds ratio is 0.90 (95% CI, 0.83 to 0.99) (P = 0.02) (I² = 0%).27

Figure 1

DISCUSSION

The frequency and morbidity associated with inpatient falls is well established, based on reproduced epidemiologic data. Reducing these adverse events could reduce morbidity, mortality, and healthcare costs, and has become the focus of most hospitals quality and patient safety initiatives. The focus of this review was to examine multidisciplinary efforts to reduce falls in acute care inpatient hospitals. Despite the importance and scope of the problem, there is a paucity of research available on this topic, with a wide literature search yielding only 6 primary research studies.

Our major finding is that multidisciplinary fall prevention strategies have a statistically significant impact on fall rates with a combined OR of 0.90. While this review demonstrates a significant benefit to multidisciplinary fall prevention strategies in the acute inpatient population, the clinical impact of these efforts may be limited. Based on rates ranging from 1.7 to 9.5 falls per 1000‐patient days, multidisciplinary interventions would reduce falls by 1 to 10 falls per 10,000‐patient days using the combined OR calculated of 0.9. Using other available incidence data regarding inpatient falls,4, 9 a reasonable baseline frequency to consider would be 8 falls per 1000 patient‐days. Assuming that prevalence, the number needed to treat (NNT) to prevent a single inpatient fall is 1250 patient days. Furthermore, based on available data, only approximately one‐third of these falls result in injury and only a minor fraction of these results in serious injury.9, 1113 The magnitude of this apparent benefit in the context of fall incidence rates raises some concerns about cost‐effectiveness given the high staffing and systems needs that multidisciplinary prevention programs require. This also suggests that there are limitations when using inpatient falls as a measure of healthcare quality given the absence of high‐quality evidence demonstrating a viable solution to the problem. At present, the Center for Medicare and Medicaid services limit reimbursement for fall‐related injuries if they occur during an acute inpatient hospitalization.28

The complexity of the interventions used may help explain the limited impact. Krauss et al23 examined compliance to their interventions and found less than ideal results. They found only 36.4% of intervention floor patients had maintained a toileting schedule compared to 24.6% on control floors. Additionally, a greater proportion of patients on control floors had a physical or occupational therapy consult, and only 1.8% more patients on intervention floors had walking aids provided. These were all strategies emphasized on the intervention floors. Similarly, Schwendimann et al9 questioned their staff's adherence to protocol after fall prevention committee audits. This may help explain why a potential benefit lost statistical significance with time, based on a natural tendency towards more participation at the beginning of a new policy. Williams et al24 reported only a 64% compliance rate with fall care plan forms and 77% rate of missing information on fall care plans. A multidisciplinary fall prevention study that did not meet inclusion criteria (based on study population) yielded strongly positive results for which the authors commented mostly on changing of the hospital culture surrounding fall prevention as a key to their success.29 Adoptability of a multidisciplinary intervention will clearly impact adherence and the intervention's ultimate effectiveness.

Single intervention strategies, not analyzed in this review, are simpler to execute and adhere to. While these types of interventions may be superior, there is extremely limited data supporting or refuting patient fall benefits in the acute care inpatient population when using simple single interventions. However, some data generated partially on acute care geriatrics wards targeting patient education only showed benefit.30

Dykes et al22 was able to improve compliance rates by removing steps in the process of executing interventions with the FPTK built into the EMR. Importantly, the FPTK was compared against very similar fall prevention strategies, the difference being that patients randomized to the FPTK arm had the assessment and interventions automatically prompted on admission in the EMR. Adherence was measured through Morse Fall Scale completion rates (81% in control units versus 94% in intervention units).22 In many ways, the utility of this study was displaying a fall risk reduction by simply enhancing compliance using health information technology with automated alerts. Additionally, both arms of the study reported low fall rates compared to previously reported data, and there may have been larger benefit seen if the FPTK was compared against no fall prevention strategy. This diminishing of effect size may have been present in all studies reviewed, as usual hospital care commonly includes basic patient safety measures.

Another potential problem with the multidisciplinary fall prevention programs included in the meta‐analysis is the inability to target interventions. Each study employed a fall risk score in an attempt to focus resources on a select group of high‐risk patients. This method is problematic given that countless risk factors for inpatient falls have been identified in the literature. Factors that have been described range from clinical characteristics to laboratory tests.31 The most consistently reproducible patient‐related risks are altered mental status (including cognitive impairment and depression), altered mobility (particularly lower limb weakness), a history of falls, and toileting needs.13, 3236 Less consistency is seen with other traditional risk factors such as age, sedating medication, and length of stay.5, 13, 32, 3638 Attempting to risk‐stratify patients using simple and accurate assessment tools developed from these risk factors has proven to be very difficult. Many tools have been developed based on identified risk factors, but perform very poorly when trying to identify patients who will fall with reasonable specificity and positive predictive value.34, 3944 In fact, it has been demonstrated that using a nurse's judgment, a physician's opinion based on a patient's likelihood to wander or a simple 2‐question tool have all performed better than sophisticated risk calculators.33, 45, 46 Therefore, it is possible that interventions could benefit from including all patients, with de‐emphasis on unproven risk stratification tools.

In contrast to our findings, a modest risk reduction has been demonstrated in several primary articles and meta‐analyses in the subacute, rehabilitation, and long‐term care populations.15, 16, 4750 Additionally, a recent study has described a 63.9% risk reduction in a population that included medical, surgical, psychiatric, and rehabilitation wards.29 One important difference between these settings and the acute inpatient populations may be the amount of time and energy that can be dedicated to fall prevention and overall care planning. Another likely factor is the added challenge of preventing falls in patients with more active medical illnesses. In the acute care setting, a patient's chief complaint may not be completely addressed at the time of first mobilization and ambulation. This may be most relevant in patients who are admitted with syncope, seizure, vertigo, and dehydration.

Our study has several limitations; most notably, the available evidence is limited in quality and quantity. Furthermore, omission of unpublished data may also lead to effect bias, though this would likely be in the direction of ineffective interventions supporting a conclusion that multidisciplinary efforts have had only a small impact on fall rates. Ideally, future studies can limit confounding variables through randomization. However, it is difficult to adequately blind when studying a multidisciplinary fall intervention that depends on patient and provider participation. As a result, none of the papers reviewed met criteria for high quality. However, almost all available data examined in this review came from large sample sizes in which thoughtful interventions were used. Since an inpatient fall will not affect the majority of patients, it was crucial for these studies to recruit a large sample size to have adequate power to detect a difference in fall rates. However, each study used risk assessment tools, which are poor indicators of who will and will not fall in the hospital.34, 39, 42 This may suggest a need for improved risk assessment tools, or be further evidence to include all patients in fall prevention regardless of risk. Quantitative synthesis of multidisciplinary fall interventions has the added limitation of comparing complex, multifaceted treatments that are not perfectly uniform. It is our opinion that interventions are semi‐standardized using the grouping methods employed in Table 3.

Preventing inpatient falls remains a difficult issue to address while convincing data is lacking. Based on current evidence, multidisciplinary fall prevention efforts on acutely ill inpatients show a possible small benefit and should be explored from a cost‐effectiveness standpoint to ensure they garner appropriate investment. Many resources are required to run such teams including nursing staff, equipment, physical and occupational therapy staff, pharmacists, and specialized staff training. We are unaware of any such cost‐effectiveness data available. Effective interventions may be those that maximize compliance through health information technology, maintain staff dedication, increase staff availability, improve risk assessment, or include all patients regardless of calculated fall risk, and take the patient's chief complaint into account in the fall prevention strategy. Where resources are limited, it appears most reasonable to focus on major risk factors for inpatient falls that have independently been shown to be detrimental to outcomes, such as delirium.51 Additionally, using inpatient fall rates as a hospital quality measure may be premature, given the lack of proven efforts to lower fall rates. Multidisciplinary fall prevention efforts on acutely ill inpatients should be further studied using high‐quality, randomized trials. It remains to be seen whether these large programs are cost‐effective, or on balance clinically effective.

Inpatient falls are the most common type of inpatient adverse event,1 persist as a significant problem nationally, and result in patient injury, increased length of stay, healthcare costs, and litigation.27 Inpatient falls remain a main focus of patient safety and a measure of quality in this era of healthcare reform and quality improvement.8 Inpatient fall rates per 1000 patient‐days range from 1.4 to 18.2.4, 9 The absolute percentage of inpatients that fall ranges from 1.3% to 7%.4, 5, 9, 10 Of inpatient falls, almost all data suggest that roughly one‐third result in some type of injury while 3%‐8% result in serious injury or death.9, 1113

Fall prevention interventions have largely been aimed at modifiable risk factors such as getting out of bed with bed alarms, toileting needs with bedside commodes, and reducing delirium through reorientation techniques. There have been several attempts at decreasing fall rates in hospitals surrounding a multidisciplinary, team‐based approach. Two Cochrane reviews and 2 meta‐analyses have partially examined this issue with mixed results.1417 However, none of these reviews focused on the acute care inpatient population. In fact, the majority of the data analyzed for inpatients was from rehabilitation wards and long‐term care wards. Additionally, there exists almost no data examining fall prevention with single interventions in the acute inpatient population, likely due to the belief that falls are multifactorial in etiology and require more comprehensive interventions.

The aim of this article is to determine the impact of team‐based, multidisciplinary quality improvement efforts to reduce inpatient falls in acute care inpatient hospitals and identify key features that determine their effectiveness.

METHODS

RESULTS

Fall Rates

Dykes et al22 and Williams et al24 found a statistically significant reduction in fall rate with falls reduced by 1.16 per 1000‐patient days and 1.5 per 1000‐patient days, respectively. Mitchell and Jones25 demonstrated a large fall reduction but had an extremely small sample size. Brandis21 found an extremely small reduction in fall rates and failed to report a P‐value. Krauss et al23 showed a trend towards reducing falls, and even showed a statistically significant reduction over the first 5 months of the study, but lost significance in the final 4 months. Similarly, Schwendimann et al9 saw more impressive fall reductions in the first year of the study that dissipated in the final 3 years of data collection.

Results from the meta‐analysis of the 6 studies comparing odds ratios are displayed quantitatively and as a forest plot in Figure 1. The figure shows results with 95% CI for each individual study and overall. There was no statistical evidence of heterogeneity between the studies or study designs. Although, due to the small number of studies included, there is poor power to detect true heterogeneity among studies. The magnitude of boxes shown is a relative sample size indicator. Using the random‐effects model, the summary odds ratio is 0.90 (95% CI, 0.83 to 0.99) (P = 0.02) (I² = 0%).27

Figure 1

DISCUSSION

The frequency and morbidity associated with inpatient falls is well established, based on reproduced epidemiologic data. Reducing these adverse events could reduce morbidity, mortality, and healthcare costs, and has become the focus of most hospitals quality and patient safety initiatives. The focus of this review was to examine multidisciplinary efforts to reduce falls in acute care inpatient hospitals. Despite the importance and scope of the problem, there is a paucity of research available on this topic, with a wide literature search yielding only 6 primary research studies.

Our major finding is that multidisciplinary fall prevention strategies have a statistically significant impact on fall rates with a combined OR of 0.90. While this review demonstrates a significant benefit to multidisciplinary fall prevention strategies in the acute inpatient population, the clinical impact of these efforts may be limited. Based on rates ranging from 1.7 to 9.5 falls per 1000‐patient days, multidisciplinary interventions would reduce falls by 1 to 10 falls per 10,000‐patient days using the combined OR calculated of 0.9. Using other available incidence data regarding inpatient falls,4, 9 a reasonable baseline frequency to consider would be 8 falls per 1000 patient‐days. Assuming that prevalence, the number needed to treat (NNT) to prevent a single inpatient fall is 1250 patient days. Furthermore, based on available data, only approximately one‐third of these falls result in injury and only a minor fraction of these results in serious injury.9, 1113 The magnitude of this apparent benefit in the context of fall incidence rates raises some concerns about cost‐effectiveness given the high staffing and systems needs that multidisciplinary prevention programs require. This also suggests that there are limitations when using inpatient falls as a measure of healthcare quality given the absence of high‐quality evidence demonstrating a viable solution to the problem. At present, the Center for Medicare and Medicaid services limit reimbursement for fall‐related injuries if they occur during an acute inpatient hospitalization.28

The complexity of the interventions used may help explain the limited impact. Krauss et al23 examined compliance to their interventions and found less than ideal results. They found only 36.4% of intervention floor patients had maintained a toileting schedule compared to 24.6% on control floors. Additionally, a greater proportion of patients on control floors had a physical or occupational therapy consult, and only 1.8% more patients on intervention floors had walking aids provided. These were all strategies emphasized on the intervention floors. Similarly, Schwendimann et al9 questioned their staff's adherence to protocol after fall prevention committee audits. This may help explain why a potential benefit lost statistical significance with time, based on a natural tendency towards more participation at the beginning of a new policy. Williams et al24 reported only a 64% compliance rate with fall care plan forms and 77% rate of missing information on fall care plans. A multidisciplinary fall prevention study that did not meet inclusion criteria (based on study population) yielded strongly positive results for which the authors commented mostly on changing of the hospital culture surrounding fall prevention as a key to their success.29 Adoptability of a multidisciplinary intervention will clearly impact adherence and the intervention's ultimate effectiveness.

Single intervention strategies, not analyzed in this review, are simpler to execute and adhere to. While these types of interventions may be superior, there is extremely limited data supporting or refuting patient fall benefits in the acute care inpatient population when using simple single interventions. However, some data generated partially on acute care geriatrics wards targeting patient education only showed benefit.30

Dykes et al22 was able to improve compliance rates by removing steps in the process of executing interventions with the FPTK built into the EMR. Importantly, the FPTK was compared against very similar fall prevention strategies, the difference being that patients randomized to the FPTK arm had the assessment and interventions automatically prompted on admission in the EMR. Adherence was measured through Morse Fall Scale completion rates (81% in control units versus 94% in intervention units).22 In many ways, the utility of this study was displaying a fall risk reduction by simply enhancing compliance using health information technology with automated alerts. Additionally, both arms of the study reported low fall rates compared to previously reported data, and there may have been larger benefit seen if the FPTK was compared against no fall prevention strategy. This diminishing of effect size may have been present in all studies reviewed, as usual hospital care commonly includes basic patient safety measures.

Another potential problem with the multidisciplinary fall prevention programs included in the meta‐analysis is the inability to target interventions. Each study employed a fall risk score in an attempt to focus resources on a select group of high‐risk patients. This method is problematic given that countless risk factors for inpatient falls have been identified in the literature. Factors that have been described range from clinical characteristics to laboratory tests.31 The most consistently reproducible patient‐related risks are altered mental status (including cognitive impairment and depression), altered mobility (particularly lower limb weakness), a history of falls, and toileting needs.13, 3236 Less consistency is seen with other traditional risk factors such as age, sedating medication, and length of stay.5, 13, 32, 3638 Attempting to risk‐stratify patients using simple and accurate assessment tools developed from these risk factors has proven to be very difficult. Many tools have been developed based on identified risk factors, but perform very poorly when trying to identify patients who will fall with reasonable specificity and positive predictive value.34, 3944 In fact, it has been demonstrated that using a nurse's judgment, a physician's opinion based on a patient's likelihood to wander or a simple 2‐question tool have all performed better than sophisticated risk calculators.33, 45, 46 Therefore, it is possible that interventions could benefit from including all patients, with de‐emphasis on unproven risk stratification tools.

In contrast to our findings, a modest risk reduction has been demonstrated in several primary articles and meta‐analyses in the subacute, rehabilitation, and long‐term care populations.15, 16, 4750 Additionally, a recent study has described a 63.9% risk reduction in a population that included medical, surgical, psychiatric, and rehabilitation wards.29 One important difference between these settings and the acute inpatient populations may be the amount of time and energy that can be dedicated to fall prevention and overall care planning. Another likely factor is the added challenge of preventing falls in patients with more active medical illnesses. In the acute care setting, a patient's chief complaint may not be completely addressed at the time of first mobilization and ambulation. This may be most relevant in patients who are admitted with syncope, seizure, vertigo, and dehydration.

Our study has several limitations; most notably, the available evidence is limited in quality and quantity. Furthermore, omission of unpublished data may also lead to effect bias, though this would likely be in the direction of ineffective interventions supporting a conclusion that multidisciplinary efforts have had only a small impact on fall rates. Ideally, future studies can limit confounding variables through randomization. However, it is difficult to adequately blind when studying a multidisciplinary fall intervention that depends on patient and provider participation. As a result, none of the papers reviewed met criteria for high quality. However, almost all available data examined in this review came from large sample sizes in which thoughtful interventions were used. Since an inpatient fall will not affect the majority of patients, it was crucial for these studies to recruit a large sample size to have adequate power to detect a difference in fall rates. However, each study used risk assessment tools, which are poor indicators of who will and will not fall in the hospital.34, 39, 42 This may suggest a need for improved risk assessment tools, or be further evidence to include all patients in fall prevention regardless of risk. Quantitative synthesis of multidisciplinary fall interventions has the added limitation of comparing complex, multifaceted treatments that are not perfectly uniform. It is our opinion that interventions are semi‐standardized using the grouping methods employed in Table 3.

Preventing inpatient falls remains a difficult issue to address while convincing data is lacking. Based on current evidence, multidisciplinary fall prevention efforts on acutely ill inpatients show a possible small benefit and should be explored from a cost‐effectiveness standpoint to ensure they garner appropriate investment. Many resources are required to run such teams including nursing staff, equipment, physical and occupational therapy staff, pharmacists, and specialized staff training. We are unaware of any such cost‐effectiveness data available. Effective interventions may be those that maximize compliance through health information technology, maintain staff dedication, increase staff availability, improve risk assessment, or include all patients regardless of calculated fall risk, and take the patient's chief complaint into account in the fall prevention strategy. Where resources are limited, it appears most reasonable to focus on major risk factors for inpatient falls that have independently been shown to be detrimental to outcomes, such as delirium.51 Additionally, using inpatient fall rates as a hospital quality measure may be premature, given the lack of proven efforts to lower fall rates. Multidisciplinary fall prevention efforts on acutely ill inpatients should be further studied using high‐quality, randomized trials. It remains to be seen whether these large programs are cost‐effective, or on balance clinically effective.

Adverse drug events (ADE), of which medication errors are one form, refer to harm caused by use of a drug. ADEs occur frequently and are associated with an increased length of stay, economic burden, and risk of death.1, 2 Classen et al and Bates et al estimate, respectively, that there are 1.2 to 1.8 preventable ADEs per 100 inpatient admissions.1, 3 Adjusting these data to current levels of yearly admissions, 380,000 to 400,000 preventable ADEs occur each year, and are projected to cost upwards of $3.5 billion annually in 2006 dollars.4

Medication reconciliation is an active process that occurs at transitions in care (admissions, transfers in level of care, and discharge) and is designed to prevent medication errors as the patient moves across the continuum of care. Medications used by the patient prior to hospitalization are considered when developing the inpatient therapeutic regimen.

Medications are ordered on admission based in part on what providers believe is the patient's home medication list (HML). A systematic review revealed that errors in medication history taking, including errors of omission and commission, are extremely common and clinically important.5 Such inaccuracies lead to unintended discrepancies between the hospital medication orders and the patient's true home medication regimen, and can result in patient harm.

Numerous studies have documented that inpatient discrepancies are common.610 From September 2004 to July 2005, data from the United States Pharmacopeia MEDMARX voluntary medication error reporting program revealed over 2000 medication errors associated with reconciliation failures: 22% occurred during admission and 12% occurred at time of discharge.11 A Canadian study demonstrated that 81 of 151 enrolled patients, who were prescribed 4 or more medications and were admitted to a medicine service, had at least 1 unintended discrepancy.6 Of those discrepancies, 38.6% were thought to have the potential to cause moderate or severe discomfort or clinical deterioration. Bates et al found that 0.9% of all inpatient medication errors lead to harm.12

The Joint Commission highlighted the importance of this problem by creating National Patient Safety Goal (NPSG) 8 in 2005, Accurately and completely reconcile medications across the continuum of care.13 This goal was modified and became effective on July 1, 2011.14 As a response, organizations have been developing physician‐led, nurse‐led, or pharmacist‐led medication reconciliation processes.8, 1522 Typically, these teams have time dedicated to producing the most accurate home list possible, a gold standard list. Examples of successful pharmacist‐led interventions to address this goal are described by investigators at Northwestern Memorial Hospital16 and Duke University Medical Center.23 Other interventions implemented to improve the reconciliation process include computerized provider order entry (CPOE) systems24 and combining information technology (IT) with process redesign involving physicians, pharmacists, and nurses.20 While the literature shows that there are multiple interventions that can reduce medication reconciliation errors, there is a dearth of evidence for interventions that are low‐cost and easily replicable.

Given that unintended medication discrepancies are common and harmful, we sought to develop a generalizable intervention. Our prospective pilot study explored whether an easily replicable nurse‐pharmacist led medication reconciliation process could efficiently and inexpensively identify unintended medication discrepancies, thereby preventing potential adverse drug events (PADEs).

METHODS

RESULTS

We enrolled 563 patients who were admitted a total of 698 times. Only the first admission for each patient was analyzed. Patient demographics are presented in Table 1. Almost 70% of our enrolled patients were less than 65 years old, 65% of the patients were black, 58% lived within 5 miles of the Johns Hopkins Hospital, and a plurality of the patients were single and received Medicare. The mean number of medications was 7.8 (SD 4.9).

The most frequent source of the home medication list was from patient verbal recall (52%). Few patients had lists of their current medications when admitted. The second most commonly used source was the electronic patient record, 36.6%, which was used to verify and complete the home list. The patient's community pharmacist, 12.5%, was contacted when other sources did not result in a complete home list. The primary care site was contacted in 6.0% of the cases. Patients were then asked to verify the HML.

Of the 563 patients, 225 (40%; 95% confidence interval [CI], 36%‐44%) had at least 1 unintended discrepancy on admission or discharge. On admission and discharge, 28% (95% CI, 25%‐30%) and 25% (95% CI, 21%‐29%) of the patients, respectively, had an unintended discrepancy. Of those 225 patients who had an unintended discrepancy, 162 (72%) had a discrepancy ranked 2 or 3 on the potential harm scale.

Overall, there were more unintended discrepancies on admission (364) than at discharge (167) (Figure 1). The paired t test showed a significant decrease (t[562] = 2.066, P = 0.039) between the number of discrepancies on admission to discharge. However, the majority of these discrepancies on admission (55%) were Rank 1 on the potential harm scale, while the majority of the discharge discrepancies (85%) were likely to cause harm (Rank 2‐3). There were many more Rank 3 discrepancies upon discharge, 39, than on admission, 13. The percentage of Rank 2‐3 discrepancies on admission and discharge were 45% versus 85%, respectively. Interclass correlation of ratings before consensus was 0.58.

Figure 1

The most common unintended discrepancies were omissions of medications at admission, 74%, and discharge, 62%, followed by discrepancies in dosing (Table 2). The majority of omission discrepancies were categorized as Rank 1. Discrepancies in frequency and dosing were most likely to be adjudicated as Rank 2 or 3. Table 3 gives examples of how discrepancies were ranked.

The only statistically significantly variable associated with the presence of discrepancies was the number of medications (odds ratio, 1.087; 95% CI, 1.044‐1.132). Each additional medication increased the odds of a discrepancy by 8.7%. Other variables, including age, race, length of stay, level of education, marital status, primary payor, and severity of illness, were not associated with prevalent discrepancies.

DISCUSSION

Inpatient medication reconciliation, an essential patient safety process, prevents potential ADEs and is mandated by The Joint Commission. Previous studies have shown that discrepancies are common occurrences for patients treated in tertiary centers,68 and those discrepancies can lead to patient discomfort or clinical deterioration.6, 8 Our current study supports this body of literature, as 40% of patients had at least 1 discrepancy on admission or discharge, and 29% of those discrepancies had the potential to result in moderate or severe discomfort or clinical deterioration. Although consistent with some findings,8 these numbers are generally lower than other studies,6, 25 where anywhere from 39% to 64% of the discrepancies were classified as Rank 2‐3.

Consistent with other studies, we found that omission was the most common type of discrepancy at admission as well as discharge.6, 8, 9, 21 In recent studies, omissions accounted for 46.5%10, 21 to 60%9 of the discrepancies. Further analysis in our study showed that the more medications a patient took, the higher the likelihood of discrepancya correlation also seen in other studies.9, 10 As the number of medications that a patient takes increases, the more difficult it becomes for all parties involved, including patients, families, and physicians, to keep an accurate recordleading to more opportunities for discrepancies.

Unintended medication discrepancies do not just occur on admission. While we identified many fewer discharge discrepancies, they were more likely to be categorized as Rank 2‐3. This is in contrast with other research that has found more discrepancies at discharge than admission.9, 11 In the current study, active medication reconciliation on admission likely led to a decrease in the number of discharge discrepancies. Even though there were fewer discharge discrepancies, the potential for harm was great and should not be underestimated.

Although many different types of interventions have been tried, this pilot study demonstrated a remarkably easy, generalizable, and inexpensive method. Other interventions have depended on wholesale reengineering of complicated processes,20, 26 pharmacists,10, 15, 16, 18, 19, 21, 22 or particular IT systems.20 Our intervention employed a nursing‐pharmacist model, which may either reduce the cost of healthcare, or at the very least, pay for itself. Each ADE is projected to cost $9300. The nurse‐pharmacist collaboration costs approximately $32 per patient. Thus, preventing only 1 ADE in 290 patient admissions would constitute a breakeven point for the interventiona goal that is likely achievable according to our study results. Even more cost‐effective would be to target only those patients at highest risk for a discrepancynamely those taking multiple home medications.10

There are several limitations to our study. First, we did not have a control group that would allow for comparison of clinical outcomes between the intervention and standard practice. Second, only potential ADEs were avoided. We were not able to determine that an ADE would definitely have occurred if the reconciliation had not taken place. Third, this study was conducted in a single department at 1 institution. As such, the results may not be generalizable to services other than general medicine or to other hospitals. Fourth, we relied on cost data from 1 inpatient study that is more than a decade old to estimate the potential savings to the healthcare system.2 This demonstrates the need for new studies of the cost of ADEs in hospital and outpatient settings. Outpatient medication discrepancies may be more or less costly than their inpatient counterparts, which would impact the cost analysis of this study. Fifth, we did not rely on the brown bag method, asking the patient's family to bring in the medication bottles, for determining the HML. That would certainly have given us another method to confirm the HML. Moreover, the nurse did not confirm the HML with a second source if she felt that the list provided by the patient was accurate. Finally, while we can intervene on discharge discrepancies, we do not control what a patient chooses to do after discharge.27 Health literacy, financial issues, deficits in communication between patients' discharge providers and their primary care providers, and many other factors affect whether patients adhere to their discharge medication list.28

Since this is not a randomized controlled trial, this pilot study requires additional testing to determine if ADEs are actually avoided and costs saved. The HML protocol could be updated to include the brown bag method or other additional steps to verify the HML. Although not inexpensive, a home visit intervention could be tested as well.29, 30

In conclusion, potentially harmful unintended medication discrepancies occurred frequently at both hospital admission and discharge. A nurse‐pharmacist collaboration to monitor and intervene on these discrepancies allowed many to be reconciled before potentially causing harm to patients. The collaboration was relatively efficient and cost‐effective, and the process potentially improves patient safety.

Adverse drug events (ADE), of which medication errors are one form, refer to harm caused by use of a drug. ADEs occur frequently and are associated with an increased length of stay, economic burden, and risk of death.1, 2 Classen et al and Bates et al estimate, respectively, that there are 1.2 to 1.8 preventable ADEs per 100 inpatient admissions.1, 3 Adjusting these data to current levels of yearly admissions, 380,000 to 400,000 preventable ADEs occur each year, and are projected to cost upwards of $3.5 billion annually in 2006 dollars.4

Medication reconciliation is an active process that occurs at transitions in care (admissions, transfers in level of care, and discharge) and is designed to prevent medication errors as the patient moves across the continuum of care. Medications used by the patient prior to hospitalization are considered when developing the inpatient therapeutic regimen.

Medications are ordered on admission based in part on what providers believe is the patient's home medication list (HML). A systematic review revealed that errors in medication history taking, including errors of omission and commission, are extremely common and clinically important.5 Such inaccuracies lead to unintended discrepancies between the hospital medication orders and the patient's true home medication regimen, and can result in patient harm.

Numerous studies have documented that inpatient discrepancies are common.610 From September 2004 to July 2005, data from the United States Pharmacopeia MEDMARX voluntary medication error reporting program revealed over 2000 medication errors associated with reconciliation failures: 22% occurred during admission and 12% occurred at time of discharge.11 A Canadian study demonstrated that 81 of 151 enrolled patients, who were prescribed 4 or more medications and were admitted to a medicine service, had at least 1 unintended discrepancy.6 Of those discrepancies, 38.6% were thought to have the potential to cause moderate or severe discomfort or clinical deterioration. Bates et al found that 0.9% of all inpatient medication errors lead to harm.12

The Joint Commission highlighted the importance of this problem by creating National Patient Safety Goal (NPSG) 8 in 2005, Accurately and completely reconcile medications across the continuum of care.13 This goal was modified and became effective on July 1, 2011.14 As a response, organizations have been developing physician‐led, nurse‐led, or pharmacist‐led medication reconciliation processes.8, 1522 Typically, these teams have time dedicated to producing the most accurate home list possible, a gold standard list. Examples of successful pharmacist‐led interventions to address this goal are described by investigators at Northwestern Memorial Hospital16 and Duke University Medical Center.23 Other interventions implemented to improve the reconciliation process include computerized provider order entry (CPOE) systems24 and combining information technology (IT) with process redesign involving physicians, pharmacists, and nurses.20 While the literature shows that there are multiple interventions that can reduce medication reconciliation errors, there is a dearth of evidence for interventions that are low‐cost and easily replicable.

Given that unintended medication discrepancies are common and harmful, we sought to develop a generalizable intervention. Our prospective pilot study explored whether an easily replicable nurse‐pharmacist led medication reconciliation process could efficiently and inexpensively identify unintended medication discrepancies, thereby preventing potential adverse drug events (PADEs).

METHODS

RESULTS

We enrolled 563 patients who were admitted a total of 698 times. Only the first admission for each patient was analyzed. Patient demographics are presented in Table 1. Almost 70% of our enrolled patients were less than 65 years old, 65% of the patients were black, 58% lived within 5 miles of the Johns Hopkins Hospital, and a plurality of the patients were single and received Medicare. The mean number of medications was 7.8 (SD 4.9).

The most frequent source of the home medication list was from patient verbal recall (52%). Few patients had lists of their current medications when admitted. The second most commonly used source was the electronic patient record, 36.6%, which was used to verify and complete the home list. The patient's community pharmacist, 12.5%, was contacted when other sources did not result in a complete home list. The primary care site was contacted in 6.0% of the cases. Patients were then asked to verify the HML.

Of the 563 patients, 225 (40%; 95% confidence interval [CI], 36%‐44%) had at least 1 unintended discrepancy on admission or discharge. On admission and discharge, 28% (95% CI, 25%‐30%) and 25% (95% CI, 21%‐29%) of the patients, respectively, had an unintended discrepancy. Of those 225 patients who had an unintended discrepancy, 162 (72%) had a discrepancy ranked 2 or 3 on the potential harm scale.

Overall, there were more unintended discrepancies on admission (364) than at discharge (167) (Figure 1). The paired t test showed a significant decrease (t[562] = 2.066, P = 0.039) between the number of discrepancies on admission to discharge. However, the majority of these discrepancies on admission (55%) were Rank 1 on the potential harm scale, while the majority of the discharge discrepancies (85%) were likely to cause harm (Rank 2‐3). There were many more Rank 3 discrepancies upon discharge, 39, than on admission, 13. The percentage of Rank 2‐3 discrepancies on admission and discharge were 45% versus 85%, respectively. Interclass correlation of ratings before consensus was 0.58.

Figure 1

The most common unintended discrepancies were omissions of medications at admission, 74%, and discharge, 62%, followed by discrepancies in dosing (Table 2). The majority of omission discrepancies were categorized as Rank 1. Discrepancies in frequency and dosing were most likely to be adjudicated as Rank 2 or 3. Table 3 gives examples of how discrepancies were ranked.

The only statistically significantly variable associated with the presence of discrepancies was the number of medications (odds ratio, 1.087; 95% CI, 1.044‐1.132). Each additional medication increased the odds of a discrepancy by 8.7%. Other variables, including age, race, length of stay, level of education, marital status, primary payor, and severity of illness, were not associated with prevalent discrepancies.

DISCUSSION

Inpatient medication reconciliation, an essential patient safety process, prevents potential ADEs and is mandated by The Joint Commission. Previous studies have shown that discrepancies are common occurrences for patients treated in tertiary centers,68 and those discrepancies can lead to patient discomfort or clinical deterioration.6, 8 Our current study supports this body of literature, as 40% of patients had at least 1 discrepancy on admission or discharge, and 29% of those discrepancies had the potential to result in moderate or severe discomfort or clinical deterioration. Although consistent with some findings,8 these numbers are generally lower than other studies,6, 25 where anywhere from 39% to 64% of the discrepancies were classified as Rank 2‐3.

Consistent with other studies, we found that omission was the most common type of discrepancy at admission as well as discharge.6, 8, 9, 21 In recent studies, omissions accounted for 46.5%10, 21 to 60%9 of the discrepancies. Further analysis in our study showed that the more medications a patient took, the higher the likelihood of discrepancya correlation also seen in other studies.9, 10 As the number of medications that a patient takes increases, the more difficult it becomes for all parties involved, including patients, families, and physicians, to keep an accurate recordleading to more opportunities for discrepancies.

Unintended medication discrepancies do not just occur on admission. While we identified many fewer discharge discrepancies, they were more likely to be categorized as Rank 2‐3. This is in contrast with other research that has found more discrepancies at discharge than admission.9, 11 In the current study, active medication reconciliation on admission likely led to a decrease in the number of discharge discrepancies. Even though there were fewer discharge discrepancies, the potential for harm was great and should not be underestimated.

Although many different types of interventions have been tried, this pilot study demonstrated a remarkably easy, generalizable, and inexpensive method. Other interventions have depended on wholesale reengineering of complicated processes,20, 26 pharmacists,10, 15, 16, 18, 19, 21, 22 or particular IT systems.20 Our intervention employed a nursing‐pharmacist model, which may either reduce the cost of healthcare, or at the very least, pay for itself. Each ADE is projected to cost $9300. The nurse‐pharmacist collaboration costs approximately $32 per patient. Thus, preventing only 1 ADE in 290 patient admissions would constitute a breakeven point for the interventiona goal that is likely achievable according to our study results. Even more cost‐effective would be to target only those patients at highest risk for a discrepancynamely those taking multiple home medications.10

There are several limitations to our study. First, we did not have a control group that would allow for comparison of clinical outcomes between the intervention and standard practice. Second, only potential ADEs were avoided. We were not able to determine that an ADE would definitely have occurred if the reconciliation had not taken place. Third, this study was conducted in a single department at 1 institution. As such, the results may not be generalizable to services other than general medicine or to other hospitals. Fourth, we relied on cost data from 1 inpatient study that is more than a decade old to estimate the potential savings to the healthcare system.2 This demonstrates the need for new studies of the cost of ADEs in hospital and outpatient settings. Outpatient medication discrepancies may be more or less costly than their inpatient counterparts, which would impact the cost analysis of this study. Fifth, we did not rely on the brown bag method, asking the patient's family to bring in the medication bottles, for determining the HML. That would certainly have given us another method to confirm the HML. Moreover, the nurse did not confirm the HML with a second source if she felt that the list provided by the patient was accurate. Finally, while we can intervene on discharge discrepancies, we do not control what a patient chooses to do after discharge.27 Health literacy, financial issues, deficits in communication between patients' discharge providers and their primary care providers, and many other factors affect whether patients adhere to their discharge medication list.28

Since this is not a randomized controlled trial, this pilot study requires additional testing to determine if ADEs are actually avoided and costs saved. The HML protocol could be updated to include the brown bag method or other additional steps to verify the HML. Although not inexpensive, a home visit intervention could be tested as well.29, 30

In conclusion, potentially harmful unintended medication discrepancies occurred frequently at both hospital admission and discharge. A nurse‐pharmacist collaboration to monitor and intervene on these discrepancies allowed many to be reconciled before potentially causing harm to patients. The collaboration was relatively efficient and cost‐effective, and the process potentially improves patient safety.

Adverse drug events (ADE), of which medication errors are one form, refer to harm caused by use of a drug. ADEs occur frequently and are associated with an increased length of stay, economic burden, and risk of death.1, 2 Classen et al and Bates et al estimate, respectively, that there are 1.2 to 1.8 preventable ADEs per 100 inpatient admissions.1, 3 Adjusting these data to current levels of yearly admissions, 380,000 to 400,000 preventable ADEs occur each year, and are projected to cost upwards of $3.5 billion annually in 2006 dollars.4

Medication reconciliation is an active process that occurs at transitions in care (admissions, transfers in level of care, and discharge) and is designed to prevent medication errors as the patient moves across the continuum of care. Medications used by the patient prior to hospitalization are considered when developing the inpatient therapeutic regimen.

Medications are ordered on admission based in part on what providers believe is the patient's home medication list (HML). A systematic review revealed that errors in medication history taking, including errors of omission and commission, are extremely common and clinically important.5 Such inaccuracies lead to unintended discrepancies between the hospital medication orders and the patient's true home medication regimen, and can result in patient harm.

Numerous studies have documented that inpatient discrepancies are common.610 From September 2004 to July 2005, data from the United States Pharmacopeia MEDMARX voluntary medication error reporting program revealed over 2000 medication errors associated with reconciliation failures: 22% occurred during admission and 12% occurred at time of discharge.11 A Canadian study demonstrated that 81 of 151 enrolled patients, who were prescribed 4 or more medications and were admitted to a medicine service, had at least 1 unintended discrepancy.6 Of those discrepancies, 38.6% were thought to have the potential to cause moderate or severe discomfort or clinical deterioration. Bates et al found that 0.9% of all inpatient medication errors lead to harm.12

The Joint Commission highlighted the importance of this problem by creating National Patient Safety Goal (NPSG) 8 in 2005, Accurately and completely reconcile medications across the continuum of care.13 This goal was modified and became effective on July 1, 2011.14 As a response, organizations have been developing physician‐led, nurse‐led, or pharmacist‐led medication reconciliation processes.8, 1522 Typically, these teams have time dedicated to producing the most accurate home list possible, a gold standard list. Examples of successful pharmacist‐led interventions to address this goal are described by investigators at Northwestern Memorial Hospital16 and Duke University Medical Center.23 Other interventions implemented to improve the reconciliation process include computerized provider order entry (CPOE) systems24 and combining information technology (IT) with process redesign involving physicians, pharmacists, and nurses.20 While the literature shows that there are multiple interventions that can reduce medication reconciliation errors, there is a dearth of evidence for interventions that are low‐cost and easily replicable.

Given that unintended medication discrepancies are common and harmful, we sought to develop a generalizable intervention. Our prospective pilot study explored whether an easily replicable nurse‐pharmacist led medication reconciliation process could efficiently and inexpensively identify unintended medication discrepancies, thereby preventing potential adverse drug events (PADEs).

METHODS

RESULTS

We enrolled 563 patients who were admitted a total of 698 times. Only the first admission for each patient was analyzed. Patient demographics are presented in Table 1. Almost 70% of our enrolled patients were less than 65 years old, 65% of the patients were black, 58% lived within 5 miles of the Johns Hopkins Hospital, and a plurality of the patients were single and received Medicare. The mean number of medications was 7.8 (SD 4.9).

The most frequent source of the home medication list was from patient verbal recall (52%). Few patients had lists of their current medications when admitted. The second most commonly used source was the electronic patient record, 36.6%, which was used to verify and complete the home list. The patient's community pharmacist, 12.5%, was contacted when other sources did not result in a complete home list. The primary care site was contacted in 6.0% of the cases. Patients were then asked to verify the HML.

Of the 563 patients, 225 (40%; 95% confidence interval [CI], 36%‐44%) had at least 1 unintended discrepancy on admission or discharge. On admission and discharge, 28% (95% CI, 25%‐30%) and 25% (95% CI, 21%‐29%) of the patients, respectively, had an unintended discrepancy. Of those 225 patients who had an unintended discrepancy, 162 (72%) had a discrepancy ranked 2 or 3 on the potential harm scale.

Overall, there were more unintended discrepancies on admission (364) than at discharge (167) (Figure 1). The paired t test showed a significant decrease (t[562] = 2.066, P = 0.039) between the number of discrepancies on admission to discharge. However, the majority of these discrepancies on admission (55%) were Rank 1 on the potential harm scale, while the majority of the discharge discrepancies (85%) were likely to cause harm (Rank 2‐3). There were many more Rank 3 discrepancies upon discharge, 39, than on admission, 13. The percentage of Rank 2‐3 discrepancies on admission and discharge were 45% versus 85%, respectively. Interclass correlation of ratings before consensus was 0.58.

Figure 1

The most common unintended discrepancies were omissions of medications at admission, 74%, and discharge, 62%, followed by discrepancies in dosing (Table 2). The majority of omission discrepancies were categorized as Rank 1. Discrepancies in frequency and dosing were most likely to be adjudicated as Rank 2 or 3. Table 3 gives examples of how discrepancies were ranked.

The only statistically significantly variable associated with the presence of discrepancies was the number of medications (odds ratio, 1.087; 95% CI, 1.044‐1.132). Each additional medication increased the odds of a discrepancy by 8.7%. Other variables, including age, race, length of stay, level of education, marital status, primary payor, and severity of illness, were not associated with prevalent discrepancies.

DISCUSSION

Inpatient medication reconciliation, an essential patient safety process, prevents potential ADEs and is mandated by The Joint Commission. Previous studies have shown that discrepancies are common occurrences for patients treated in tertiary centers,68 and those discrepancies can lead to patient discomfort or clinical deterioration.6, 8 Our current study supports this body of literature, as 40% of patients had at least 1 discrepancy on admission or discharge, and 29% of those discrepancies had the potential to result in moderate or severe discomfort or clinical deterioration. Although consistent with some findings,8 these numbers are generally lower than other studies,6, 25 where anywhere from 39% to 64% of the discrepancies were classified as Rank 2‐3.

Consistent with other studies, we found that omission was the most common type of discrepancy at admission as well as discharge.6, 8, 9, 21 In recent studies, omissions accounted for 46.5%10, 21 to 60%9 of the discrepancies. Further analysis in our study showed that the more medications a patient took, the higher the likelihood of discrepancya correlation also seen in other studies.9, 10 As the number of medications that a patient takes increases, the more difficult it becomes for all parties involved, including patients, families, and physicians, to keep an accurate recordleading to more opportunities for discrepancies.

Unintended medication discrepancies do not just occur on admission. While we identified many fewer discharge discrepancies, they were more likely to be categorized as Rank 2‐3. This is in contrast with other research that has found more discrepancies at discharge than admission.9, 11 In the current study, active medication reconciliation on admission likely led to a decrease in the number of discharge discrepancies. Even though there were fewer discharge discrepancies, the potential for harm was great and should not be underestimated.

Although many different types of interventions have been tried, this pilot study demonstrated a remarkably easy, generalizable, and inexpensive method. Other interventions have depended on wholesale reengineering of complicated processes,20, 26 pharmacists,10, 15, 16, 18, 19, 21, 22 or particular IT systems.20 Our intervention employed a nursing‐pharmacist model, which may either reduce the cost of healthcare, or at the very least, pay for itself. Each ADE is projected to cost $9300. The nurse‐pharmacist collaboration costs approximately $32 per patient. Thus, preventing only 1 ADE in 290 patient admissions would constitute a breakeven point for the interventiona goal that is likely achievable according to our study results. Even more cost‐effective would be to target only those patients at highest risk for a discrepancynamely those taking multiple home medications.10

There are several limitations to our study. First, we did not have a control group that would allow for comparison of clinical outcomes between the intervention and standard practice. Second, only potential ADEs were avoided. We were not able to determine that an ADE would definitely have occurred if the reconciliation had not taken place. Third, this study was conducted in a single department at 1 institution. As such, the results may not be generalizable to services other than general medicine or to other hospitals. Fourth, we relied on cost data from 1 inpatient study that is more than a decade old to estimate the potential savings to the healthcare system.2 This demonstrates the need for new studies of the cost of ADEs in hospital and outpatient settings. Outpatient medication discrepancies may be more or less costly than their inpatient counterparts, which would impact the cost analysis of this study. Fifth, we did not rely on the brown bag method, asking the patient's family to bring in the medication bottles, for determining the HML. That would certainly have given us another method to confirm the HML. Moreover, the nurse did not confirm the HML with a second source if she felt that the list provided by the patient was accurate. Finally, while we can intervene on discharge discrepancies, we do not control what a patient chooses to do after discharge.27 Health literacy, financial issues, deficits in communication between patients' discharge providers and their primary care providers, and many other factors affect whether patients adhere to their discharge medication list.28

Since this is not a randomized controlled trial, this pilot study requires additional testing to determine if ADEs are actually avoided and costs saved. The HML protocol could be updated to include the brown bag method or other additional steps to verify the HML. Although not inexpensive, a home visit intervention could be tested as well.29, 30

In conclusion, potentially harmful unintended medication discrepancies occurred frequently at both hospital admission and discharge. A nurse‐pharmacist collaboration to monitor and intervene on these discrepancies allowed many to be reconciled before potentially causing harm to patients. The collaboration was relatively efficient and cost‐effective, and the process potentially improves patient safety.

Caring for patients with acute decompensated heart failure (ADHF) is one of the core competencies of practice in hospitalist medicine. Congestive heart failure remains the most common discharge diagnosis as recorded in the National Hospital Discharge Survey, with over 1.1 million hospitalizations for heart failure in 2004.1 Furthermore, with the disproportionate growth in the population over age 65 that will occur over the next 20 years, heart failure prevalence will grow from its current value of 2.8% to 3.5% by 2030.2 This will result in an additional 3 million Americans with chronic heart failure, thereby sustaining ADHF as the most common reason for hospital admission. Despite an average hospital stay of 5 days, the readmission rate for heart failure was 26.9% at 30 days in a 2003‐2004 analysis of Medicare data.3 This high readmission rate is the target of reform as part of the recently passed Patient Protection and Accountability Act. Starting in fiscal year 2013, acute‐care hospitals with higher‐than‐expected readmission rates for heart failure will have a reduction in reimbursement for these admissions.4 Thus, there is substantial incentive for hospitalists to focus on providing the highest quality of care for patients with ADHF. Here we review the most recent evidence applicable to hospitalists for the diagnosis, risk stratification, and management of patients presenting with ADHF.

DIAGNOSIS

The hospitalist can establish the ADHF diagnosis efficiently by applying a structured approach based on the patient's symptoms, history, physical examination, and laboratory testing. The typical symptoms of ADHF include dyspnea, orthopnea, paroxysmal nocturnal dyspnea (PND), and lower extremity edema. In particular, patients complaining of PND and/or orthopnea are likely to have ADHF.5, 6 Patients may also report chest congestion or chest pain in an atypical pattern. A history of rapid weight gain suggests fluid overload, hence determination of the patient's dry weight is important to establish a target for congestive therapy. Patients with advanced systolic heart failure may also complain of nausea, abdominal pain, and abdominal fullness from ascites.7 In a patient with dyspnea, a history of heart failure, myocardial infarction, or coronary artery disease, all make the diagnosis of ADHF more likely.5

Performing a careful physical examination on a patient presenting with suspected ADHF will not only establish the diagnosis of heart failure, but also determine the hemodynamic profile. Patients presenting with ADHF can be separated into 4 hemodynamic profiles, based on vital sign and physical exam parameters: the presence or absence of congestion (wet or dry), and the presence or absence of adequate perfusion (warm or cold) (Figure 1).8 Parameters indicating the presence of congestion include: orthopnea, elevated jugular venous pulsation (JVP), lower extremity edema, hepatojugular reflux, ascites, and a loud P2 heart sound. Notably, rales are an uncommon physical finding in patients with ADHF, likely because pulmonary lymphatics compensate for chronically elevated filling pressures in such patients.9, 10 Parameters indicating inadequate perfusion include: hypotension (mean arterial pressure 60 mmHg), proportional pulse pressure 25%, cool extremities, altered mental status, and poor urine output (0.5 mL/kg/hr). We recommend assigning the patient to 1 of these 4 hemodynamic profiles, as the profile correlates with invasive hemodynamic measurements of pulmonary capillary wedge pressure and cardiac index, guides management, and predicts outcome.

Figure 1

Natriuretic peptide testing may help establish or exclude a diagnosis of ADHF. A recent expert consensus paper on natriuretic peptide testing recommends cutpoints for both B‐type natriuretic peptide (BNP) and N‐terminal proBNP (NT‐BNP) that indicate a very low (BNP 100 or NT‐BNP 300), intermediate (BNP 100‐400 or NT‐BNP 300‐1800), and high (BNP >400 or NT‐BNP >1800) probability of heart failure11 (Figure 2). However, 2 common conditions affect the utility of BNP testing. First, obese patients have lower levels, and thus a lower rule‐out cutpoint of 54 pg/mL is recommended when using BNP, whereas the cutpoint for NT‐BNP remains the same.12, 13 Second, in patients with renal dysfunction, levels are increased, and thus higher rule‐out cutpoints of 200 pg/mL (for BNP) and 1200 pg/mL (for NT‐BNP) are recommended for patients with a glomerular filtration rate 60 mL/min.14, 15 For patients with longstanding heart failure and chronically elevated levels of natriuretic peptides, there is a correlation between BNP levels and left ventricular filling pressure,16 but the change is more helpful than the absolute levels; a 50% increase over baseline, in conjunction with symptoms, usually reflects ADHF.11

Figure 2

Chest radiography will establish the presence or absence of pulmonary congestion. Classic teaching is that congestion starts with cephalization (pulmonary capillary wedge pressure 10‐15 mmHg), progresses to Kerley B lines (15‐20 mmHg), then to interstitial edema (20‐25 mmHg), and finally to alveolar edema (>25 mmHg).17 In patients presenting with dyspnea, any of these findings helps to establish the diagnosis of ADHF.5

MECHANISMS AND TERMINOLOGY

Data from ADHF registries show that hemodynamically stable patients presenting to the hospital with ADHF are an approximately equal mix of heart failure with reduced ejection fraction (HFrEF; ejection fraction 50%) and heart failure with preserved ejection fraction (HFpEF; ejection fraction 50%).18, 19 The important differences between these groups with regards to pathophysiology and etiology have been reviewed elsewhere.20 Establishing the heart failure mechanism (ie, reduced or preserved EF) is important because the medical management is distinct. Patients with HFrEF are more likely to be male, younger in age, to have ischemic heart disease, and to present with normal or low blood pressure. Patients with HFpEF are more likely to be female, older in age, to have hypertension or diabetes mellitus, and to present with elevated blood pressure.18, 19

The terminology used for inpatient heart failure coding has been the subject of renewed focus. For fiscal year 2008, the Centers for Medicare and Medicaid Services (CMS) overhauled its Diagnosis Related Group (DRG) system to better account for the severity of illness of hospitalized patients.21 In this revision, the existing DRG codes for heart failure were subdivided into 3 severity subclasses: major complication, complication, and non‐complication. Payment to hospitals for a heart failure DRG was changed to be proportional to the level of complication. Thus, for the first time, the clinicians' assessment of the acuity of heart failure determines the level of payment to the hospital. Not surprisingly, this has led to initiatives by hospitals to improve clinicians' coding of inpatients hospitalized with heart failure. A major impediment is that there are no established criteria for the application of each DRG code. Table 1 presents recommended clinical criteria for the application of these codes to patients with ADHF.

PRECIPITANTS AND ETIOLOGY

For patients presenting for the first time with a diagnosis of ADHF (de novo), a thorough evaluation should be performed to determine the mechanism and etiology of the patient's left ventricular dysfunction. After the initial history and physical exam, the American College of Cardiology/American Heart Association (ACC/AHA) guidelines recommend checking basic laboratory studies, an electrocardiogram, and an echocardiogram.22 The full assessment recommended by the ACC/AHA is detailed in Supporting Online Table 1 (in the online version of this article). Cardiac ischemia is the most common etiology of HFrEF, accounting for about 50% of cases. The common, non‐ischemic causes of systolic heart failure include atrial fibrillation, aortic stenosis, illicit cardiotoxic drugs (cocaine, methamphetamine), medical cardiotoxic drugs (adriamycin), as well as primary myocardial disorders such as myocarditis, idiopathic, or peripartum cardiomyopathy. HFpEF is most commonly associated with long‐standing hypertension and diabetes mellitus, but can also be caused by infiltrative, hypertrophic, and constrictive cardiomyopathies.

For patients with a history of heart failure, it is important to identify the precipitant for the decompensation, as it may be treated or avoided in the future. When no clear precipitant is identified, this is most concerning, as it indicates the patient's tenuous cardiac function. In the Organized Program to Initiate Lifesaving Treatment in Hospitalized Patients with Heart Failure (OPTIMIZE‐HF) registry, approximately 61% of patients were found to have at least 1 precipitating factor.23 The most common precipitants were respiratory process in 15.3%, acute coronary syndrome in 14.7%, arrhythmia in 13.5%, uncontrolled hypertension in 10.7%, medication non‐compliance in 8.9%, worsening renal function in 8.0%, and dietary non‐compliance in 5.2%.

RISK STRATIFICATION

Patients hospitalized with ADHF are at a significantly elevated risk for death, both during their hospitalization and after discharge. Numerous studies have shown that multiple clinical parameters assessed during the hospitalization, such as vital signs and laboratory values, predict outcome.6, 8, 24, 25 Some of the most elegant parameters are physical exam findings. As introduced above, the wet‐cold hemodynamic profile assessed at admission predicts increased mortality and urgent transplantation at 1 year.8 One of the most powerful risk stratification schemes for in‐hospital mortality is that developed from the Acute Decompensated Heart Failure (ADHERE) national registry. Three clinical parameters, blood urea nitrogen (BUN) >43 mg/dL, systolic blood pressure 115 mmHg, and serum creatinine >2.75 mg/dL, stratified patients into risk groups. Patients exhibiting all 3 parameters had a 22% in‐hospital mortality compared with 2% for patients with none of the 3 parameters.24

BNP and troponin also have a role in risk stratification of patients with ADHF. In the ADHERE registry, for every increase in the BNP of 400 pg/mL, the odds of risk‐adjusted mortality increased by 9%, in patients with both HFrEF and HFpEF.26 Similarly, an elevated admission troponin was associated with an in‐hospital mortality of 8.0%, versus 2.7% for troponin‐negative patients27; notably almost half of patients with a positive troponin had no history of ischemic heart disease. In the future, refinement and widespread application of these risk stratification methods should allow clinicians to triage patients to determine their location (eg, observation unit, inpatient, intensive care unit) and type of treatment (eg, oral or intravenous diuretic, vasodilator, inotrope).28

In the community, hospitalists care for many patients with ADHF without input from a cardiologist.29 However, there are several situations where the patient is at an increased risk of adverse outcomes, and therefore in which we recommend consulting a cardiologist (Table 2). Patients with hypotension, a cold hemodynamic profile, or worsening renal function due to poor cardiac function are at an especially elevated risk and should be considered for advanced therapies such as mechanical circulatory support or heart transplantation.

CONGESTION AND DIURESIS

The syndrome of heart failure is due primarily to elevation of left ventricular filling pressures resulting in congestion, and therapies aimed at reducing congestion are of primary importance.30 For 50 years, treatment with diuretic medications has been the mainstay of therapy for patients admitted with ADHF. Vasodilator agents, specifically nitroglycerin and sodium nitroprusside, may also be beneficial in patients presenting with ADHF and hypertension.22 In 1 study, patients with acute pulmonary edema treated with high‐dose nitroglycerin experienced fewer adverse events as compared to those treated with high‐dose diuretics alone, suggesting that nitrates can more rapidly decrease congestion and thereby improve outcomes.31 Unfortunately vasodilators are underutilized, with only 5.8% of patients with elevated blood pressure (>160 mmHg) treated with nitrates in the OPTIMIZE‐HF registry.9

Diuretic choice, dosing, and administration method have traditionally been highly variable between practitioners. Oral diuretics are generally not preferred initially for patients with ADHF because of concerns of inadequate absorption from an edematous bowel and slow onset of action.32 For a patient who is not on diuretics as an outpatient, an initial dose of 40 mg intravenous furosemide is reasonable. For a patient with chronic heart failure on outpatient loop diuretic therapy, the Diuretic Optimization Strategies Evaluation (DOSE) study provides insight into diuretic dosing and administration. Patients were randomized to an administration route (bolus dosing every 12 hours or continuous infusion) and a dosing strategy (low‐dose or high‐dose).33 There were no differences in the primary endpoint of patient‐reported global assessment of symptoms, or the primary safety endpoint of change in serum creatinine from baseline to 72 hours between the bolus and continuous infusion groups or between the low‐dose and high‐dose groups. However, patients in the high‐dose group had decreased dyspnea at 72 hours, decreased body weight at 72 hours, increased fluid loss at 72 hours, and decreased NT‐BNP at 72 hours. These improvements came at the expense of a mild increase in creatinine. Therefore, in hospitalized patients with ADHF on outpatient furosemide, these data support initiation of high‐dose furosemide with a daily intravenous dose equal to 2.5 times their daily outpatient oral dose, using either bolus or continuous infusion.

All patients being treated with diuretic therapy should have close fluid intake and output monitoring, fluid restriction of 1500 to 2000 mL per day, a 2 gram sodium diet, and at least daily electrolyte monitoring. For patients with inadequate diuresis (generally less than 1 L per day in a patient with moderate volume overload), several options are available. If the urinary response to a furosemide dose is inadequate, the dose should be doubled and the urinary response followed. If there has been inadequate diuresis in a patient with a low serum albumin or significant proteinuria, furosemide should be switched to bumetanide, which is not protein‐bound and thus will achieve higher concentrations in the tubule.34

Longstanding treatment with loop diuretics leads to decreased renal responsiveness and an increased dose required to maintain euvolemia. Patients taking furosemide 80 mg daily or above (or an equivalent dose of other loop diuretics) are designated as diuretic‐resistant.35, 36 Diuretic resistance is associated with more severe heart failure, more advanced chronic kidney disease, and worsening renal function with the use of intravenous diuretics.35, 37, 38 There are no consensus recommendations available to guide the management of diuretic resistance, but several options exist. First, a thiazide diuretic, such as metolazone, can be given before the loop diuretic.39 This combination is frequently able to initiate a brisk diuresis, but patients require close monitoring for hypokalemia and worsening renal function. Recently, ultrafiltration has emerged as an option. In the Ultrafiltration versus Intravenous Diuretics for Patients Hospitalized for Acute Decompensated Congestive Heart Failure (UNLOAD) trial, patients with congestion treated with ultrafiltration had more weight and fluid loss at 48 hours compared to patients treated with intravenous furosemide, without any significant differences in renal function.40 For patients with oliguria and renal dysfunction, initiation of renal replacement therapy may be needed. We present an algorithm for the management of diuretic resistance in Figure 3.

Figure 3

The endpoints for discontinuation of diuretic therapy remain unclear. Traditionally, alleviation of the patient's congestive symptoms, edema, and attainment of the patient's self‐reported dry weight have served as endpoints for diuretic therapy. A more accurate approach may be daily assessments of the JVP, as normalization of the JVP may be a more accurate method to assess for euvolemia. When euvolemia has been achieved, patients should be switched to maintenance therapy at a diuretic dose of one‐fourth to one‐half the total daily dose used for diuresis. Patients should be observed for 24 hours on oral diuretic therapy to ensure that their fluid intake and output are balanced. Generally, we aim for slightly negative fluid balance (less than 500 mL) on an oral diuretic regimen prior to discharge, assuming some relaxation of the salt and fluid restriction once the patient is discharged home.

NEUROHORMONAL THERAPIES

Activation of neurohormonal systems, specifically the renin‐angiotensin‐aldosterone and beta‐adrenergic pathways, are the major mechanisms for disease progression in HFrEF, and agents which block these pathways improve functional status and survival in these patients. In the OPTIMIZE‐HF registry, patients treated with beta‐blockers on admission had a lower in‐hospital mortality.25 Although beta‐blockers are often discontinued in patients with ADHF, continuation of beta‐blocker treatment is associated with decreased mortality and rehospitalization at 60 to 90 days.41 While beta‐blocker initiation is often deferred to the outpatient setting, patients who receive a beta‐blocker at hospital discharge are 31 times more likely to be treated with a beta‐blocker at 60 to 90 day follow‐up.42 Only 3 agents, metoprolol succinate, carvedilol, and bisoprolol, have survival benefit in large clinical trials of systolic heart failure, and therefore are the only recommended agents.22 In the hospital, hypotension is a common reason for suspension or discontinuation of beta‐blocker therapy. However, in the absence of symptoms such as light‐headedness, patients with systolic blood pressure as low as 85 mmHg will benefit from beta‐blocker treatment.43 Thus, we recommend continuation or initiation of an evidence‐based beta‐blocker for all patients hospitalized with systolic heart failure in the absence of symptomatic hypotension, systolic blood pressure 85 mmHg, second or third degree heart block, or the need for intravenous inotropic therapy.

Inhibitors of the renin‐angiotensin‐aldosterone system also have an important role in patients with HFrEF. Patients treated with angiotensin converting enzyme inhibitors (ACEI) on admission have a lower in‐hospital mortality25 and a lower likelihood of readmission or death within 60 to 90 days.44 In practice, ACEI or angiotensin receptor blocker (ARB) treatment is frequently suspended or discontinued during treatment with diuretics out of concerns for worsening renal function, an association not borne out in trials.38, 45, 46 For patients that are not able to tolerate an indicated therapy, such as a beta‐blocker, ACEI, or ARB, the specific contraindication to treatment should be documented in the medical record.

For patients with HFpEF, no therapy has been shown to improve survival.19, 47 The mainstays of therapy are management of congestion, hypertension, and ventricular rate for patients with atrial fibrillation.22 Research into novel therapies for diastolic heart failure is ongoing.48

DISCHARGE

Patients hospitalized for ADHF are at an increased risk for adverse events following discharge. In an analysis of data from the Candesartan in Heart Failure Assessment of Reduction in Mortality and Morbidity (CHARM) trial, the risk of death was 6‐fold higher in the first month after discharge and remained elevated at 2‐fold higher at 2 years after hospitalization, as compared to persons never hospitalized.49 As yet, no model can accurately predict which ADHF patients will require readmission, though multiple clinical factors have been identified.50 In the OPTIMIZE‐HF registry, increasing admission serum creatinine, a history of chronic obstructive pulmonary disease or cerebrovascular disease, hospitalization for heart failure within the last 6 months, as well as treatment with nitrates, digoxin, diuretics, or mechanical ventilation, were all predictors of mortality and rehospitalization within 60 to 90 days after discharge.44 Furthermore, a BNP level of greater than 350 pg/mL or less than a 50% reduction in NT‐BNP during the hospital stay is also associated with an increased risk for rehospitalization or death.51, 52

Unfortunately, few interventions reduce heart failure readmission rates. In a recent analysis of Medicare claims data, hospitals with the highest rates of early follow‐up after discharge (defined as a clinic visit within 7 days of discharge) had decreased rates of readmission within 30 days.53 Thus, early follow‐up after discharge is essential. Not surprisingly, non‐compliance with weight self‐monitoring leads to increased readmission and mortality rates, and therefore patient education is essential.54 The benefit of home telemonitoring programs remains controversial and requires further study.55, 56 At our center, patients are required to follow up with their internist or cardiologist within 7 days of discharge, and the patient's discharge medication list, discharge weight, and laboratory studies on the day of discharge are faxed to the outpatient provider's office to ensure a seamless transition of care.

PERFORMANCE MEASURES AND GUIDELINES

Performance measures are being assessed with greater frequency in medicine to ensure that clinicians perform key assessments and provide treatments that can improve outcomes. Acute and chronic heart failure were 2 of the first areas to be assessed. In 1996, CMS developed a set of 4 measures for inpatient heart failure care (see Supporting Online Table 2 in the online version of this article).57 Each hospital's performance for these 4 measures is now published at the CMS website. The ACC, AHA, and the American Medical Association's Physician Consortium for Performance Improvement (AMA‐PCPI) released a joint heart failure performance measurement set in 2011. This set removes 3 older recommendations (anticoagulation for patients with atrial fibrillation, discharge instructions, and smoking cessation counseling) and adds 2 new recommendations: prescription of an appropriate beta‐blocker at discharge and arrangement of a postdischarge follow‐up appointment.58, 59 The ACC will publish guidelines based on the ACC/AHA/AMA‐PCPI measure set in early 2012. Of the extant performance measures, both ACEI/ARB and beta‐blocker therapy at discharge are associated with improved outcomes.60, 61

CONCLUSION

With the aging of the population, hospitalizations for ADHF are projected to increase substantially, creating a greater necessity for hospitalists to diagnose, risk stratify, and manage inpatients with heart failure. Once the heart failure diagnosis has been established, determining the etiology of the decompensation and estimating the patient's risk for in‐hospital and postdischarge adverse events is essential. For patients with reduced systolic function, treatment with neurohormonal therapies, even while hospitalized, improves outcomes. Patients should be scheduled for follow‐up within 7 days after discharge to ensure clinical stability. Hospitalists should understand and adhere to the current performance measures for heart failure, as efforts tying payment to the quality of care are likely to evolve.

Caring for patients with acute decompensated heart failure (ADHF) is one of the core competencies of practice in hospitalist medicine. Congestive heart failure remains the most common discharge diagnosis as recorded in the National Hospital Discharge Survey, with over 1.1 million hospitalizations for heart failure in 2004.1 Furthermore, with the disproportionate growth in the population over age 65 that will occur over the next 20 years, heart failure prevalence will grow from its current value of 2.8% to 3.5% by 2030.2 This will result in an additional 3 million Americans with chronic heart failure, thereby sustaining ADHF as the most common reason for hospital admission. Despite an average hospital stay of 5 days, the readmission rate for heart failure was 26.9% at 30 days in a 2003‐2004 analysis of Medicare data.3 This high readmission rate is the target of reform as part of the recently passed Patient Protection and Accountability Act. Starting in fiscal year 2013, acute‐care hospitals with higher‐than‐expected readmission rates for heart failure will have a reduction in reimbursement for these admissions.4 Thus, there is substantial incentive for hospitalists to focus on providing the highest quality of care for patients with ADHF. Here we review the most recent evidence applicable to hospitalists for the diagnosis, risk stratification, and management of patients presenting with ADHF.

DIAGNOSIS

The hospitalist can establish the ADHF diagnosis efficiently by applying a structured approach based on the patient's symptoms, history, physical examination, and laboratory testing. The typical symptoms of ADHF include dyspnea, orthopnea, paroxysmal nocturnal dyspnea (PND), and lower extremity edema. In particular, patients complaining of PND and/or orthopnea are likely to have ADHF.5, 6 Patients may also report chest congestion or chest pain in an atypical pattern. A history of rapid weight gain suggests fluid overload, hence determination of the patient's dry weight is important to establish a target for congestive therapy. Patients with advanced systolic heart failure may also complain of nausea, abdominal pain, and abdominal fullness from ascites.7 In a patient with dyspnea, a history of heart failure, myocardial infarction, or coronary artery disease, all make the diagnosis of ADHF more likely.5

Performing a careful physical examination on a patient presenting with suspected ADHF will not only establish the diagnosis of heart failure, but also determine the hemodynamic profile. Patients presenting with ADHF can be separated into 4 hemodynamic profiles, based on vital sign and physical exam parameters: the presence or absence of congestion (wet or dry), and the presence or absence of adequate perfusion (warm or cold) (Figure 1).8 Parameters indicating the presence of congestion include: orthopnea, elevated jugular venous pulsation (JVP), lower extremity edema, hepatojugular reflux, ascites, and a loud P2 heart sound. Notably, rales are an uncommon physical finding in patients with ADHF, likely because pulmonary lymphatics compensate for chronically elevated filling pressures in such patients.9, 10 Parameters indicating inadequate perfusion include: hypotension (mean arterial pressure 60 mmHg), proportional pulse pressure 25%, cool extremities, altered mental status, and poor urine output (0.5 mL/kg/hr). We recommend assigning the patient to 1 of these 4 hemodynamic profiles, as the profile correlates with invasive hemodynamic measurements of pulmonary capillary wedge pressure and cardiac index, guides management, and predicts outcome.

Figure 1

Natriuretic peptide testing may help establish or exclude a diagnosis of ADHF. A recent expert consensus paper on natriuretic peptide testing recommends cutpoints for both B‐type natriuretic peptide (BNP) and N‐terminal proBNP (NT‐BNP) that indicate a very low (BNP 100 or NT‐BNP 300), intermediate (BNP 100‐400 or NT‐BNP 300‐1800), and high (BNP >400 or NT‐BNP >1800) probability of heart failure11 (Figure 2). However, 2 common conditions affect the utility of BNP testing. First, obese patients have lower levels, and thus a lower rule‐out cutpoint of 54 pg/mL is recommended when using BNP, whereas the cutpoint for NT‐BNP remains the same.12, 13 Second, in patients with renal dysfunction, levels are increased, and thus higher rule‐out cutpoints of 200 pg/mL (for BNP) and 1200 pg/mL (for NT‐BNP) are recommended for patients with a glomerular filtration rate 60 mL/min.14, 15 For patients with longstanding heart failure and chronically elevated levels of natriuretic peptides, there is a correlation between BNP levels and left ventricular filling pressure,16 but the change is more helpful than the absolute levels; a 50% increase over baseline, in conjunction with symptoms, usually reflects ADHF.11

Figure 2

Chest radiography will establish the presence or absence of pulmonary congestion. Classic teaching is that congestion starts with cephalization (pulmonary capillary wedge pressure 10‐15 mmHg), progresses to Kerley B lines (15‐20 mmHg), then to interstitial edema (20‐25 mmHg), and finally to alveolar edema (>25 mmHg).17 In patients presenting with dyspnea, any of these findings helps to establish the diagnosis of ADHF.5

MECHANISMS AND TERMINOLOGY

Data from ADHF registries show that hemodynamically stable patients presenting to the hospital with ADHF are an approximately equal mix of heart failure with reduced ejection fraction (HFrEF; ejection fraction 50%) and heart failure with preserved ejection fraction (HFpEF; ejection fraction 50%).18, 19 The important differences between these groups with regards to pathophysiology and etiology have been reviewed elsewhere.20 Establishing the heart failure mechanism (ie, reduced or preserved EF) is important because the medical management is distinct. Patients with HFrEF are more likely to be male, younger in age, to have ischemic heart disease, and to present with normal or low blood pressure. Patients with HFpEF are more likely to be female, older in age, to have hypertension or diabetes mellitus, and to present with elevated blood pressure.18, 19

The terminology used for inpatient heart failure coding has been the subject of renewed focus. For fiscal year 2008, the Centers for Medicare and Medicaid Services (CMS) overhauled its Diagnosis Related Group (DRG) system to better account for the severity of illness of hospitalized patients.21 In this revision, the existing DRG codes for heart failure were subdivided into 3 severity subclasses: major complication, complication, and non‐complication. Payment to hospitals for a heart failure DRG was changed to be proportional to the level of complication. Thus, for the first time, the clinicians' assessment of the acuity of heart failure determines the level of payment to the hospital. Not surprisingly, this has led to initiatives by hospitals to improve clinicians' coding of inpatients hospitalized with heart failure. A major impediment is that there are no established criteria for the application of each DRG code. Table 1 presents recommended clinical criteria for the application of these codes to patients with ADHF.

PRECIPITANTS AND ETIOLOGY

For patients presenting for the first time with a diagnosis of ADHF (de novo), a thorough evaluation should be performed to determine the mechanism and etiology of the patient's left ventricular dysfunction. After the initial history and physical exam, the American College of Cardiology/American Heart Association (ACC/AHA) guidelines recommend checking basic laboratory studies, an electrocardiogram, and an echocardiogram.22 The full assessment recommended by the ACC/AHA is detailed in Supporting Online Table 1 (in the online version of this article). Cardiac ischemia is the most common etiology of HFrEF, accounting for about 50% of cases. The common, non‐ischemic causes of systolic heart failure include atrial fibrillation, aortic stenosis, illicit cardiotoxic drugs (cocaine, methamphetamine), medical cardiotoxic drugs (adriamycin), as well as primary myocardial disorders such as myocarditis, idiopathic, or peripartum cardiomyopathy. HFpEF is most commonly associated with long‐standing hypertension and diabetes mellitus, but can also be caused by infiltrative, hypertrophic, and constrictive cardiomyopathies.

For patients with a history of heart failure, it is important to identify the precipitant for the decompensation, as it may be treated or avoided in the future. When no clear precipitant is identified, this is most concerning, as it indicates the patient's tenuous cardiac function. In the Organized Program to Initiate Lifesaving Treatment in Hospitalized Patients with Heart Failure (OPTIMIZE‐HF) registry, approximately 61% of patients were found to have at least 1 precipitating factor.23 The most common precipitants were respiratory process in 15.3%, acute coronary syndrome in 14.7%, arrhythmia in 13.5%, uncontrolled hypertension in 10.7%, medication non‐compliance in 8.9%, worsening renal function in 8.0%, and dietary non‐compliance in 5.2%.

RISK STRATIFICATION

Patients hospitalized with ADHF are at a significantly elevated risk for death, both during their hospitalization and after discharge. Numerous studies have shown that multiple clinical parameters assessed during the hospitalization, such as vital signs and laboratory values, predict outcome.6, 8, 24, 25 Some of the most elegant parameters are physical exam findings. As introduced above, the wet‐cold hemodynamic profile assessed at admission predicts increased mortality and urgent transplantation at 1 year.8 One of the most powerful risk stratification schemes for in‐hospital mortality is that developed from the Acute Decompensated Heart Failure (ADHERE) national registry. Three clinical parameters, blood urea nitrogen (BUN) >43 mg/dL, systolic blood pressure 115 mmHg, and serum creatinine >2.75 mg/dL, stratified patients into risk groups. Patients exhibiting all 3 parameters had a 22% in‐hospital mortality compared with 2% for patients with none of the 3 parameters.24

BNP and troponin also have a role in risk stratification of patients with ADHF. In the ADHERE registry, for every increase in the BNP of 400 pg/mL, the odds of risk‐adjusted mortality increased by 9%, in patients with both HFrEF and HFpEF.26 Similarly, an elevated admission troponin was associated with an in‐hospital mortality of 8.0%, versus 2.7% for troponin‐negative patients27; notably almost half of patients with a positive troponin had no history of ischemic heart disease. In the future, refinement and widespread application of these risk stratification methods should allow clinicians to triage patients to determine their location (eg, observation unit, inpatient, intensive care unit) and type of treatment (eg, oral or intravenous diuretic, vasodilator, inotrope).28

In the community, hospitalists care for many patients with ADHF without input from a cardiologist.29 However, there are several situations where the patient is at an increased risk of adverse outcomes, and therefore in which we recommend consulting a cardiologist (Table 2). Patients with hypotension, a cold hemodynamic profile, or worsening renal function due to poor cardiac function are at an especially elevated risk and should be considered for advanced therapies such as mechanical circulatory support or heart transplantation.

CONGESTION AND DIURESIS

The syndrome of heart failure is due primarily to elevation of left ventricular filling pressures resulting in congestion, and therapies aimed at reducing congestion are of primary importance.30 For 50 years, treatment with diuretic medications has been the mainstay of therapy for patients admitted with ADHF. Vasodilator agents, specifically nitroglycerin and sodium nitroprusside, may also be beneficial in patients presenting with ADHF and hypertension.22 In 1 study, patients with acute pulmonary edema treated with high‐dose nitroglycerin experienced fewer adverse events as compared to those treated with high‐dose diuretics alone, suggesting that nitrates can more rapidly decrease congestion and thereby improve outcomes.31 Unfortunately vasodilators are underutilized, with only 5.8% of patients with elevated blood pressure (>160 mmHg) treated with nitrates in the OPTIMIZE‐HF registry.9

Diuretic choice, dosing, and administration method have traditionally been highly variable between practitioners. Oral diuretics are generally not preferred initially for patients with ADHF because of concerns of inadequate absorption from an edematous bowel and slow onset of action.32 For a patient who is not on diuretics as an outpatient, an initial dose of 40 mg intravenous furosemide is reasonable. For a patient with chronic heart failure on outpatient loop diuretic therapy, the Diuretic Optimization Strategies Evaluation (DOSE) study provides insight into diuretic dosing and administration. Patients were randomized to an administration route (bolus dosing every 12 hours or continuous infusion) and a dosing strategy (low‐dose or high‐dose).33 There were no differences in the primary endpoint of patient‐reported global assessment of symptoms, or the primary safety endpoint of change in serum creatinine from baseline to 72 hours between the bolus and continuous infusion groups or between the low‐dose and high‐dose groups. However, patients in the high‐dose group had decreased dyspnea at 72 hours, decreased body weight at 72 hours, increased fluid loss at 72 hours, and decreased NT‐BNP at 72 hours. These improvements came at the expense of a mild increase in creatinine. Therefore, in hospitalized patients with ADHF on outpatient furosemide, these data support initiation of high‐dose furosemide with a daily intravenous dose equal to 2.5 times their daily outpatient oral dose, using either bolus or continuous infusion.

All patients being treated with diuretic therapy should have close fluid intake and output monitoring, fluid restriction of 1500 to 2000 mL per day, a 2 gram sodium diet, and at least daily electrolyte monitoring. For patients with inadequate diuresis (generally less than 1 L per day in a patient with moderate volume overload), several options are available. If the urinary response to a furosemide dose is inadequate, the dose should be doubled and the urinary response followed. If there has been inadequate diuresis in a patient with a low serum albumin or significant proteinuria, furosemide should be switched to bumetanide, which is not protein‐bound and thus will achieve higher concentrations in the tubule.34

Longstanding treatment with loop diuretics leads to decreased renal responsiveness and an increased dose required to maintain euvolemia. Patients taking furosemide 80 mg daily or above (or an equivalent dose of other loop diuretics) are designated as diuretic‐resistant.35, 36 Diuretic resistance is associated with more severe heart failure, more advanced chronic kidney disease, and worsening renal function with the use of intravenous diuretics.35, 37, 38 There are no consensus recommendations available to guide the management of diuretic resistance, but several options exist. First, a thiazide diuretic, such as metolazone, can be given before the loop diuretic.39 This combination is frequently able to initiate a brisk diuresis, but patients require close monitoring for hypokalemia and worsening renal function. Recently, ultrafiltration has emerged as an option. In the Ultrafiltration versus Intravenous Diuretics for Patients Hospitalized for Acute Decompensated Congestive Heart Failure (UNLOAD) trial, patients with congestion treated with ultrafiltration had more weight and fluid loss at 48 hours compared to patients treated with intravenous furosemide, without any significant differences in renal function.40 For patients with oliguria and renal dysfunction, initiation of renal replacement therapy may be needed. We present an algorithm for the management of diuretic resistance in Figure 3.

Figure 3

The endpoints for discontinuation of diuretic therapy remain unclear. Traditionally, alleviation of the patient's congestive symptoms, edema, and attainment of the patient's self‐reported dry weight have served as endpoints for diuretic therapy. A more accurate approach may be daily assessments of the JVP, as normalization of the JVP may be a more accurate method to assess for euvolemia. When euvolemia has been achieved, patients should be switched to maintenance therapy at a diuretic dose of one‐fourth to one‐half the total daily dose used for diuresis. Patients should be observed for 24 hours on oral diuretic therapy to ensure that their fluid intake and output are balanced. Generally, we aim for slightly negative fluid balance (less than 500 mL) on an oral diuretic regimen prior to discharge, assuming some relaxation of the salt and fluid restriction once the patient is discharged home.

NEUROHORMONAL THERAPIES

Activation of neurohormonal systems, specifically the renin‐angiotensin‐aldosterone and beta‐adrenergic pathways, are the major mechanisms for disease progression in HFrEF, and agents which block these pathways improve functional status and survival in these patients. In the OPTIMIZE‐HF registry, patients treated with beta‐blockers on admission had a lower in‐hospital mortality.25 Although beta‐blockers are often discontinued in patients with ADHF, continuation of beta‐blocker treatment is associated with decreased mortality and rehospitalization at 60 to 90 days.41 While beta‐blocker initiation is often deferred to the outpatient setting, patients who receive a beta‐blocker at hospital discharge are 31 times more likely to be treated with a beta‐blocker at 60 to 90 day follow‐up.42 Only 3 agents, metoprolol succinate, carvedilol, and bisoprolol, have survival benefit in large clinical trials of systolic heart failure, and therefore are the only recommended agents.22 In the hospital, hypotension is a common reason for suspension or discontinuation of beta‐blocker therapy. However, in the absence of symptoms such as light‐headedness, patients with systolic blood pressure as low as 85 mmHg will benefit from beta‐blocker treatment.43 Thus, we recommend continuation or initiation of an evidence‐based beta‐blocker for all patients hospitalized with systolic heart failure in the absence of symptomatic hypotension, systolic blood pressure 85 mmHg, second or third degree heart block, or the need for intravenous inotropic therapy.

Inhibitors of the renin‐angiotensin‐aldosterone system also have an important role in patients with HFrEF. Patients treated with angiotensin converting enzyme inhibitors (ACEI) on admission have a lower in‐hospital mortality25 and a lower likelihood of readmission or death within 60 to 90 days.44 In practice, ACEI or angiotensin receptor blocker (ARB) treatment is frequently suspended or discontinued during treatment with diuretics out of concerns for worsening renal function, an association not borne out in trials.38, 45, 46 For patients that are not able to tolerate an indicated therapy, such as a beta‐blocker, ACEI, or ARB, the specific contraindication to treatment should be documented in the medical record.

For patients with HFpEF, no therapy has been shown to improve survival.19, 47 The mainstays of therapy are management of congestion, hypertension, and ventricular rate for patients with atrial fibrillation.22 Research into novel therapies for diastolic heart failure is ongoing.48

DISCHARGE

Patients hospitalized for ADHF are at an increased risk for adverse events following discharge. In an analysis of data from the Candesartan in Heart Failure Assessment of Reduction in Mortality and Morbidity (CHARM) trial, the risk of death was 6‐fold higher in the first month after discharge and remained elevated at 2‐fold higher at 2 years after hospitalization, as compared to persons never hospitalized.49 As yet, no model can accurately predict which ADHF patients will require readmission, though multiple clinical factors have been identified.50 In the OPTIMIZE‐HF registry, increasing admission serum creatinine, a history of chronic obstructive pulmonary disease or cerebrovascular disease, hospitalization for heart failure within the last 6 months, as well as treatment with nitrates, digoxin, diuretics, or mechanical ventilation, were all predictors of mortality and rehospitalization within 60 to 90 days after discharge.44 Furthermore, a BNP level of greater than 350 pg/mL or less than a 50% reduction in NT‐BNP during the hospital stay is also associated with an increased risk for rehospitalization or death.51, 52

Unfortunately, few interventions reduce heart failure readmission rates. In a recent analysis of Medicare claims data, hospitals with the highest rates of early follow‐up after discharge (defined as a clinic visit within 7 days of discharge) had decreased rates of readmission within 30 days.53 Thus, early follow‐up after discharge is essential. Not surprisingly, non‐compliance with weight self‐monitoring leads to increased readmission and mortality rates, and therefore patient education is essential.54 The benefit of home telemonitoring programs remains controversial and requires further study.55, 56 At our center, patients are required to follow up with their internist or cardiologist within 7 days of discharge, and the patient's discharge medication list, discharge weight, and laboratory studies on the day of discharge are faxed to the outpatient provider's office to ensure a seamless transition of care.

PERFORMANCE MEASURES AND GUIDELINES

Performance measures are being assessed with greater frequency in medicine to ensure that clinicians perform key assessments and provide treatments that can improve outcomes. Acute and chronic heart failure were 2 of the first areas to be assessed. In 1996, CMS developed a set of 4 measures for inpatient heart failure care (see Supporting Online Table 2 in the online version of this article).57 Each hospital's performance for these 4 measures is now published at the CMS website. The ACC, AHA, and the American Medical Association's Physician Consortium for Performance Improvement (AMA‐PCPI) released a joint heart failure performance measurement set in 2011. This set removes 3 older recommendations (anticoagulation for patients with atrial fibrillation, discharge instructions, and smoking cessation counseling) and adds 2 new recommendations: prescription of an appropriate beta‐blocker at discharge and arrangement of a postdischarge follow‐up appointment.58, 59 The ACC will publish guidelines based on the ACC/AHA/AMA‐PCPI measure set in early 2012. Of the extant performance measures, both ACEI/ARB and beta‐blocker therapy at discharge are associated with improved outcomes.60, 61

CONCLUSION

With the aging of the population, hospitalizations for ADHF are projected to increase substantially, creating a greater necessity for hospitalists to diagnose, risk stratify, and manage inpatients with heart failure. Once the heart failure diagnosis has been established, determining the etiology of the decompensation and estimating the patient's risk for in‐hospital and postdischarge adverse events is essential. For patients with reduced systolic function, treatment with neurohormonal therapies, even while hospitalized, improves outcomes. Patients should be scheduled for follow‐up within 7 days after discharge to ensure clinical stability. Hospitalists should understand and adhere to the current performance measures for heart failure, as efforts tying payment to the quality of care are likely to evolve.

Caring for patients with acute decompensated heart failure (ADHF) is one of the core competencies of practice in hospitalist medicine. Congestive heart failure remains the most common discharge diagnosis as recorded in the National Hospital Discharge Survey, with over 1.1 million hospitalizations for heart failure in 2004.1 Furthermore, with the disproportionate growth in the population over age 65 that will occur over the next 20 years, heart failure prevalence will grow from its current value of 2.8% to 3.5% by 2030.2 This will result in an additional 3 million Americans with chronic heart failure, thereby sustaining ADHF as the most common reason for hospital admission. Despite an average hospital stay of 5 days, the readmission rate for heart failure was 26.9% at 30 days in a 2003‐2004 analysis of Medicare data.3 This high readmission rate is the target of reform as part of the recently passed Patient Protection and Accountability Act. Starting in fiscal year 2013, acute‐care hospitals with higher‐than‐expected readmission rates for heart failure will have a reduction in reimbursement for these admissions.4 Thus, there is substantial incentive for hospitalists to focus on providing the highest quality of care for patients with ADHF. Here we review the most recent evidence applicable to hospitalists for the diagnosis, risk stratification, and management of patients presenting with ADHF.

DIAGNOSIS

The hospitalist can establish the ADHF diagnosis efficiently by applying a structured approach based on the patient's symptoms, history, physical examination, and laboratory testing. The typical symptoms of ADHF include dyspnea, orthopnea, paroxysmal nocturnal dyspnea (PND), and lower extremity edema. In particular, patients complaining of PND and/or orthopnea are likely to have ADHF.5, 6 Patients may also report chest congestion or chest pain in an atypical pattern. A history of rapid weight gain suggests fluid overload, hence determination of the patient's dry weight is important to establish a target for congestive therapy. Patients with advanced systolic heart failure may also complain of nausea, abdominal pain, and abdominal fullness from ascites.7 In a patient with dyspnea, a history of heart failure, myocardial infarction, or coronary artery disease, all make the diagnosis of ADHF more likely.5

Performing a careful physical examination on a patient presenting with suspected ADHF will not only establish the diagnosis of heart failure, but also determine the hemodynamic profile. Patients presenting with ADHF can be separated into 4 hemodynamic profiles, based on vital sign and physical exam parameters: the presence or absence of congestion (wet or dry), and the presence or absence of adequate perfusion (warm or cold) (Figure 1).8 Parameters indicating the presence of congestion include: orthopnea, elevated jugular venous pulsation (JVP), lower extremity edema, hepatojugular reflux, ascites, and a loud P2 heart sound. Notably, rales are an uncommon physical finding in patients with ADHF, likely because pulmonary lymphatics compensate for chronically elevated filling pressures in such patients.9, 10 Parameters indicating inadequate perfusion include: hypotension (mean arterial pressure 60 mmHg), proportional pulse pressure 25%, cool extremities, altered mental status, and poor urine output (0.5 mL/kg/hr). We recommend assigning the patient to 1 of these 4 hemodynamic profiles, as the profile correlates with invasive hemodynamic measurements of pulmonary capillary wedge pressure and cardiac index, guides management, and predicts outcome.

Figure 1

Natriuretic peptide testing may help establish or exclude a diagnosis of ADHF. A recent expert consensus paper on natriuretic peptide testing recommends cutpoints for both B‐type natriuretic peptide (BNP) and N‐terminal proBNP (NT‐BNP) that indicate a very low (BNP 100 or NT‐BNP 300), intermediate (BNP 100‐400 or NT‐BNP 300‐1800), and high (BNP >400 or NT‐BNP >1800) probability of heart failure11 (Figure 2). However, 2 common conditions affect the utility of BNP testing. First, obese patients have lower levels, and thus a lower rule‐out cutpoint of 54 pg/mL is recommended when using BNP, whereas the cutpoint for NT‐BNP remains the same.12, 13 Second, in patients with renal dysfunction, levels are increased, and thus higher rule‐out cutpoints of 200 pg/mL (for BNP) and 1200 pg/mL (for NT‐BNP) are recommended for patients with a glomerular filtration rate 60 mL/min.14, 15 For patients with longstanding heart failure and chronically elevated levels of natriuretic peptides, there is a correlation between BNP levels and left ventricular filling pressure,16 but the change is more helpful than the absolute levels; a 50% increase over baseline, in conjunction with symptoms, usually reflects ADHF.11

Figure 2

Chest radiography will establish the presence or absence of pulmonary congestion. Classic teaching is that congestion starts with cephalization (pulmonary capillary wedge pressure 10‐15 mmHg), progresses to Kerley B lines (15‐20 mmHg), then to interstitial edema (20‐25 mmHg), and finally to alveolar edema (>25 mmHg).17 In patients presenting with dyspnea, any of these findings helps to establish the diagnosis of ADHF.5

MECHANISMS AND TERMINOLOGY

Data from ADHF registries show that hemodynamically stable patients presenting to the hospital with ADHF are an approximately equal mix of heart failure with reduced ejection fraction (HFrEF; ejection fraction 50%) and heart failure with preserved ejection fraction (HFpEF; ejection fraction 50%).18, 19 The important differences between these groups with regards to pathophysiology and etiology have been reviewed elsewhere.20 Establishing the heart failure mechanism (ie, reduced or preserved EF) is important because the medical management is distinct. Patients with HFrEF are more likely to be male, younger in age, to have ischemic heart disease, and to present with normal or low blood pressure. Patients with HFpEF are more likely to be female, older in age, to have hypertension or diabetes mellitus, and to present with elevated blood pressure.18, 19

The terminology used for inpatient heart failure coding has been the subject of renewed focus. For fiscal year 2008, the Centers for Medicare and Medicaid Services (CMS) overhauled its Diagnosis Related Group (DRG) system to better account for the severity of illness of hospitalized patients.21 In this revision, the existing DRG codes for heart failure were subdivided into 3 severity subclasses: major complication, complication, and non‐complication. Payment to hospitals for a heart failure DRG was changed to be proportional to the level of complication. Thus, for the first time, the clinicians' assessment of the acuity of heart failure determines the level of payment to the hospital. Not surprisingly, this has led to initiatives by hospitals to improve clinicians' coding of inpatients hospitalized with heart failure. A major impediment is that there are no established criteria for the application of each DRG code. Table 1 presents recommended clinical criteria for the application of these codes to patients with ADHF.

PRECIPITANTS AND ETIOLOGY

For patients presenting for the first time with a diagnosis of ADHF (de novo), a thorough evaluation should be performed to determine the mechanism and etiology of the patient's left ventricular dysfunction. After the initial history and physical exam, the American College of Cardiology/American Heart Association (ACC/AHA) guidelines recommend checking basic laboratory studies, an electrocardiogram, and an echocardiogram.22 The full assessment recommended by the ACC/AHA is detailed in Supporting Online Table 1 (in the online version of this article). Cardiac ischemia is the most common etiology of HFrEF, accounting for about 50% of cases. The common, non‐ischemic causes of systolic heart failure include atrial fibrillation, aortic stenosis, illicit cardiotoxic drugs (cocaine, methamphetamine), medical cardiotoxic drugs (adriamycin), as well as primary myocardial disorders such as myocarditis, idiopathic, or peripartum cardiomyopathy. HFpEF is most commonly associated with long‐standing hypertension and diabetes mellitus, but can also be caused by infiltrative, hypertrophic, and constrictive cardiomyopathies.

For patients with a history of heart failure, it is important to identify the precipitant for the decompensation, as it may be treated or avoided in the future. When no clear precipitant is identified, this is most concerning, as it indicates the patient's tenuous cardiac function. In the Organized Program to Initiate Lifesaving Treatment in Hospitalized Patients with Heart Failure (OPTIMIZE‐HF) registry, approximately 61% of patients were found to have at least 1 precipitating factor.23 The most common precipitants were respiratory process in 15.3%, acute coronary syndrome in 14.7%, arrhythmia in 13.5%, uncontrolled hypertension in 10.7%, medication non‐compliance in 8.9%, worsening renal function in 8.0%, and dietary non‐compliance in 5.2%.

RISK STRATIFICATION

Patients hospitalized with ADHF are at a significantly elevated risk for death, both during their hospitalization and after discharge. Numerous studies have shown that multiple clinical parameters assessed during the hospitalization, such as vital signs and laboratory values, predict outcome.6, 8, 24, 25 Some of the most elegant parameters are physical exam findings. As introduced above, the wet‐cold hemodynamic profile assessed at admission predicts increased mortality and urgent transplantation at 1 year.8 One of the most powerful risk stratification schemes for in‐hospital mortality is that developed from the Acute Decompensated Heart Failure (ADHERE) national registry. Three clinical parameters, blood urea nitrogen (BUN) >43 mg/dL, systolic blood pressure 115 mmHg, and serum creatinine >2.75 mg/dL, stratified patients into risk groups. Patients exhibiting all 3 parameters had a 22% in‐hospital mortality compared with 2% for patients with none of the 3 parameters.24

BNP and troponin also have a role in risk stratification of patients with ADHF. In the ADHERE registry, for every increase in the BNP of 400 pg/mL, the odds of risk‐adjusted mortality increased by 9%, in patients with both HFrEF and HFpEF.26 Similarly, an elevated admission troponin was associated with an in‐hospital mortality of 8.0%, versus 2.7% for troponin‐negative patients27; notably almost half of patients with a positive troponin had no history of ischemic heart disease. In the future, refinement and widespread application of these risk stratification methods should allow clinicians to triage patients to determine their location (eg, observation unit, inpatient, intensive care unit) and type of treatment (eg, oral or intravenous diuretic, vasodilator, inotrope).28

In the community, hospitalists care for many patients with ADHF without input from a cardiologist.29 However, there are several situations where the patient is at an increased risk of adverse outcomes, and therefore in which we recommend consulting a cardiologist (Table 2). Patients with hypotension, a cold hemodynamic profile, or worsening renal function due to poor cardiac function are at an especially elevated risk and should be considered for advanced therapies such as mechanical circulatory support or heart transplantation.

CONGESTION AND DIURESIS

The syndrome of heart failure is due primarily to elevation of left ventricular filling pressures resulting in congestion, and therapies aimed at reducing congestion are of primary importance.30 For 50 years, treatment with diuretic medications has been the mainstay of therapy for patients admitted with ADHF. Vasodilator agents, specifically nitroglycerin and sodium nitroprusside, may also be beneficial in patients presenting with ADHF and hypertension.22 In 1 study, patients with acute pulmonary edema treated with high‐dose nitroglycerin experienced fewer adverse events as compared to those treated with high‐dose diuretics alone, suggesting that nitrates can more rapidly decrease congestion and thereby improve outcomes.31 Unfortunately vasodilators are underutilized, with only 5.8% of patients with elevated blood pressure (>160 mmHg) treated with nitrates in the OPTIMIZE‐HF registry.9

Diuretic choice, dosing, and administration method have traditionally been highly variable between practitioners. Oral diuretics are generally not preferred initially for patients with ADHF because of concerns of inadequate absorption from an edematous bowel and slow onset of action.32 For a patient who is not on diuretics as an outpatient, an initial dose of 40 mg intravenous furosemide is reasonable. For a patient with chronic heart failure on outpatient loop diuretic therapy, the Diuretic Optimization Strategies Evaluation (DOSE) study provides insight into diuretic dosing and administration. Patients were randomized to an administration route (bolus dosing every 12 hours or continuous infusion) and a dosing strategy (low‐dose or high‐dose).33 There were no differences in the primary endpoint of patient‐reported global assessment of symptoms, or the primary safety endpoint of change in serum creatinine from baseline to 72 hours between the bolus and continuous infusion groups or between the low‐dose and high‐dose groups. However, patients in the high‐dose group had decreased dyspnea at 72 hours, decreased body weight at 72 hours, increased fluid loss at 72 hours, and decreased NT‐BNP at 72 hours. These improvements came at the expense of a mild increase in creatinine. Therefore, in hospitalized patients with ADHF on outpatient furosemide, these data support initiation of high‐dose furosemide with a daily intravenous dose equal to 2.5 times their daily outpatient oral dose, using either bolus or continuous infusion.

All patients being treated with diuretic therapy should have close fluid intake and output monitoring, fluid restriction of 1500 to 2000 mL per day, a 2 gram sodium diet, and at least daily electrolyte monitoring. For patients with inadequate diuresis (generally less than 1 L per day in a patient with moderate volume overload), several options are available. If the urinary response to a furosemide dose is inadequate, the dose should be doubled and the urinary response followed. If there has been inadequate diuresis in a patient with a low serum albumin or significant proteinuria, furosemide should be switched to bumetanide, which is not protein‐bound and thus will achieve higher concentrations in the tubule.34

Longstanding treatment with loop diuretics leads to decreased renal responsiveness and an increased dose required to maintain euvolemia. Patients taking furosemide 80 mg daily or above (or an equivalent dose of other loop diuretics) are designated as diuretic‐resistant.35, 36 Diuretic resistance is associated with more severe heart failure, more advanced chronic kidney disease, and worsening renal function with the use of intravenous diuretics.35, 37, 38 There are no consensus recommendations available to guide the management of diuretic resistance, but several options exist. First, a thiazide diuretic, such as metolazone, can be given before the loop diuretic.39 This combination is frequently able to initiate a brisk diuresis, but patients require close monitoring for hypokalemia and worsening renal function. Recently, ultrafiltration has emerged as an option. In the Ultrafiltration versus Intravenous Diuretics for Patients Hospitalized for Acute Decompensated Congestive Heart Failure (UNLOAD) trial, patients with congestion treated with ultrafiltration had more weight and fluid loss at 48 hours compared to patients treated with intravenous furosemide, without any significant differences in renal function.40 For patients with oliguria and renal dysfunction, initiation of renal replacement therapy may be needed. We present an algorithm for the management of diuretic resistance in Figure 3.

Figure 3

The endpoints for discontinuation of diuretic therapy remain unclear. Traditionally, alleviation of the patient's congestive symptoms, edema, and attainment of the patient's self‐reported dry weight have served as endpoints for diuretic therapy. A more accurate approach may be daily assessments of the JVP, as normalization of the JVP may be a more accurate method to assess for euvolemia. When euvolemia has been achieved, patients should be switched to maintenance therapy at a diuretic dose of one‐fourth to one‐half the total daily dose used for diuresis. Patients should be observed for 24 hours on oral diuretic therapy to ensure that their fluid intake and output are balanced. Generally, we aim for slightly negative fluid balance (less than 500 mL) on an oral diuretic regimen prior to discharge, assuming some relaxation of the salt and fluid restriction once the patient is discharged home.

NEUROHORMONAL THERAPIES

Activation of neurohormonal systems, specifically the renin‐angiotensin‐aldosterone and beta‐adrenergic pathways, are the major mechanisms for disease progression in HFrEF, and agents which block these pathways improve functional status and survival in these patients. In the OPTIMIZE‐HF registry, patients treated with beta‐blockers on admission had a lower in‐hospital mortality.25 Although beta‐blockers are often discontinued in patients with ADHF, continuation of beta‐blocker treatment is associated with decreased mortality and rehospitalization at 60 to 90 days.41 While beta‐blocker initiation is often deferred to the outpatient setting, patients who receive a beta‐blocker at hospital discharge are 31 times more likely to be treated with a beta‐blocker at 60 to 90 day follow‐up.42 Only 3 agents, metoprolol succinate, carvedilol, and bisoprolol, have survival benefit in large clinical trials of systolic heart failure, and therefore are the only recommended agents.22 In the hospital, hypotension is a common reason for suspension or discontinuation of beta‐blocker therapy. However, in the absence of symptoms such as light‐headedness, patients with systolic blood pressure as low as 85 mmHg will benefit from beta‐blocker treatment.43 Thus, we recommend continuation or initiation of an evidence‐based beta‐blocker for all patients hospitalized with systolic heart failure in the absence of symptomatic hypotension, systolic blood pressure 85 mmHg, second or third degree heart block, or the need for intravenous inotropic therapy.

Inhibitors of the renin‐angiotensin‐aldosterone system also have an important role in patients with HFrEF. Patients treated with angiotensin converting enzyme inhibitors (ACEI) on admission have a lower in‐hospital mortality25 and a lower likelihood of readmission or death within 60 to 90 days.44 In practice, ACEI or angiotensin receptor blocker (ARB) treatment is frequently suspended or discontinued during treatment with diuretics out of concerns for worsening renal function, an association not borne out in trials.38, 45, 46 For patients that are not able to tolerate an indicated therapy, such as a beta‐blocker, ACEI, or ARB, the specific contraindication to treatment should be documented in the medical record.

For patients with HFpEF, no therapy has been shown to improve survival.19, 47 The mainstays of therapy are management of congestion, hypertension, and ventricular rate for patients with atrial fibrillation.22 Research into novel therapies for diastolic heart failure is ongoing.48

DISCHARGE

Patients hospitalized for ADHF are at an increased risk for adverse events following discharge. In an analysis of data from the Candesartan in Heart Failure Assessment of Reduction in Mortality and Morbidity (CHARM) trial, the risk of death was 6‐fold higher in the first month after discharge and remained elevated at 2‐fold higher at 2 years after hospitalization, as compared to persons never hospitalized.49 As yet, no model can accurately predict which ADHF patients will require readmission, though multiple clinical factors have been identified.50 In the OPTIMIZE‐HF registry, increasing admission serum creatinine, a history of chronic obstructive pulmonary disease or cerebrovascular disease, hospitalization for heart failure within the last 6 months, as well as treatment with nitrates, digoxin, diuretics, or mechanical ventilation, were all predictors of mortality and rehospitalization within 60 to 90 days after discharge.44 Furthermore, a BNP level of greater than 350 pg/mL or less than a 50% reduction in NT‐BNP during the hospital stay is also associated with an increased risk for rehospitalization or death.51, 52

Unfortunately, few interventions reduce heart failure readmission rates. In a recent analysis of Medicare claims data, hospitals with the highest rates of early follow‐up after discharge (defined as a clinic visit within 7 days of discharge) had decreased rates of readmission within 30 days.53 Thus, early follow‐up after discharge is essential. Not surprisingly, non‐compliance with weight self‐monitoring leads to increased readmission and mortality rates, and therefore patient education is essential.54 The benefit of home telemonitoring programs remains controversial and requires further study.55, 56 At our center, patients are required to follow up with their internist or cardiologist within 7 days of discharge, and the patient's discharge medication list, discharge weight, and laboratory studies on the day of discharge are faxed to the outpatient provider's office to ensure a seamless transition of care.

PERFORMANCE MEASURES AND GUIDELINES

Performance measures are being assessed with greater frequency in medicine to ensure that clinicians perform key assessments and provide treatments that can improve outcomes. Acute and chronic heart failure were 2 of the first areas to be assessed. In 1996, CMS developed a set of 4 measures for inpatient heart failure care (see Supporting Online Table 2 in the online version of this article).57 Each hospital's performance for these 4 measures is now published at the CMS website. The ACC, AHA, and the American Medical Association's Physician Consortium for Performance Improvement (AMA‐PCPI) released a joint heart failure performance measurement set in 2011. This set removes 3 older recommendations (anticoagulation for patients with atrial fibrillation, discharge instructions, and smoking cessation counseling) and adds 2 new recommendations: prescription of an appropriate beta‐blocker at discharge and arrangement of a postdischarge follow‐up appointment.58, 59 The ACC will publish guidelines based on the ACC/AHA/AMA‐PCPI measure set in early 2012. Of the extant performance measures, both ACEI/ARB and beta‐blocker therapy at discharge are associated with improved outcomes.60, 61

CONCLUSION

With the aging of the population, hospitalizations for ADHF are projected to increase substantially, creating a greater necessity for hospitalists to diagnose, risk stratify, and manage inpatients with heart failure. Once the heart failure diagnosis has been established, determining the etiology of the decompensation and estimating the patient's risk for in‐hospital and postdischarge adverse events is essential. For patients with reduced systolic function, treatment with neurohormonal therapies, even while hospitalized, improves outcomes. Patients should be scheduled for follow‐up within 7 days after discharge to ensure clinical stability. Hospitalists should understand and adhere to the current performance measures for heart failure, as efforts tying payment to the quality of care are likely to evolve.