Acute coronary syndromes (ACS) are potentially lethal and present with a wide variety of symptoms. As such, physicians frequently order cardiac biomarkers, such as cardiac troponin, for patients with acute complaints. Elevated troponin is associated with higher risk of mortality regardless of the causes, which can be myriad, both chronic and acute.[1] Among patients with an elevated troponin, distinguishing ACS from non‐ACS can be challenging.

Making the distinction between ACS and non‐ACS troponin elevation is crucial because the underlying pathophysiology and subsequent management strategies are markedly different.[2] According to evidence‐based practice guidelines, ACS is managed with antiplatelet drugs, statins, and percutaneous coronary intervention, improving clinical outcomes.[3] In contrast, care for patients with non‐ACS troponin elevations is usually supportive, with a focus on the underlying conditions. The lack of specific treatment options for such patients is concerning given that several series have suggested that non‐ACS troponin patients may have a higher mortality risk than ACS patients.[4, 5, 6] Non‐ACS troponin elevation can be the result of a multitude of conditions.[7, 8] What remains unclear at this point is whether the excess mortality observed with non‐ACS troponin elevation is due to myocardial damage or to the underlying conditions that predispose to troponin release.

Using data from a quality improvement (QI) project collected at our Veterans Affairs (VA) medical center, we investigated the mortality risk associated with ACS and non‐ACS troponin elevation including an analysis of factors associated with mortality. We hypothesized that non‐ACS troponin elevation will have a higher mortality risk than troponin elevation due to ACS, and that important contributors to this relationship could be identified to provide direction for future investigation directed at modifying this mortality risk.

METHODS

We analyzed data that were prospectively collected for a quality initiative between 2006 and 2007. The project was a collaborative endeavor between cardiology, hospital medicine, and emergency medicine with the process goal of better identifying patients with ACS to hopefully improve outcomes. The QI team was consulted in real time to assist with treatment recommendations; no retrospective decisions were made regarding whether or not ACS was present. As the goal of the project was to improve cardiovascular outcomes, consultative advice was freely provided, and no physicians or teams were subject to any adverse repercussions for their diagnoses or management decisions.

A cardiologist‐led team was created to improve quality of care for myocardial infarction patients by evaluating all patients at our facility with an elevated troponin. On a daily basis, a specialist clinical coordinator (nurse practitioner or physician assistant) received a list of all patients with elevated troponin from the chemistry lab. The coordinator reviewed the patients' medical records with a cardiologist. A positive troponin was defined as a troponin T level of greater than 0.03 ng/mL (99th percentile at our facility). Each attending cardiologist prospectively determined if troponin elevation was related to clinical findings consistent with an ACS based on review of the patients' symptoms (duration, quality, severity, chronicity, and alleviating/aggravating factors), medical history, and noninvasive cardiac testing including electrocardiograms, cardiac biomarkers, and any other available imaging tests.

We have previously demonstrated that the cardiologists at our facility have a similar rate of diagnosing ACS.[9] All cardiologists at our facility maintain current American Board of Internal Medicine certification in cardiovascular disease and have academic appointments at the University of Florida College of Medicine. All patients were followed prospectively, and data on their medical history, acute evaluation, and outcomes were tracked in an electronic database. Given the higher risk of mortality with ST‐elevation myocardial infarction, such patients were excluded from this investigation. By definition, patients with unstable angina do not have elevated biomarkers and thus would not have been included in the database to begin with. Prospectively recorded data elements included: age, gender, chief complaint, tobacco use, presence of hypertension, hyperlipidemia, prior coronary disease, chronic kidney disease, diabetes mellitus, cardiac troponin values, serum creatinine, electrocardiogram (ECG) variables, Thrombolysis in Myocardial Infarction (TIMI) score, and if the patient was placed under hospice care or an active do‐not‐resuscitate (DNR) order. Additional data elements gathered at a later date included maximum temperature, white blood cell count, N‐terminal pro‐brain natriuretic peptide (NT‐proBNP), administration of advanced cardiac life support (ACLS), and admission to an intensive care unit (ICU). All consecutive patients with elevated troponin were included in the database; if patients were included more than once, we used their index evaluation only. All patients with troponin elevation after revascularization (percutaneous coronary intervention or coronary bypass surgery) were excluded. Our investigational design was reviewed by our institutional review board, who waived the requirement for formal written informed consent and approved use of data from this QI project for research purposes.

We focused this investigation on an analysis of all‐cause mortality in February 2014. We analyzed mortality at 30 days, 1 year, and 6 years. As secondary outcomes we analyzed the likelihood of the patients' chief complaint for the diagnosis of ACS and evaluated predictors of mortality based on Cox proportional hazard modeling. Mortality within the VA system is reliably tracked and compares favorably to the Social Security National Death Index Master File for accuracy.[10, 11]

Categorical variables were compared by ² test. The Student t test was used to compare normally distributed continuous variables, and nonparametric tests were used for non‐normal distributions as appropriate. Mortality data at 30 days, 1 year, and 6 years were compared by log‐rank test and Kaplan‐Meier graphs. A formal power analysis was not performed; the entire available population was included. A Cox proportional hazard model was created to estimate mortality risk at each time point. Variables included in our Cox regression model were age, gender, history of coronary artery disease (CAD), hypertension, diabetes mellitus or hyperlipidemia, ACS diagnosis, dynamic ECG changes, TIMI risk score, initial troponin level, creatinine level at time of initial troponin (per mg/dL), presence of fever, maximum white blood cell count, NT‐proBNP level (per 1000 pg/mL), if ACLS was performed, if the patient was under hospice care, if there was a DNR order, and if they required ICU admission. This model was also constructed independently for the ACS and non‐ACS cohorts for mortality at 1 year. A forward stepwise model was used. Statistical results were considered significant at P < 0.05. Statistical analyses were performed using SPSS version 21 (IBM, Armonk, NY).

RESULTS

Among the 761 patients, 502 (66.0%) were classified as non‐ACS and 259 (34.0%) as ACS (Table 1). The mean age was higher in the non‐ACS group (71 years vs 69 years in the ACS group, P = 0.006). Hypertension, diabetes mellitus, and prior CAD were frequent in both groups and not significantly different. Median initial troponin T was higher in the ACS group (0.12 ng/mL vs 0.06 ng/mL, P < 0.001) as were the frequency of a TIMI risk score >2 (92.5% vs 74.3%, P < 0.001) and new ECG changes (29.7% vs 8.2%, P < 0.001). Hospice, DNR orders, and administration of ACLS were not different between groups; however, admission to the ICU was more frequent in the ACS group (44.8% vs 31.9%, P < 0.001). Chest pain was the symptom with the highest positive predictive value for the diagnosis of ACS (63.3%), whereas the least predictive was altered mental status or confusion (18.0%) (Figure 1).

Figure 1

Mortality at 30 days was not different between the 2 groups, but mortality was higher for the non‐ACS cohort at 1 year and at 6 years (Table 1). Kaplan‐Meier curves demonstrate that mortality for the 2 cohorts begins to diverge between 30 and 60 days until approximately 2 years when the curves again are parallel (Figure 2).

Figure 2

In Cox proportional hazards models, 5 factors were associated with higher mortality at 30 days, 1 year, and at 6 years: age, hospice, DNR order, need for ACLS, and admission to the ICU (Table 2). Additionally, at 1 and 6 years, NT‐proBNP and non‐ACS were associated with higher mortality. At 6 years, creatinine was an additional significant factor. We separated the ACS and non‐ACS cohorts and performed the same model for 1‐year mortality (Table 3). The models yielded similar factors associated with higher mortality: hospice, DNR order, need for ACLS, age, and NT‐proBNP, with ICU admission being significant only in the non‐ACS cohort.

DISCUSSION

Our findings confirm the important, but perhaps not well‐recognized, fact that an elevated troponin without ACS is associated with higher mortality than with ACS. This has been previously observed in veteran and nonveteran populations.[4, 6, 8, 12] The novel finding from our investigation is that mortality risk with troponin elevation is most strongly associated with unmodifiable clinical factors that are plausible explanations of risk. Furthermore, the distribution of these factors between our 2 cohorts does not sufficiently explain the difference in risk between ACS and non‐ACS patients.

At each time point we evaluated, ICU admission and need for ACLS were associated with mortality. These are indicators of a severely ill population and are not surprising to find associated with mortality. Many hospitals have instituted some form of pre‐code approach or rapid response team to identify patients before they need ACLS. These efforts, although well meaning, have not yielded convincing results of effectiveness.[13] Hospice and DNR patients were also, not surprisingly, associated with higher mortality. Although these factors were statistically significant, the low prevalence suggests that they are not clinically impactful on the primary questions of the investigation. These factors can be altered but are not intended as modifiable as they reflect the wishes of patients and their decision makers. The distribution of the factors in our model, however, did not adequately explain the higher risk of death with non‐ACS troponin elevation. For example, ACLS administration, hospice care, and DNR orders were strong predictors but were similar between the groups. ICU admission was actually more common with ACS patients, despite strong association with mortality. Age and NT‐proBNP were associated with mortality and higher in the non‐ACS group; however the magnitude of hazard was less than for the other factors. These findings lead us back to the possible explanation that non‐ACS troponin elevation stands as an independent risk factor, and that ACS patients have a distinct advantage in the myriad treatments available. If ACS patients were misdiagnosed as non‐ACS and failed to receive appropriate treatments, that might have contributed to higher mortality; however, we consider that unlikely given that the goal of the QI project was to minimize missed ACS diagnoses.

The overall mortality risk in our study was high: 12.7% at 30 days and 37.6% at 1 year. This reflects the high‐risk population with elevated troponin seen at our facility with ages nearly 70 years and high prevalence of multiple cardiovascular risk factors. Despite a high event rate, many clinically relevant risk factors were not retained in our Cox hazard model. Among sepsis patients, elevation in troponin is associated with mortality[14]; however, in our population neither fever or white blood cell count were significant mortality factors. The relationship between chronic kidney disease and troponin is complex. Renal dysfunction may result in troponin elevation and troponin elevation is a predictor of risk within kidney disease patients.[15] In our study, we did not evaluate chronic kidney disease as a predictor, instead opting to use the serum creatinine. This was not associated with mortality except at the 6‐year time point.

The TIMI score was not associated with mortality in either the overall population or the ACS cohort. The proportion of patients in our cohort with TIMI score under 3 was 16.5% as compared with 21.6% in the original derivation study.[16] The limited data on the prognostic value of the TIMI score within a veteran population suggest a modest predictive capacity.[17] Our data raises the possibility that TIMI is not an optimal choice; however, our analysis only includes all‐cause mortality, different from the original intended use of TIMI, predicting a variety of major cardiac events.

Our data confirm that ACS can be detected in a wide range of clinical presentations. Within our population of troponin positive patients, those with chest pain were most likely to be diagnosed with ACS, although one‐third of chest pain patients were felt to have a non‐ACS diagnosis. On the opposite end of the spectrum, an elevation in troponin with altered mental status or confusion was rarely diagnosed as ACSonly 18% of the time. Many symptoms were poor predictors of ACS; however, none were low enough to disregard. Our data would suggest that most patients with elevated troponin warrant evaluation by a cardiovascular expert.

Our study population came from a single VA hospital that is comprised of elderly and predominantly male patients limiting applicability to other populations. Despite this, other investigations in younger populations and with a higher proportion of women have found similar mortality trends.[4, 8, 12] We did not have sufficient data to determine the cause of death or to further classify as cardiac versus noncardiac; knowledge of the cause of the specific death may better inform future investigations into this important clinical question. Our investigation did not use a standardized definition to determine ACS, a notable limitation that could introduce bias or variation in care. Because all determinations about ACS were made prospectively as part of a QI project, we have little reason to suspect any systematic bias to the determination of ACS. With regard to variation in care, we have previously presented data demonstrating consistent rates of ACS diagnosis across the physicians at our facility.

Based on our investigation and others on this topic, non‐ACS troponin elevation is a common, high‐risk clinical scenario. In our cohort, non‐ACS troponin elevation is about twice as frequent as ACS, and the problem is likely to grow dramatically within the next few years as ultrasensitive troponin assays are eventually approved for use in the United States. These assays are much more sensitive than the current assays, and may make it challenging to distinguish between someone with an acute supply/demand mismatch from someone with an elevated troponin due to chronic, but stable, illness such as CAD, heart failure, or diabetes. Non‐ACS troponin elevation remains poorly understood, with no viable treatment options other than addressing the pathophysiology resulting in the troponin elevation. Due to the heterogeneity of the diagnoses and pathophysiological conditions that result in elevated troponin, a unifying treatment is not likely feasible.

In conclusion, in this elderly, male veteran population, the mortality impact associated with a cardiac troponin elevation was not limited to ACS, as mortality was high among those without ACS. Factors independently associated with this non‐ACS mortality risk were plausible, but did not elucidate the reasons why non‐ACS troponin elevation carries a higher risk. Attempting to better understand the biological basis for the troponin elevation in these non‐ACS patients is a critical unmet need.

Acute coronary syndromes (ACS) are potentially lethal and present with a wide variety of symptoms. As such, physicians frequently order cardiac biomarkers, such as cardiac troponin, for patients with acute complaints. Elevated troponin is associated with higher risk of mortality regardless of the causes, which can be myriad, both chronic and acute.[1] Among patients with an elevated troponin, distinguishing ACS from non‐ACS can be challenging.

Making the distinction between ACS and non‐ACS troponin elevation is crucial because the underlying pathophysiology and subsequent management strategies are markedly different.[2] According to evidence‐based practice guidelines, ACS is managed with antiplatelet drugs, statins, and percutaneous coronary intervention, improving clinical outcomes.[3] In contrast, care for patients with non‐ACS troponin elevations is usually supportive, with a focus on the underlying conditions. The lack of specific treatment options for such patients is concerning given that several series have suggested that non‐ACS troponin patients may have a higher mortality risk than ACS patients.[4, 5, 6] Non‐ACS troponin elevation can be the result of a multitude of conditions.[7, 8] What remains unclear at this point is whether the excess mortality observed with non‐ACS troponin elevation is due to myocardial damage or to the underlying conditions that predispose to troponin release.

Using data from a quality improvement (QI) project collected at our Veterans Affairs (VA) medical center, we investigated the mortality risk associated with ACS and non‐ACS troponin elevation including an analysis of factors associated with mortality. We hypothesized that non‐ACS troponin elevation will have a higher mortality risk than troponin elevation due to ACS, and that important contributors to this relationship could be identified to provide direction for future investigation directed at modifying this mortality risk.

METHODS

We analyzed data that were prospectively collected for a quality initiative between 2006 and 2007. The project was a collaborative endeavor between cardiology, hospital medicine, and emergency medicine with the process goal of better identifying patients with ACS to hopefully improve outcomes. The QI team was consulted in real time to assist with treatment recommendations; no retrospective decisions were made regarding whether or not ACS was present. As the goal of the project was to improve cardiovascular outcomes, consultative advice was freely provided, and no physicians or teams were subject to any adverse repercussions for their diagnoses or management decisions.

A cardiologist‐led team was created to improve quality of care for myocardial infarction patients by evaluating all patients at our facility with an elevated troponin. On a daily basis, a specialist clinical coordinator (nurse practitioner or physician assistant) received a list of all patients with elevated troponin from the chemistry lab. The coordinator reviewed the patients' medical records with a cardiologist. A positive troponin was defined as a troponin T level of greater than 0.03 ng/mL (99th percentile at our facility). Each attending cardiologist prospectively determined if troponin elevation was related to clinical findings consistent with an ACS based on review of the patients' symptoms (duration, quality, severity, chronicity, and alleviating/aggravating factors), medical history, and noninvasive cardiac testing including electrocardiograms, cardiac biomarkers, and any other available imaging tests.

We have previously demonstrated that the cardiologists at our facility have a similar rate of diagnosing ACS.[9] All cardiologists at our facility maintain current American Board of Internal Medicine certification in cardiovascular disease and have academic appointments at the University of Florida College of Medicine. All patients were followed prospectively, and data on their medical history, acute evaluation, and outcomes were tracked in an electronic database. Given the higher risk of mortality with ST‐elevation myocardial infarction, such patients were excluded from this investigation. By definition, patients with unstable angina do not have elevated biomarkers and thus would not have been included in the database to begin with. Prospectively recorded data elements included: age, gender, chief complaint, tobacco use, presence of hypertension, hyperlipidemia, prior coronary disease, chronic kidney disease, diabetes mellitus, cardiac troponin values, serum creatinine, electrocardiogram (ECG) variables, Thrombolysis in Myocardial Infarction (TIMI) score, and if the patient was placed under hospice care or an active do‐not‐resuscitate (DNR) order. Additional data elements gathered at a later date included maximum temperature, white blood cell count, N‐terminal pro‐brain natriuretic peptide (NT‐proBNP), administration of advanced cardiac life support (ACLS), and admission to an intensive care unit (ICU). All consecutive patients with elevated troponin were included in the database; if patients were included more than once, we used their index evaluation only. All patients with troponin elevation after revascularization (percutaneous coronary intervention or coronary bypass surgery) were excluded. Our investigational design was reviewed by our institutional review board, who waived the requirement for formal written informed consent and approved use of data from this QI project for research purposes.

We focused this investigation on an analysis of all‐cause mortality in February 2014. We analyzed mortality at 30 days, 1 year, and 6 years. As secondary outcomes we analyzed the likelihood of the patients' chief complaint for the diagnosis of ACS and evaluated predictors of mortality based on Cox proportional hazard modeling. Mortality within the VA system is reliably tracked and compares favorably to the Social Security National Death Index Master File for accuracy.[10, 11]

Categorical variables were compared by ² test. The Student t test was used to compare normally distributed continuous variables, and nonparametric tests were used for non‐normal distributions as appropriate. Mortality data at 30 days, 1 year, and 6 years were compared by log‐rank test and Kaplan‐Meier graphs. A formal power analysis was not performed; the entire available population was included. A Cox proportional hazard model was created to estimate mortality risk at each time point. Variables included in our Cox regression model were age, gender, history of coronary artery disease (CAD), hypertension, diabetes mellitus or hyperlipidemia, ACS diagnosis, dynamic ECG changes, TIMI risk score, initial troponin level, creatinine level at time of initial troponin (per mg/dL), presence of fever, maximum white blood cell count, NT‐proBNP level (per 1000 pg/mL), if ACLS was performed, if the patient was under hospice care, if there was a DNR order, and if they required ICU admission. This model was also constructed independently for the ACS and non‐ACS cohorts for mortality at 1 year. A forward stepwise model was used. Statistical results were considered significant at P < 0.05. Statistical analyses were performed using SPSS version 21 (IBM, Armonk, NY).

RESULTS

Among the 761 patients, 502 (66.0%) were classified as non‐ACS and 259 (34.0%) as ACS (Table 1). The mean age was higher in the non‐ACS group (71 years vs 69 years in the ACS group, P = 0.006). Hypertension, diabetes mellitus, and prior CAD were frequent in both groups and not significantly different. Median initial troponin T was higher in the ACS group (0.12 ng/mL vs 0.06 ng/mL, P < 0.001) as were the frequency of a TIMI risk score >2 (92.5% vs 74.3%, P < 0.001) and new ECG changes (29.7% vs 8.2%, P < 0.001). Hospice, DNR orders, and administration of ACLS were not different between groups; however, admission to the ICU was more frequent in the ACS group (44.8% vs 31.9%, P < 0.001). Chest pain was the symptom with the highest positive predictive value for the diagnosis of ACS (63.3%), whereas the least predictive was altered mental status or confusion (18.0%) (Figure 1).

Figure 1

Mortality at 30 days was not different between the 2 groups, but mortality was higher for the non‐ACS cohort at 1 year and at 6 years (Table 1). Kaplan‐Meier curves demonstrate that mortality for the 2 cohorts begins to diverge between 30 and 60 days until approximately 2 years when the curves again are parallel (Figure 2).

Figure 2

In Cox proportional hazards models, 5 factors were associated with higher mortality at 30 days, 1 year, and at 6 years: age, hospice, DNR order, need for ACLS, and admission to the ICU (Table 2). Additionally, at 1 and 6 years, NT‐proBNP and non‐ACS were associated with higher mortality. At 6 years, creatinine was an additional significant factor. We separated the ACS and non‐ACS cohorts and performed the same model for 1‐year mortality (Table 3). The models yielded similar factors associated with higher mortality: hospice, DNR order, need for ACLS, age, and NT‐proBNP, with ICU admission being significant only in the non‐ACS cohort.

DISCUSSION

Our findings confirm the important, but perhaps not well‐recognized, fact that an elevated troponin without ACS is associated with higher mortality than with ACS. This has been previously observed in veteran and nonveteran populations.[4, 6, 8, 12] The novel finding from our investigation is that mortality risk with troponin elevation is most strongly associated with unmodifiable clinical factors that are plausible explanations of risk. Furthermore, the distribution of these factors between our 2 cohorts does not sufficiently explain the difference in risk between ACS and non‐ACS patients.

At each time point we evaluated, ICU admission and need for ACLS were associated with mortality. These are indicators of a severely ill population and are not surprising to find associated with mortality. Many hospitals have instituted some form of pre‐code approach or rapid response team to identify patients before they need ACLS. These efforts, although well meaning, have not yielded convincing results of effectiveness.[13] Hospice and DNR patients were also, not surprisingly, associated with higher mortality. Although these factors were statistically significant, the low prevalence suggests that they are not clinically impactful on the primary questions of the investigation. These factors can be altered but are not intended as modifiable as they reflect the wishes of patients and their decision makers. The distribution of the factors in our model, however, did not adequately explain the higher risk of death with non‐ACS troponin elevation. For example, ACLS administration, hospice care, and DNR orders were strong predictors but were similar between the groups. ICU admission was actually more common with ACS patients, despite strong association with mortality. Age and NT‐proBNP were associated with mortality and higher in the non‐ACS group; however the magnitude of hazard was less than for the other factors. These findings lead us back to the possible explanation that non‐ACS troponin elevation stands as an independent risk factor, and that ACS patients have a distinct advantage in the myriad treatments available. If ACS patients were misdiagnosed as non‐ACS and failed to receive appropriate treatments, that might have contributed to higher mortality; however, we consider that unlikely given that the goal of the QI project was to minimize missed ACS diagnoses.

The overall mortality risk in our study was high: 12.7% at 30 days and 37.6% at 1 year. This reflects the high‐risk population with elevated troponin seen at our facility with ages nearly 70 years and high prevalence of multiple cardiovascular risk factors. Despite a high event rate, many clinically relevant risk factors were not retained in our Cox hazard model. Among sepsis patients, elevation in troponin is associated with mortality[14]; however, in our population neither fever or white blood cell count were significant mortality factors. The relationship between chronic kidney disease and troponin is complex. Renal dysfunction may result in troponin elevation and troponin elevation is a predictor of risk within kidney disease patients.[15] In our study, we did not evaluate chronic kidney disease as a predictor, instead opting to use the serum creatinine. This was not associated with mortality except at the 6‐year time point.

The TIMI score was not associated with mortality in either the overall population or the ACS cohort. The proportion of patients in our cohort with TIMI score under 3 was 16.5% as compared with 21.6% in the original derivation study.[16] The limited data on the prognostic value of the TIMI score within a veteran population suggest a modest predictive capacity.[17] Our data raises the possibility that TIMI is not an optimal choice; however, our analysis only includes all‐cause mortality, different from the original intended use of TIMI, predicting a variety of major cardiac events.

Our data confirm that ACS can be detected in a wide range of clinical presentations. Within our population of troponin positive patients, those with chest pain were most likely to be diagnosed with ACS, although one‐third of chest pain patients were felt to have a non‐ACS diagnosis. On the opposite end of the spectrum, an elevation in troponin with altered mental status or confusion was rarely diagnosed as ACSonly 18% of the time. Many symptoms were poor predictors of ACS; however, none were low enough to disregard. Our data would suggest that most patients with elevated troponin warrant evaluation by a cardiovascular expert.

Our study population came from a single VA hospital that is comprised of elderly and predominantly male patients limiting applicability to other populations. Despite this, other investigations in younger populations and with a higher proportion of women have found similar mortality trends.[4, 8, 12] We did not have sufficient data to determine the cause of death or to further classify as cardiac versus noncardiac; knowledge of the cause of the specific death may better inform future investigations into this important clinical question. Our investigation did not use a standardized definition to determine ACS, a notable limitation that could introduce bias or variation in care. Because all determinations about ACS were made prospectively as part of a QI project, we have little reason to suspect any systematic bias to the determination of ACS. With regard to variation in care, we have previously presented data demonstrating consistent rates of ACS diagnosis across the physicians at our facility.

Based on our investigation and others on this topic, non‐ACS troponin elevation is a common, high‐risk clinical scenario. In our cohort, non‐ACS troponin elevation is about twice as frequent as ACS, and the problem is likely to grow dramatically within the next few years as ultrasensitive troponin assays are eventually approved for use in the United States. These assays are much more sensitive than the current assays, and may make it challenging to distinguish between someone with an acute supply/demand mismatch from someone with an elevated troponin due to chronic, but stable, illness such as CAD, heart failure, or diabetes. Non‐ACS troponin elevation remains poorly understood, with no viable treatment options other than addressing the pathophysiology resulting in the troponin elevation. Due to the heterogeneity of the diagnoses and pathophysiological conditions that result in elevated troponin, a unifying treatment is not likely feasible.

In conclusion, in this elderly, male veteran population, the mortality impact associated with a cardiac troponin elevation was not limited to ACS, as mortality was high among those without ACS. Factors independently associated with this non‐ACS mortality risk were plausible, but did not elucidate the reasons why non‐ACS troponin elevation carries a higher risk. Attempting to better understand the biological basis for the troponin elevation in these non‐ACS patients is a critical unmet need.

Acute coronary syndromes (ACS) are potentially lethal and present with a wide variety of symptoms. As such, physicians frequently order cardiac biomarkers, such as cardiac troponin, for patients with acute complaints. Elevated troponin is associated with higher risk of mortality regardless of the causes, which can be myriad, both chronic and acute.[1] Among patients with an elevated troponin, distinguishing ACS from non‐ACS can be challenging.

Making the distinction between ACS and non‐ACS troponin elevation is crucial because the underlying pathophysiology and subsequent management strategies are markedly different.[2] According to evidence‐based practice guidelines, ACS is managed with antiplatelet drugs, statins, and percutaneous coronary intervention, improving clinical outcomes.[3] In contrast, care for patients with non‐ACS troponin elevations is usually supportive, with a focus on the underlying conditions. The lack of specific treatment options for such patients is concerning given that several series have suggested that non‐ACS troponin patients may have a higher mortality risk than ACS patients.[4, 5, 6] Non‐ACS troponin elevation can be the result of a multitude of conditions.[7, 8] What remains unclear at this point is whether the excess mortality observed with non‐ACS troponin elevation is due to myocardial damage or to the underlying conditions that predispose to troponin release.

Using data from a quality improvement (QI) project collected at our Veterans Affairs (VA) medical center, we investigated the mortality risk associated with ACS and non‐ACS troponin elevation including an analysis of factors associated with mortality. We hypothesized that non‐ACS troponin elevation will have a higher mortality risk than troponin elevation due to ACS, and that important contributors to this relationship could be identified to provide direction for future investigation directed at modifying this mortality risk.

METHODS

We analyzed data that were prospectively collected for a quality initiative between 2006 and 2007. The project was a collaborative endeavor between cardiology, hospital medicine, and emergency medicine with the process goal of better identifying patients with ACS to hopefully improve outcomes. The QI team was consulted in real time to assist with treatment recommendations; no retrospective decisions were made regarding whether or not ACS was present. As the goal of the project was to improve cardiovascular outcomes, consultative advice was freely provided, and no physicians or teams were subject to any adverse repercussions for their diagnoses or management decisions.

A cardiologist‐led team was created to improve quality of care for myocardial infarction patients by evaluating all patients at our facility with an elevated troponin. On a daily basis, a specialist clinical coordinator (nurse practitioner or physician assistant) received a list of all patients with elevated troponin from the chemistry lab. The coordinator reviewed the patients' medical records with a cardiologist. A positive troponin was defined as a troponin T level of greater than 0.03 ng/mL (99th percentile at our facility). Each attending cardiologist prospectively determined if troponin elevation was related to clinical findings consistent with an ACS based on review of the patients' symptoms (duration, quality, severity, chronicity, and alleviating/aggravating factors), medical history, and noninvasive cardiac testing including electrocardiograms, cardiac biomarkers, and any other available imaging tests.

We have previously demonstrated that the cardiologists at our facility have a similar rate of diagnosing ACS.[9] All cardiologists at our facility maintain current American Board of Internal Medicine certification in cardiovascular disease and have academic appointments at the University of Florida College of Medicine. All patients were followed prospectively, and data on their medical history, acute evaluation, and outcomes were tracked in an electronic database. Given the higher risk of mortality with ST‐elevation myocardial infarction, such patients were excluded from this investigation. By definition, patients with unstable angina do not have elevated biomarkers and thus would not have been included in the database to begin with. Prospectively recorded data elements included: age, gender, chief complaint, tobacco use, presence of hypertension, hyperlipidemia, prior coronary disease, chronic kidney disease, diabetes mellitus, cardiac troponin values, serum creatinine, electrocardiogram (ECG) variables, Thrombolysis in Myocardial Infarction (TIMI) score, and if the patient was placed under hospice care or an active do‐not‐resuscitate (DNR) order. Additional data elements gathered at a later date included maximum temperature, white blood cell count, N‐terminal pro‐brain natriuretic peptide (NT‐proBNP), administration of advanced cardiac life support (ACLS), and admission to an intensive care unit (ICU). All consecutive patients with elevated troponin were included in the database; if patients were included more than once, we used their index evaluation only. All patients with troponin elevation after revascularization (percutaneous coronary intervention or coronary bypass surgery) were excluded. Our investigational design was reviewed by our institutional review board, who waived the requirement for formal written informed consent and approved use of data from this QI project for research purposes.

We focused this investigation on an analysis of all‐cause mortality in February 2014. We analyzed mortality at 30 days, 1 year, and 6 years. As secondary outcomes we analyzed the likelihood of the patients' chief complaint for the diagnosis of ACS and evaluated predictors of mortality based on Cox proportional hazard modeling. Mortality within the VA system is reliably tracked and compares favorably to the Social Security National Death Index Master File for accuracy.[10, 11]

Categorical variables were compared by ² test. The Student t test was used to compare normally distributed continuous variables, and nonparametric tests were used for non‐normal distributions as appropriate. Mortality data at 30 days, 1 year, and 6 years were compared by log‐rank test and Kaplan‐Meier graphs. A formal power analysis was not performed; the entire available population was included. A Cox proportional hazard model was created to estimate mortality risk at each time point. Variables included in our Cox regression model were age, gender, history of coronary artery disease (CAD), hypertension, diabetes mellitus or hyperlipidemia, ACS diagnosis, dynamic ECG changes, TIMI risk score, initial troponin level, creatinine level at time of initial troponin (per mg/dL), presence of fever, maximum white blood cell count, NT‐proBNP level (per 1000 pg/mL), if ACLS was performed, if the patient was under hospice care, if there was a DNR order, and if they required ICU admission. This model was also constructed independently for the ACS and non‐ACS cohorts for mortality at 1 year. A forward stepwise model was used. Statistical results were considered significant at P < 0.05. Statistical analyses were performed using SPSS version 21 (IBM, Armonk, NY).

RESULTS

Among the 761 patients, 502 (66.0%) were classified as non‐ACS and 259 (34.0%) as ACS (Table 1). The mean age was higher in the non‐ACS group (71 years vs 69 years in the ACS group, P = 0.006). Hypertension, diabetes mellitus, and prior CAD were frequent in both groups and not significantly different. Median initial troponin T was higher in the ACS group (0.12 ng/mL vs 0.06 ng/mL, P < 0.001) as were the frequency of a TIMI risk score >2 (92.5% vs 74.3%, P < 0.001) and new ECG changes (29.7% vs 8.2%, P < 0.001). Hospice, DNR orders, and administration of ACLS were not different between groups; however, admission to the ICU was more frequent in the ACS group (44.8% vs 31.9%, P < 0.001). Chest pain was the symptom with the highest positive predictive value for the diagnosis of ACS (63.3%), whereas the least predictive was altered mental status or confusion (18.0%) (Figure 1).

Figure 1

Mortality at 30 days was not different between the 2 groups, but mortality was higher for the non‐ACS cohort at 1 year and at 6 years (Table 1). Kaplan‐Meier curves demonstrate that mortality for the 2 cohorts begins to diverge between 30 and 60 days until approximately 2 years when the curves again are parallel (Figure 2).

Figure 2

In Cox proportional hazards models, 5 factors were associated with higher mortality at 30 days, 1 year, and at 6 years: age, hospice, DNR order, need for ACLS, and admission to the ICU (Table 2). Additionally, at 1 and 6 years, NT‐proBNP and non‐ACS were associated with higher mortality. At 6 years, creatinine was an additional significant factor. We separated the ACS and non‐ACS cohorts and performed the same model for 1‐year mortality (Table 3). The models yielded similar factors associated with higher mortality: hospice, DNR order, need for ACLS, age, and NT‐proBNP, with ICU admission being significant only in the non‐ACS cohort.

DISCUSSION

Our findings confirm the important, but perhaps not well‐recognized, fact that an elevated troponin without ACS is associated with higher mortality than with ACS. This has been previously observed in veteran and nonveteran populations.[4, 6, 8, 12] The novel finding from our investigation is that mortality risk with troponin elevation is most strongly associated with unmodifiable clinical factors that are plausible explanations of risk. Furthermore, the distribution of these factors between our 2 cohorts does not sufficiently explain the difference in risk between ACS and non‐ACS patients.

At each time point we evaluated, ICU admission and need for ACLS were associated with mortality. These are indicators of a severely ill population and are not surprising to find associated with mortality. Many hospitals have instituted some form of pre‐code approach or rapid response team to identify patients before they need ACLS. These efforts, although well meaning, have not yielded convincing results of effectiveness.[13] Hospice and DNR patients were also, not surprisingly, associated with higher mortality. Although these factors were statistically significant, the low prevalence suggests that they are not clinically impactful on the primary questions of the investigation. These factors can be altered but are not intended as modifiable as they reflect the wishes of patients and their decision makers. The distribution of the factors in our model, however, did not adequately explain the higher risk of death with non‐ACS troponin elevation. For example, ACLS administration, hospice care, and DNR orders were strong predictors but were similar between the groups. ICU admission was actually more common with ACS patients, despite strong association with mortality. Age and NT‐proBNP were associated with mortality and higher in the non‐ACS group; however the magnitude of hazard was less than for the other factors. These findings lead us back to the possible explanation that non‐ACS troponin elevation stands as an independent risk factor, and that ACS patients have a distinct advantage in the myriad treatments available. If ACS patients were misdiagnosed as non‐ACS and failed to receive appropriate treatments, that might have contributed to higher mortality; however, we consider that unlikely given that the goal of the QI project was to minimize missed ACS diagnoses.

The overall mortality risk in our study was high: 12.7% at 30 days and 37.6% at 1 year. This reflects the high‐risk population with elevated troponin seen at our facility with ages nearly 70 years and high prevalence of multiple cardiovascular risk factors. Despite a high event rate, many clinically relevant risk factors were not retained in our Cox hazard model. Among sepsis patients, elevation in troponin is associated with mortality[14]; however, in our population neither fever or white blood cell count were significant mortality factors. The relationship between chronic kidney disease and troponin is complex. Renal dysfunction may result in troponin elevation and troponin elevation is a predictor of risk within kidney disease patients.[15] In our study, we did not evaluate chronic kidney disease as a predictor, instead opting to use the serum creatinine. This was not associated with mortality except at the 6‐year time point.

The TIMI score was not associated with mortality in either the overall population or the ACS cohort. The proportion of patients in our cohort with TIMI score under 3 was 16.5% as compared with 21.6% in the original derivation study.[16] The limited data on the prognostic value of the TIMI score within a veteran population suggest a modest predictive capacity.[17] Our data raises the possibility that TIMI is not an optimal choice; however, our analysis only includes all‐cause mortality, different from the original intended use of TIMI, predicting a variety of major cardiac events.

Our data confirm that ACS can be detected in a wide range of clinical presentations. Within our population of troponin positive patients, those with chest pain were most likely to be diagnosed with ACS, although one‐third of chest pain patients were felt to have a non‐ACS diagnosis. On the opposite end of the spectrum, an elevation in troponin with altered mental status or confusion was rarely diagnosed as ACSonly 18% of the time. Many symptoms were poor predictors of ACS; however, none were low enough to disregard. Our data would suggest that most patients with elevated troponin warrant evaluation by a cardiovascular expert.

Our study population came from a single VA hospital that is comprised of elderly and predominantly male patients limiting applicability to other populations. Despite this, other investigations in younger populations and with a higher proportion of women have found similar mortality trends.[4, 8, 12] We did not have sufficient data to determine the cause of death or to further classify as cardiac versus noncardiac; knowledge of the cause of the specific death may better inform future investigations into this important clinical question. Our investigation did not use a standardized definition to determine ACS, a notable limitation that could introduce bias or variation in care. Because all determinations about ACS were made prospectively as part of a QI project, we have little reason to suspect any systematic bias to the determination of ACS. With regard to variation in care, we have previously presented data demonstrating consistent rates of ACS diagnosis across the physicians at our facility.

Based on our investigation and others on this topic, non‐ACS troponin elevation is a common, high‐risk clinical scenario. In our cohort, non‐ACS troponin elevation is about twice as frequent as ACS, and the problem is likely to grow dramatically within the next few years as ultrasensitive troponin assays are eventually approved for use in the United States. These assays are much more sensitive than the current assays, and may make it challenging to distinguish between someone with an acute supply/demand mismatch from someone with an elevated troponin due to chronic, but stable, illness such as CAD, heart failure, or diabetes. Non‐ACS troponin elevation remains poorly understood, with no viable treatment options other than addressing the pathophysiology resulting in the troponin elevation. Due to the heterogeneity of the diagnoses and pathophysiological conditions that result in elevated troponin, a unifying treatment is not likely feasible.

In conclusion, in this elderly, male veteran population, the mortality impact associated with a cardiac troponin elevation was not limited to ACS, as mortality was high among those without ACS. Factors independently associated with this non‐ACS mortality risk were plausible, but did not elucidate the reasons why non‐ACS troponin elevation carries a higher risk. Attempting to better understand the biological basis for the troponin elevation in these non‐ACS patients is a critical unmet need.

Rapid response systems (RRSs) have been developed to identify and treat deteriorating patients on general hospital units.[1] The most commonly proposed approach to the problem of identifying and stabilizing deteriorating hospitalized patients includes some combination of an early warning system to detect the deterioration and an RRS to deal with it. We previously demonstrated that a relatively simple hospital‐specific prediction model employing routine laboratory values and vital sign data is capable of predicting clinical deterioration, the need for intensive care unit (ICU) transfer, and hospital mortality in patients admitted to general medicine units.[2, 3, 4, 5, 6]

Hospital readmissions within 30 days of hospital discharge occur often and are difficult to predict. Starting in 2013, readmission penalties have been applied to specific conditions in the United States (acute myocardial infarction, heart failure, and pneumonia), with the expectation that additional conditions will be added to this group in years to come.[7, 8] Unfortunately, interventions developed to date have not been universally successful in preventing hospital readmissions for various medical conditions and patient types.[9] One potential explanation for this is the inability to reliably predict which patients are at risk for readmission to better target preventative interventions. Predictors of hospital readmission can be disease specific, such as the presence of multivessel disease in patients hospitalized with myocardial infarction,[10] or more general, such as lack of available medical follow‐up postdischarge.[11] Therefore, we performed a study to determine whether the occurrence of automated clinical deterioration alerts (CDAs) predicted 30‐day hospital readmission.

METHODS

RESULTS

The final cohort had 3015 patients with a mean age of 57.5 17.5 years and 47.8% males. The most common reasons for hospital admission were infection or sepsis syndrome including pneumonia and urinary tract infections (23.6%), congestive heart failure or other cardiac conditions (18.4%), respiratory distress including chronic obstructive pulmonary disease (16.2%), acute or chronic renal failure (9.7%), gastrointestinal disorders (8.4%), and diabetes mellitus management (7.4%). Overall, there were 567 (18.8%) patients who were readmitted within 30 days of their hospital discharge date.

Table 2 shows the characteristics of patients readmitted within 30 days and of patients not requiring hospital readmission within 30 days. Patients requiring hospital readmission within 30 days were younger and had significantly more comorbidities as manifested by significantly greater Charlson scores and individual comorbidities including coronary artery disease, congestive heart disease, peripheral vascular disease, connective tissue disease, cirrhosis, diabetes mellitus with end‐organ complications, renal failure, and metastatic cancer. Patients with a 30‐day readmission had significantly longer duration of hospitalization, more emergency department visits in the 6 months prior to the index hospitalization, lower minimum hemoglobin measurements, higher minimum serum creatinine values, and were more likely to have Medicare or Medicaid insurance compared to patients without a 30‐day readmission.

There were 1141 (34.4%) patients that triggered a CDA. Patients triggering a CDA were significantly more likely to have a 30‐day readmission compared to those who did not trigger a CDA (23.6% vs 15.9%; P < 0.001). Patients triggering a CDA were also significantly more likely to be readmitted within 60 days (31.7% vs 22.1%; P < 0.001) and 90 days (35.8% vs 26.2%; P < 0.001) compared to patients who did not trigger a CDA. Multiple logistic regression identified the triggering of a CDA to be independently associated with 30‐day readmission (OR: 1.40; 95% CI: 1.26‐1.55; P = 0.001) (Table 3). Other independent predictors of 30‐day readmission were: an emergency department visit in the previous 6 months, increasing age in 1‐year increments, presence of connective tissue disease, diabetes mellitus with end‐organ complications, chronic renal disease, cirrhosis, and metastatic cancer (Hosmer‐Lemeshow goodness of fit test, 0.363). Figure 1 reveals the ROC curves for the logistic regression model (Table 3) with and without the CDA variable. As the ROC curves document, the 2 models had similar sensitivity for the entire range of specificities. Reflecting this, the area under the ROC curve for the model inclusive of the CDA variable equaled 0.675 (95% CI: 0.649‐0.700), whereas the area under the ROC curve for the model excluding the CDA variable equaled 0.658 (95% CI: 0.632‐0.684).

Figure 1

DISCUSSION

We demonstrated that the occurrence of an automated CDA is associated with increased risk for 30‐day hospital readmission. However, the addition of the CDA variable to the other variables identified to be independently associated with 30‐day readmission (Table 3) did not significantly add to the overall predictive accuracy of the derived logistic regression model. Other investigators have previously attempted to develop automated predictors of hospital readmission. Amarasingham et al. developed a real‐time electronic predictive model that identifies hospitalized heart failure patients at high risk for readmission or death from clinical and nonclinical risk factors present on admission.[13] Their electronic model demonstrated good discrimination for 30‐day mortality and readmission and performed as well, or better than, models developed by the Center for Medicaid and Medicare Services and the Acute Decompensated Heart Failure Registry. Similarly, Baillie et al. developed an automated prediction model that was effectively integrated into an existing EMR and identified patients on admission who were at risk for readmission within 30 days of discharge.[14] Our automated CDA differs from these previous risk predictors by surveying patients throughout their hospital stay as opposed to identifying risk for readmission at a single time point.

Several limitations of our study should be recognized. First, this was a noninterventional study aimed at examining the ability of CDAs to predict hospital readmission. Future studies are needed to assess whether the use of enhanced readmission prediction algorithms can be utilized to avert hospital readmissions. Second, the data derive from a single center, and this necessarily limits the generalizability of our findings. As such, our results may not reflect what one might see at other institutions. For example, Barnes‐Jewish Hospital has a regional referral pattern that includes community hospitals, regional long‐term acute care hospitals, nursing homes, and chronic wound, dialysis, and infusion clinics. This may explain, in part, the relatively high rate of hospital readmission observed in our cohort. Third, there is the possibility that CDAs were associated with readmission by chance given the number of potential predictor variables examined. The importance of CDAs as a determinant of rehospitalization requires confirmation in other independent populations. Fourth, it is likely that we did not capture all hospital readmissions, primarily those occurring outside of our hospital system. Therefore, we may have underestimated the actual rates of readmission for this cohort. Finally, we cannot be certain that all important predictors of hospital readmission were captured in this study.

The development of an accurate real‐time early warning system has the potential to identify patients at risk for various adverse outcomes including clinical deterioration, hospital death, and postdischarge readmission. By identifying patients at greatest risk for readmission, valuable healthcare resources can be better targeted to such populations. Our findings suggest that existing readmission predictors may suboptimally risk‐stratify patients, and it may be important to include additional clinical variables if pay for performance and other across‐institution comparisons are to be fair to institutions that care for more seriously ill patients. The variables identified as predictors of 30‐day hospital readmission in our study, with the exception of a CDA, are all readily identifiable clinical characteristics. The modest incremental value of a CDA to these clinical characteristics suggests that they would suffice for the identification of patients at high risk for hospital readmission. This is especially important for safety‐net institutions not routinely employing automated CDAs. These safety‐net hospitals provide a disproportionate level of care for patients who otherwise would have difficulty obtaining inpatient medical care and disproportionately carry the greatest burden of hospital readmissions.[15]

Rapid response systems (RRSs) have been developed to identify and treat deteriorating patients on general hospital units.[1] The most commonly proposed approach to the problem of identifying and stabilizing deteriorating hospitalized patients includes some combination of an early warning system to detect the deterioration and an RRS to deal with it. We previously demonstrated that a relatively simple hospital‐specific prediction model employing routine laboratory values and vital sign data is capable of predicting clinical deterioration, the need for intensive care unit (ICU) transfer, and hospital mortality in patients admitted to general medicine units.[2, 3, 4, 5, 6]

Hospital readmissions within 30 days of hospital discharge occur often and are difficult to predict. Starting in 2013, readmission penalties have been applied to specific conditions in the United States (acute myocardial infarction, heart failure, and pneumonia), with the expectation that additional conditions will be added to this group in years to come.[7, 8] Unfortunately, interventions developed to date have not been universally successful in preventing hospital readmissions for various medical conditions and patient types.[9] One potential explanation for this is the inability to reliably predict which patients are at risk for readmission to better target preventative interventions. Predictors of hospital readmission can be disease specific, such as the presence of multivessel disease in patients hospitalized with myocardial infarction,[10] or more general, such as lack of available medical follow‐up postdischarge.[11] Therefore, we performed a study to determine whether the occurrence of automated clinical deterioration alerts (CDAs) predicted 30‐day hospital readmission.

METHODS

RESULTS

The final cohort had 3015 patients with a mean age of 57.5 17.5 years and 47.8% males. The most common reasons for hospital admission were infection or sepsis syndrome including pneumonia and urinary tract infections (23.6%), congestive heart failure or other cardiac conditions (18.4%), respiratory distress including chronic obstructive pulmonary disease (16.2%), acute or chronic renal failure (9.7%), gastrointestinal disorders (8.4%), and diabetes mellitus management (7.4%). Overall, there were 567 (18.8%) patients who were readmitted within 30 days of their hospital discharge date.

Table 2 shows the characteristics of patients readmitted within 30 days and of patients not requiring hospital readmission within 30 days. Patients requiring hospital readmission within 30 days were younger and had significantly more comorbidities as manifested by significantly greater Charlson scores and individual comorbidities including coronary artery disease, congestive heart disease, peripheral vascular disease, connective tissue disease, cirrhosis, diabetes mellitus with end‐organ complications, renal failure, and metastatic cancer. Patients with a 30‐day readmission had significantly longer duration of hospitalization, more emergency department visits in the 6 months prior to the index hospitalization, lower minimum hemoglobin measurements, higher minimum serum creatinine values, and were more likely to have Medicare or Medicaid insurance compared to patients without a 30‐day readmission.

There were 1141 (34.4%) patients that triggered a CDA. Patients triggering a CDA were significantly more likely to have a 30‐day readmission compared to those who did not trigger a CDA (23.6% vs 15.9%; P < 0.001). Patients triggering a CDA were also significantly more likely to be readmitted within 60 days (31.7% vs 22.1%; P < 0.001) and 90 days (35.8% vs 26.2%; P < 0.001) compared to patients who did not trigger a CDA. Multiple logistic regression identified the triggering of a CDA to be independently associated with 30‐day readmission (OR: 1.40; 95% CI: 1.26‐1.55; P = 0.001) (Table 3). Other independent predictors of 30‐day readmission were: an emergency department visit in the previous 6 months, increasing age in 1‐year increments, presence of connective tissue disease, diabetes mellitus with end‐organ complications, chronic renal disease, cirrhosis, and metastatic cancer (Hosmer‐Lemeshow goodness of fit test, 0.363). Figure 1 reveals the ROC curves for the logistic regression model (Table 3) with and without the CDA variable. As the ROC curves document, the 2 models had similar sensitivity for the entire range of specificities. Reflecting this, the area under the ROC curve for the model inclusive of the CDA variable equaled 0.675 (95% CI: 0.649‐0.700), whereas the area under the ROC curve for the model excluding the CDA variable equaled 0.658 (95% CI: 0.632‐0.684).

Figure 1

DISCUSSION

We demonstrated that the occurrence of an automated CDA is associated with increased risk for 30‐day hospital readmission. However, the addition of the CDA variable to the other variables identified to be independently associated with 30‐day readmission (Table 3) did not significantly add to the overall predictive accuracy of the derived logistic regression model. Other investigators have previously attempted to develop automated predictors of hospital readmission. Amarasingham et al. developed a real‐time electronic predictive model that identifies hospitalized heart failure patients at high risk for readmission or death from clinical and nonclinical risk factors present on admission.[13] Their electronic model demonstrated good discrimination for 30‐day mortality and readmission and performed as well, or better than, models developed by the Center for Medicaid and Medicare Services and the Acute Decompensated Heart Failure Registry. Similarly, Baillie et al. developed an automated prediction model that was effectively integrated into an existing EMR and identified patients on admission who were at risk for readmission within 30 days of discharge.[14] Our automated CDA differs from these previous risk predictors by surveying patients throughout their hospital stay as opposed to identifying risk for readmission at a single time point.

Several limitations of our study should be recognized. First, this was a noninterventional study aimed at examining the ability of CDAs to predict hospital readmission. Future studies are needed to assess whether the use of enhanced readmission prediction algorithms can be utilized to avert hospital readmissions. Second, the data derive from a single center, and this necessarily limits the generalizability of our findings. As such, our results may not reflect what one might see at other institutions. For example, Barnes‐Jewish Hospital has a regional referral pattern that includes community hospitals, regional long‐term acute care hospitals, nursing homes, and chronic wound, dialysis, and infusion clinics. This may explain, in part, the relatively high rate of hospital readmission observed in our cohort. Third, there is the possibility that CDAs were associated with readmission by chance given the number of potential predictor variables examined. The importance of CDAs as a determinant of rehospitalization requires confirmation in other independent populations. Fourth, it is likely that we did not capture all hospital readmissions, primarily those occurring outside of our hospital system. Therefore, we may have underestimated the actual rates of readmission for this cohort. Finally, we cannot be certain that all important predictors of hospital readmission were captured in this study.

The development of an accurate real‐time early warning system has the potential to identify patients at risk for various adverse outcomes including clinical deterioration, hospital death, and postdischarge readmission. By identifying patients at greatest risk for readmission, valuable healthcare resources can be better targeted to such populations. Our findings suggest that existing readmission predictors may suboptimally risk‐stratify patients, and it may be important to include additional clinical variables if pay for performance and other across‐institution comparisons are to be fair to institutions that care for more seriously ill patients. The variables identified as predictors of 30‐day hospital readmission in our study, with the exception of a CDA, are all readily identifiable clinical characteristics. The modest incremental value of a CDA to these clinical characteristics suggests that they would suffice for the identification of patients at high risk for hospital readmission. This is especially important for safety‐net institutions not routinely employing automated CDAs. These safety‐net hospitals provide a disproportionate level of care for patients who otherwise would have difficulty obtaining inpatient medical care and disproportionately carry the greatest burden of hospital readmissions.[15]

Rapid response systems (RRSs) have been developed to identify and treat deteriorating patients on general hospital units.[1] The most commonly proposed approach to the problem of identifying and stabilizing deteriorating hospitalized patients includes some combination of an early warning system to detect the deterioration and an RRS to deal with it. We previously demonstrated that a relatively simple hospital‐specific prediction model employing routine laboratory values and vital sign data is capable of predicting clinical deterioration, the need for intensive care unit (ICU) transfer, and hospital mortality in patients admitted to general medicine units.[2, 3, 4, 5, 6]

Hospital readmissions within 30 days of hospital discharge occur often and are difficult to predict. Starting in 2013, readmission penalties have been applied to specific conditions in the United States (acute myocardial infarction, heart failure, and pneumonia), with the expectation that additional conditions will be added to this group in years to come.[7, 8] Unfortunately, interventions developed to date have not been universally successful in preventing hospital readmissions for various medical conditions and patient types.[9] One potential explanation for this is the inability to reliably predict which patients are at risk for readmission to better target preventative interventions. Predictors of hospital readmission can be disease specific, such as the presence of multivessel disease in patients hospitalized with myocardial infarction,[10] or more general, such as lack of available medical follow‐up postdischarge.[11] Therefore, we performed a study to determine whether the occurrence of automated clinical deterioration alerts (CDAs) predicted 30‐day hospital readmission.

METHODS

RESULTS

The final cohort had 3015 patients with a mean age of 57.5 17.5 years and 47.8% males. The most common reasons for hospital admission were infection or sepsis syndrome including pneumonia and urinary tract infections (23.6%), congestive heart failure or other cardiac conditions (18.4%), respiratory distress including chronic obstructive pulmonary disease (16.2%), acute or chronic renal failure (9.7%), gastrointestinal disorders (8.4%), and diabetes mellitus management (7.4%). Overall, there were 567 (18.8%) patients who were readmitted within 30 days of their hospital discharge date.

Table 2 shows the characteristics of patients readmitted within 30 days and of patients not requiring hospital readmission within 30 days. Patients requiring hospital readmission within 30 days were younger and had significantly more comorbidities as manifested by significantly greater Charlson scores and individual comorbidities including coronary artery disease, congestive heart disease, peripheral vascular disease, connective tissue disease, cirrhosis, diabetes mellitus with end‐organ complications, renal failure, and metastatic cancer. Patients with a 30‐day readmission had significantly longer duration of hospitalization, more emergency department visits in the 6 months prior to the index hospitalization, lower minimum hemoglobin measurements, higher minimum serum creatinine values, and were more likely to have Medicare or Medicaid insurance compared to patients without a 30‐day readmission.

There were 1141 (34.4%) patients that triggered a CDA. Patients triggering a CDA were significantly more likely to have a 30‐day readmission compared to those who did not trigger a CDA (23.6% vs 15.9%; P < 0.001). Patients triggering a CDA were also significantly more likely to be readmitted within 60 days (31.7% vs 22.1%; P < 0.001) and 90 days (35.8% vs 26.2%; P < 0.001) compared to patients who did not trigger a CDA. Multiple logistic regression identified the triggering of a CDA to be independently associated with 30‐day readmission (OR: 1.40; 95% CI: 1.26‐1.55; P = 0.001) (Table 3). Other independent predictors of 30‐day readmission were: an emergency department visit in the previous 6 months, increasing age in 1‐year increments, presence of connective tissue disease, diabetes mellitus with end‐organ complications, chronic renal disease, cirrhosis, and metastatic cancer (Hosmer‐Lemeshow goodness of fit test, 0.363). Figure 1 reveals the ROC curves for the logistic regression model (Table 3) with and without the CDA variable. As the ROC curves document, the 2 models had similar sensitivity for the entire range of specificities. Reflecting this, the area under the ROC curve for the model inclusive of the CDA variable equaled 0.675 (95% CI: 0.649‐0.700), whereas the area under the ROC curve for the model excluding the CDA variable equaled 0.658 (95% CI: 0.632‐0.684).

Figure 1

DISCUSSION

We demonstrated that the occurrence of an automated CDA is associated with increased risk for 30‐day hospital readmission. However, the addition of the CDA variable to the other variables identified to be independently associated with 30‐day readmission (Table 3) did not significantly add to the overall predictive accuracy of the derived logistic regression model. Other investigators have previously attempted to develop automated predictors of hospital readmission. Amarasingham et al. developed a real‐time electronic predictive model that identifies hospitalized heart failure patients at high risk for readmission or death from clinical and nonclinical risk factors present on admission.[13] Their electronic model demonstrated good discrimination for 30‐day mortality and readmission and performed as well, or better than, models developed by the Center for Medicaid and Medicare Services and the Acute Decompensated Heart Failure Registry. Similarly, Baillie et al. developed an automated prediction model that was effectively integrated into an existing EMR and identified patients on admission who were at risk for readmission within 30 days of discharge.[14] Our automated CDA differs from these previous risk predictors by surveying patients throughout their hospital stay as opposed to identifying risk for readmission at a single time point.

Several limitations of our study should be recognized. First, this was a noninterventional study aimed at examining the ability of CDAs to predict hospital readmission. Future studies are needed to assess whether the use of enhanced readmission prediction algorithms can be utilized to avert hospital readmissions. Second, the data derive from a single center, and this necessarily limits the generalizability of our findings. As such, our results may not reflect what one might see at other institutions. For example, Barnes‐Jewish Hospital has a regional referral pattern that includes community hospitals, regional long‐term acute care hospitals, nursing homes, and chronic wound, dialysis, and infusion clinics. This may explain, in part, the relatively high rate of hospital readmission observed in our cohort. Third, there is the possibility that CDAs were associated with readmission by chance given the number of potential predictor variables examined. The importance of CDAs as a determinant of rehospitalization requires confirmation in other independent populations. Fourth, it is likely that we did not capture all hospital readmissions, primarily those occurring outside of our hospital system. Therefore, we may have underestimated the actual rates of readmission for this cohort. Finally, we cannot be certain that all important predictors of hospital readmission were captured in this study.

The development of an accurate real‐time early warning system has the potential to identify patients at risk for various adverse outcomes including clinical deterioration, hospital death, and postdischarge readmission. By identifying patients at greatest risk for readmission, valuable healthcare resources can be better targeted to such populations. Our findings suggest that existing readmission predictors may suboptimally risk‐stratify patients, and it may be important to include additional clinical variables if pay for performance and other across‐institution comparisons are to be fair to institutions that care for more seriously ill patients. The variables identified as predictors of 30‐day hospital readmission in our study, with the exception of a CDA, are all readily identifiable clinical characteristics. The modest incremental value of a CDA to these clinical characteristics suggests that they would suffice for the identification of patients at high risk for hospital readmission. This is especially important for safety‐net institutions not routinely employing automated CDAs. These safety‐net hospitals provide a disproportionate level of care for patients who otherwise would have difficulty obtaining inpatient medical care and disproportionately carry the greatest burden of hospital readmissions.[15]

Geriatric syndromes are common clinical conditions in older adults that do not fall into specific disease categories. Unlike the traditional definition of a syndrome, geriatric syndrome refers to a condition that is mediated by multiple shared underlying risk factors.[1, 2] Conditions commonly referred to as geriatric syndromes include delirium, cognitive impairment, falls, unintentional weight loss, depressive symptoms, and incontinence. Even though many perceive it as medical misnomer,[3] geriatric syndromes have been shown to negatively impact quality of life and activities of daily living in older adults.[2] They are also associated with adverse outcomes such as increased healthcare utilization, functional decline, and mortality, even after adjusting for age and disease severity.[4, 5, 6] Hospitalized older adults, including those discharged to skilled nursing facilities (SNFs)[7, 8] are particularly at high risk for new‐onset or exacerbation of geriatric syndromes and poor outcomes.[7, 9, 10] However, hospital providers seldom assess, manage, or document geriatric syndromes because they are often overshadowed by disease conditions that lead to an acute episode requiring hospitalization (e.g., heart disease).[11]

Pharmacotherapy is the cornerstone of hospital treatment, and it is well‐known that it affects multiple physiologic systems causing side effects apart from the condition they are approved to treat. Given that geriatric syndromes are a result of impairments in multiple organ systems, it is plausible that pharmacotherapy may initiate or worsen these syndromes.[12] Medication‐related problems in older adults are well known. Polypharmacy and adverse drug events (as a result of drug‐drug/disease interactions and changes in pharmacokinetics and pharmacodynamics) are prevalent in multimorbid elderly patients.[13, 14, 15, 16] The prescribing cascade[17] increases the medication burden and may be a contributing factor for geriatric syndromes in hospitalized patients.[18] For instance, laxatives may be prescribed to counteract constipation caused by anticholinergic drugs.

The American Geriatric Society (AGS) Beers list[19, 20] and similar criteria[21] provide excellent resources to identify medications with potentially harmful interactions or adverse effects in older adults. Although these lists include medicines associated with a specific geriatric syndrome, they were not developed to explicitly link medicines across multiple geriatric syndromes, regardless of indication or appropriateness. For example, medications that effect important geriatric syndromes like unintentional weight/appetite loss, depression, and urinary incontinence are not extensively covered. In addition, disease‐appropriate medications (eg, ‐blockers for systolic heart failure), that may be associated with a geriatric syndrome (eg, falls) are not included; however, they may be important to consider for a patient and clinician who are weighing the disease benefits compared to the geriatric syndrome‐related risks. Finally, the AGS 2015 Beers criteria panel mentions the limitation that many medication associations may be excluded because older adults are less represented in clinical trials.[20] Clinicians are currently limited in identifying medications potentially contributing to a broad set of geriatric syndromes in their patients without a specific list of medications associated with geriatric syndromes (MAGS).[20]

In response to this gap, identifying these medications is important and should be a starting point in efforts toward prevention and treatment of geriatric syndromes. The 2 main objectives of this study were to first identify medications that may meaningfully contribute to 6 geriatric syndromes and subsequently describe the frequency of these medications in a population transitioning from acute care to postacute care to highlight the need and potential impact of such a list.

METHODS

This study included 2 phases that aligned with our 2 primary objectives. Phase 1 involved identifying medications associated with 6 geriatric syndromes, and phase 2 included a cross‐sectional analysis of the prevalence of these medications in a sample of patients discharged to SNFs.

Phase 1: Development of the MAGS List

Figure 1 depicts the underlying conceptual model and approach that was used in phase 1. The interaction between the patient factors and medication leads to polypharmacy that contributes to geriatric syndromes and additional adverse outcomes. As a starting point for mitigating geriatric syndromes, we used an iterative analytical process to identify a list of medications associated with the following geriatric syndromes that were documented to be highly prevalent in patients discharged to SNFs: cognitive impairment, delirium, falls, unintentional weight and/or appetite loss, urinary incontinence, and depression.[8] To be inclusive and sensitive, our approach differed from traditional systematic reviews, and in fact was meant to bring together much of the established systematic literature about disparate geriatric syndromes in 1 place, because patients often do not experience a geriatric syndrome in isolation, but rather experience multiple geriatric syndromes.[8] The MAGS list had 3 main inclusion criteria (Figure 1): (1) evidence in the published literature (systematic reviews, cohort studies, randomized clinical trials) that the medication is related to the syndrome, (2) expert panel opinion, and (3) drug databases (Lexicomp Online database[22] and/or US Food and Drug Administration [FDA]approved package inserts).[23] We generated an initial list of medications based on these 3 main criteria to identify medications with significant associations to each geriatric syndrome. The list was further expanded and vetted using an iterative review of each medication list as it related to each geriatric syndrome through a series of group meetings focused around each geriatric syndrome. Following further discussion, we obtained agreement among all team members for medications included in the final list. For each geriatric syndrome, we excluded medications from consideration if they were used to treat the same geriatric syndrome (eg, ‐adrenergic blockers used to treat incontinence in men were listed as associated with incontinence only in women). We classified medications according to the Established Pharmacologic Class available at the FDA website. We also compared our final MAGS list with the 2015 AGS Beer's list[20] by identifying medications that were related to the 6 geriatric syndromes. This included Beers[20]‐cited rationale of anticholinergic, extrapyramidal symptoms, orthostatic hypotension (eg, falls), high‐risk adverse central nervous system effects, sedating, cognitive decline (eg, antipsychotics), delirium, falls, fractures, incontinence, and gastrointestinal (eg, nausea, vomiting). Specifically, we assessed whether the medications were included as inappropriate by the AGS Beers 2015[20] list and also whether they documented the syndrome association for that medication.

Figure 1

RESULTS

DISCUSSION

An iterative process of evidence review by a multidisciplinary panel resulted in a list of 513 medications associated with 6 common geriatric syndromes. This analysis demonstrated that hospitalized, older patients discharged to SNFs were frequently prescribed MAGS. The rate of prescribing ranged from 100% of patients with a medication associated with falls to 36% for unintentional weight loss. Moreover, an alarming 43% of all medications at hospital discharge were MAGS. For this vulnerable population, the combination of high prevalence of MAGS and high risk of geriatric syndromes emphasize a need to critically review the risks and benefits of MAGS throughout hospitalization and at the time of discharge.

A body of evidence demonstrates that many drugs in a typical older adult regimen have no specific clinical indication, are considered inappropriate, or have uncertain efficacy in the geriatric population.[24, 25, 26] This study builds on the foundational work described in landmark reviews such as the AGS Beers[20] and STOPP/START[21] (Screening Tool of Older Persons' Potentially Inappropriate Prescriptions/Screening Tool to Alert doctors to Right, i.e. appropriate indicated Treatment) criteria. Both of these tools, however, were specifically designed as screening tools to identify medications considered unsafe for older adults under most circumstances and within specific illness states.[19, 20, 21] They are most often utilized when starting a medication to avoid acute adverse events. In contrast, the MAGS list was developed to be inclusive of medications that are often appropriate for many medical diagnoses but may also contribute to underlying geriatric syndromes that are more chronic in nature. In addition, inclusion of such medicines increases the sensitivity of screening for medications that can be targeted through patient‐centered deprescribing efforts when clinically appropriate.

A major strength of this study is that we bring together evidence across a spectrum of geriatric syndromes commonly experienced by hospitalized elders. In addition to evaluating multiple syndromes, we applied multiple modalities; particularly the use of an iterative review process by a multidisciplinary team of experts and using Lexicomp and FDA insert packages for linking medications to specific geriatric conditions. The inclusion criteria were broadened beyond single sources of evidence in an effort to capture a comprehensive list of medications. As a result, the MAGS list can be implemented as a screening tool for deprescribing interventions and assessing medication appropriateness to address individual or clusters of geriatric syndromes within a patient.

In addition to expanding this knowledge base, clinical relevance of the MAGS list is highlighted by its application to a sample of hospitalized older adults discharged to SNFs, a cohort known to experience geriatric syndromes. In fact, 43% of patients' medications at hospital discharge were MAGS. Importantly, due to the cross‐sectional nature of this study, we cannot be certain if the medication caused or potentiated each of the geriatric syndromes. However, hospitals and SNFs are devoting major resources toward reduction of falls, avoidance of urinary catheter use, and reduction of preventable readmissions. These efforts can be complemented by considering the number of medications associated with falls, urinary incontinence, and overall MAGS burden. The striking prevalence of MAGS demonstrates a rigorous need to weigh the risks and benefits of these medications. Above all, the intent of this study is not to propose that any MAGS be reflexively stopped, but rather that the MAGS list should facilitate a holistic approach to care for the complex older adult. For example, standard therapies such as gabapentin may be appropriate for treating neuralgic pain but may also contribute to falls and urinary incontinence. Thus, alternative pain treatments could be selected in place of gabapentin for a 75‐year old patient who is experiencing recurrent falls and increasing incontinence. Therefore, the MAGS list enables a patient‐provider discussion wherein medications' therapeutic benefits can be weighed against risks posed by specific clusters of geriatric syndromes, potential impact on quality of life, and consistency with goals of care.

This study has some limitations. First, although we examined a broad number of geriatric syndromes, several other geriatric syndromes experienced by hospitalized older adults were not addressed including: fecal incontinence, insomnia, and functional impairment. These syndromes were intentionally excluded from the study a priori due to reasons of feasibility and scope. Second, unlike the Beer's 2015 criteria, the MAGS list does not sub‐classify associations of medications with geriatric syndromes for patients with specific diseases (eg, heart failure). In fact, our MAGS list included medications often indicated in treating these diagnoses. A clinician must work with the patient to weigh the disease‐specific benefits of some medications with the potential effect on geriatric syndrome symptoms and outcomes. Third, the instrument has a very high sensitivity, which was intended to generate an inclusive list of medications that enables providers to weigh risks of geriatric syndromes with the intended indication benefit. The objective is not to use this list as a reflexive tool but rather help clinicians identify a starting point to address geriatric syndromes in their patients to make patient‐centered medication decisions. Although the MAGS list is intentionally large (sensitive), the advent of advanced bioinformatics can enable MAGS to be assessed in the future for both clinical and research purposes. Fourth, FDA insert packages and Lexicomp databases report anything experienced by the patient while on the particular medication, but it might not necessarily imply a causative link. The high use of MAGS and the specific geriatric syndrome may coexist due to the high prevalence and interplay of multimorbidity, polypharmacy, and geriatric syndromes in this population. Last, the list was developed by expert panel members predominantly from a single institution, which may introduce bias. Despite these limitations, the prevalence of these medications in a sample of patients transitioning from acute to postacute care highlights the utility of the MAGS list in future clinical research and quality improvement endeavors.

In conclusion, the MAGS list provides a comprehensive and sensitive indicator of medications associated with any of 6 geriatric syndromes regardless of medication indication and appropriateness. The MAGS list provides an overall degree of medication burden with respect to geriatric syndromes and a foundation for future research to assess the relationship between the presence of geriatric syndromes and syndrome‐associated medications. The MAGS list is an important first step in summarizing the data that link medications to geriatric syndromes. Future studies are needed to broaden the analysis of MAGS for other common geriatric syndromes and to identify new and emerging medications not present during the time of this analysis. The MAGS list has the potential to facilitate deprescribing efforts needed to combat the epidemic of overprescribing that may be contributing to the burden of geriatric syndromes among older patients.

Geriatric syndromes are common clinical conditions in older adults that do not fall into specific disease categories. Unlike the traditional definition of a syndrome, geriatric syndrome refers to a condition that is mediated by multiple shared underlying risk factors.[1, 2] Conditions commonly referred to as geriatric syndromes include delirium, cognitive impairment, falls, unintentional weight loss, depressive symptoms, and incontinence. Even though many perceive it as medical misnomer,[3] geriatric syndromes have been shown to negatively impact quality of life and activities of daily living in older adults.[2] They are also associated with adverse outcomes such as increased healthcare utilization, functional decline, and mortality, even after adjusting for age and disease severity.[4, 5, 6] Hospitalized older adults, including those discharged to skilled nursing facilities (SNFs)[7, 8] are particularly at high risk for new‐onset or exacerbation of geriatric syndromes and poor outcomes.[7, 9, 10] However, hospital providers seldom assess, manage, or document geriatric syndromes because they are often overshadowed by disease conditions that lead to an acute episode requiring hospitalization (e.g., heart disease).[11]

Pharmacotherapy is the cornerstone of hospital treatment, and it is well‐known that it affects multiple physiologic systems causing side effects apart from the condition they are approved to treat. Given that geriatric syndromes are a result of impairments in multiple organ systems, it is plausible that pharmacotherapy may initiate or worsen these syndromes.[12] Medication‐related problems in older adults are well known. Polypharmacy and adverse drug events (as a result of drug‐drug/disease interactions and changes in pharmacokinetics and pharmacodynamics) are prevalent in multimorbid elderly patients.[13, 14, 15, 16] The prescribing cascade[17] increases the medication burden and may be a contributing factor for geriatric syndromes in hospitalized patients.[18] For instance, laxatives may be prescribed to counteract constipation caused by anticholinergic drugs.

The American Geriatric Society (AGS) Beers list[19, 20] and similar criteria[21] provide excellent resources to identify medications with potentially harmful interactions or adverse effects in older adults. Although these lists include medicines associated with a specific geriatric syndrome, they were not developed to explicitly link medicines across multiple geriatric syndromes, regardless of indication or appropriateness. For example, medications that effect important geriatric syndromes like unintentional weight/appetite loss, depression, and urinary incontinence are not extensively covered. In addition, disease‐appropriate medications (eg, ‐blockers for systolic heart failure), that may be associated with a geriatric syndrome (eg, falls) are not included; however, they may be important to consider for a patient and clinician who are weighing the disease benefits compared to the geriatric syndrome‐related risks. Finally, the AGS 2015 Beers criteria panel mentions the limitation that many medication associations may be excluded because older adults are less represented in clinical trials.[20] Clinicians are currently limited in identifying medications potentially contributing to a broad set of geriatric syndromes in their patients without a specific list of medications associated with geriatric syndromes (MAGS).[20]

In response to this gap, identifying these medications is important and should be a starting point in efforts toward prevention and treatment of geriatric syndromes. The 2 main objectives of this study were to first identify medications that may meaningfully contribute to 6 geriatric syndromes and subsequently describe the frequency of these medications in a population transitioning from acute care to postacute care to highlight the need and potential impact of such a list.

METHODS

This study included 2 phases that aligned with our 2 primary objectives. Phase 1 involved identifying medications associated with 6 geriatric syndromes, and phase 2 included a cross‐sectional analysis of the prevalence of these medications in a sample of patients discharged to SNFs.

Phase 1: Development of the MAGS List

Figure 1 depicts the underlying conceptual model and approach that was used in phase 1. The interaction between the patient factors and medication leads to polypharmacy that contributes to geriatric syndromes and additional adverse outcomes. As a starting point for mitigating geriatric syndromes, we used an iterative analytical process to identify a list of medications associated with the following geriatric syndromes that were documented to be highly prevalent in patients discharged to SNFs: cognitive impairment, delirium, falls, unintentional weight and/or appetite loss, urinary incontinence, and depression.[8] To be inclusive and sensitive, our approach differed from traditional systematic reviews, and in fact was meant to bring together much of the established systematic literature about disparate geriatric syndromes in 1 place, because patients often do not experience a geriatric syndrome in isolation, but rather experience multiple geriatric syndromes.[8] The MAGS list had 3 main inclusion criteria (Figure 1): (1) evidence in the published literature (systematic reviews, cohort studies, randomized clinical trials) that the medication is related to the syndrome, (2) expert panel opinion, and (3) drug databases (Lexicomp Online database[22] and/or US Food and Drug Administration [FDA]approved package inserts).[23] We generated an initial list of medications based on these 3 main criteria to identify medications with significant associations to each geriatric syndrome. The list was further expanded and vetted using an iterative review of each medication list as it related to each geriatric syndrome through a series of group meetings focused around each geriatric syndrome. Following further discussion, we obtained agreement among all team members for medications included in the final list. For each geriatric syndrome, we excluded medications from consideration if they were used to treat the same geriatric syndrome (eg, ‐adrenergic blockers used to treat incontinence in men were listed as associated with incontinence only in women). We classified medications according to the Established Pharmacologic Class available at the FDA website. We also compared our final MAGS list with the 2015 AGS Beer's list[20] by identifying medications that were related to the 6 geriatric syndromes. This included Beers[20]‐cited rationale of anticholinergic, extrapyramidal symptoms, orthostatic hypotension (eg, falls), high‐risk adverse central nervous system effects, sedating, cognitive decline (eg, antipsychotics), delirium, falls, fractures, incontinence, and gastrointestinal (eg, nausea, vomiting). Specifically, we assessed whether the medications were included as inappropriate by the AGS Beers 2015[20] list and also whether they documented the syndrome association for that medication.

Figure 1

RESULTS

DISCUSSION

An iterative process of evidence review by a multidisciplinary panel resulted in a list of 513 medications associated with 6 common geriatric syndromes. This analysis demonstrated that hospitalized, older patients discharged to SNFs were frequently prescribed MAGS. The rate of prescribing ranged from 100% of patients with a medication associated with falls to 36% for unintentional weight loss. Moreover, an alarming 43% of all medications at hospital discharge were MAGS. For this vulnerable population, the combination of high prevalence of MAGS and high risk of geriatric syndromes emphasize a need to critically review the risks and benefits of MAGS throughout hospitalization and at the time of discharge.

A body of evidence demonstrates that many drugs in a typical older adult regimen have no specific clinical indication, are considered inappropriate, or have uncertain efficacy in the geriatric population.[24, 25, 26] This study builds on the foundational work described in landmark reviews such as the AGS Beers[20] and STOPP/START[21] (Screening Tool of Older Persons' Potentially Inappropriate Prescriptions/Screening Tool to Alert doctors to Right, i.e. appropriate indicated Treatment) criteria. Both of these tools, however, were specifically designed as screening tools to identify medications considered unsafe for older adults under most circumstances and within specific illness states.[19, 20, 21] They are most often utilized when starting a medication to avoid acute adverse events. In contrast, the MAGS list was developed to be inclusive of medications that are often appropriate for many medical diagnoses but may also contribute to underlying geriatric syndromes that are more chronic in nature. In addition, inclusion of such medicines increases the sensitivity of screening for medications that can be targeted through patient‐centered deprescribing efforts when clinically appropriate.

A major strength of this study is that we bring together evidence across a spectrum of geriatric syndromes commonly experienced by hospitalized elders. In addition to evaluating multiple syndromes, we applied multiple modalities; particularly the use of an iterative review process by a multidisciplinary team of experts and using Lexicomp and FDA insert packages for linking medications to specific geriatric conditions. The inclusion criteria were broadened beyond single sources of evidence in an effort to capture a comprehensive list of medications. As a result, the MAGS list can be implemented as a screening tool for deprescribing interventions and assessing medication appropriateness to address individual or clusters of geriatric syndromes within a patient.

In addition to expanding this knowledge base, clinical relevance of the MAGS list is highlighted by its application to a sample of hospitalized older adults discharged to SNFs, a cohort known to experience geriatric syndromes. In fact, 43% of patients' medications at hospital discharge were MAGS. Importantly, due to the cross‐sectional nature of this study, we cannot be certain if the medication caused or potentiated each of the geriatric syndromes. However, hospitals and SNFs are devoting major resources toward reduction of falls, avoidance of urinary catheter use, and reduction of preventable readmissions. These efforts can be complemented by considering the number of medications associated with falls, urinary incontinence, and overall MAGS burden. The striking prevalence of MAGS demonstrates a rigorous need to weigh the risks and benefits of these medications. Above all, the intent of this study is not to propose that any MAGS be reflexively stopped, but rather that the MAGS list should facilitate a holistic approach to care for the complex older adult. For example, standard therapies such as gabapentin may be appropriate for treating neuralgic pain but may also contribute to falls and urinary incontinence. Thus, alternative pain treatments could be selected in place of gabapentin for a 75‐year old patient who is experiencing recurrent falls and increasing incontinence. Therefore, the MAGS list enables a patient‐provider discussion wherein medications' therapeutic benefits can be weighed against risks posed by specific clusters of geriatric syndromes, potential impact on quality of life, and consistency with goals of care.

This study has some limitations. First, although we examined a broad number of geriatric syndromes, several other geriatric syndromes experienced by hospitalized older adults were not addressed including: fecal incontinence, insomnia, and functional impairment. These syndromes were intentionally excluded from the study a priori due to reasons of feasibility and scope. Second, unlike the Beer's 2015 criteria, the MAGS list does not sub‐classify associations of medications with geriatric syndromes for patients with specific diseases (eg, heart failure). In fact, our MAGS list included medications often indicated in treating these diagnoses. A clinician must work with the patient to weigh the disease‐specific benefits of some medications with the potential effect on geriatric syndrome symptoms and outcomes. Third, the instrument has a very high sensitivity, which was intended to generate an inclusive list of medications that enables providers to weigh risks of geriatric syndromes with the intended indication benefit. The objective is not to use this list as a reflexive tool but rather help clinicians identify a starting point to address geriatric syndromes in their patients to make patient‐centered medication decisions. Although the MAGS list is intentionally large (sensitive), the advent of advanced bioinformatics can enable MAGS to be assessed in the future for both clinical and research purposes. Fourth, FDA insert packages and Lexicomp databases report anything experienced by the patient while on the particular medication, but it might not necessarily imply a causative link. The high use of MAGS and the specific geriatric syndrome may coexist due to the high prevalence and interplay of multimorbidity, polypharmacy, and geriatric syndromes in this population. Last, the list was developed by expert panel members predominantly from a single institution, which may introduce bias. Despite these limitations, the prevalence of these medications in a sample of patients transitioning from acute to postacute care highlights the utility of the MAGS list in future clinical research and quality improvement endeavors.

In conclusion, the MAGS list provides a comprehensive and sensitive indicator of medications associated with any of 6 geriatric syndromes regardless of medication indication and appropriateness. The MAGS list provides an overall degree of medication burden with respect to geriatric syndromes and a foundation for future research to assess the relationship between the presence of geriatric syndromes and syndrome‐associated medications. The MAGS list is an important first step in summarizing the data that link medications to geriatric syndromes. Future studies are needed to broaden the analysis of MAGS for other common geriatric syndromes and to identify new and emerging medications not present during the time of this analysis. The MAGS list has the potential to facilitate deprescribing efforts needed to combat the epidemic of overprescribing that may be contributing to the burden of geriatric syndromes among older patients.

Geriatric syndromes are common clinical conditions in older adults that do not fall into specific disease categories. Unlike the traditional definition of a syndrome, geriatric syndrome refers to a condition that is mediated by multiple shared underlying risk factors.[1, 2] Conditions commonly referred to as geriatric syndromes include delirium, cognitive impairment, falls, unintentional weight loss, depressive symptoms, and incontinence. Even though many perceive it as medical misnomer,[3] geriatric syndromes have been shown to negatively impact quality of life and activities of daily living in older adults.[2] They are also associated with adverse outcomes such as increased healthcare utilization, functional decline, and mortality, even after adjusting for age and disease severity.[4, 5, 6] Hospitalized older adults, including those discharged to skilled nursing facilities (SNFs)[7, 8] are particularly at high risk for new‐onset or exacerbation of geriatric syndromes and poor outcomes.[7, 9, 10] However, hospital providers seldom assess, manage, or document geriatric syndromes because they are often overshadowed by disease conditions that lead to an acute episode requiring hospitalization (e.g., heart disease).[11]

Pharmacotherapy is the cornerstone of hospital treatment, and it is well‐known that it affects multiple physiologic systems causing side effects apart from the condition they are approved to treat. Given that geriatric syndromes are a result of impairments in multiple organ systems, it is plausible that pharmacotherapy may initiate or worsen these syndromes.[12] Medication‐related problems in older adults are well known. Polypharmacy and adverse drug events (as a result of drug‐drug/disease interactions and changes in pharmacokinetics and pharmacodynamics) are prevalent in multimorbid elderly patients.[13, 14, 15, 16] The prescribing cascade[17] increases the medication burden and may be a contributing factor for geriatric syndromes in hospitalized patients.[18] For instance, laxatives may be prescribed to counteract constipation caused by anticholinergic drugs.

The American Geriatric Society (AGS) Beers list[19, 20] and similar criteria[21] provide excellent resources to identify medications with potentially harmful interactions or adverse effects in older adults. Although these lists include medicines associated with a specific geriatric syndrome, they were not developed to explicitly link medicines across multiple geriatric syndromes, regardless of indication or appropriateness. For example, medications that effect important geriatric syndromes like unintentional weight/appetite loss, depression, and urinary incontinence are not extensively covered. In addition, disease‐appropriate medications (eg, ‐blockers for systolic heart failure), that may be associated with a geriatric syndrome (eg, falls) are not included; however, they may be important to consider for a patient and clinician who are weighing the disease benefits compared to the geriatric syndrome‐related risks. Finally, the AGS 2015 Beers criteria panel mentions the limitation that many medication associations may be excluded because older adults are less represented in clinical trials.[20] Clinicians are currently limited in identifying medications potentially contributing to a broad set of geriatric syndromes in their patients without a specific list of medications associated with geriatric syndromes (MAGS).[20]

In response to this gap, identifying these medications is important and should be a starting point in efforts toward prevention and treatment of geriatric syndromes. The 2 main objectives of this study were to first identify medications that may meaningfully contribute to 6 geriatric syndromes and subsequently describe the frequency of these medications in a population transitioning from acute care to postacute care to highlight the need and potential impact of such a list.

METHODS

This study included 2 phases that aligned with our 2 primary objectives. Phase 1 involved identifying medications associated with 6 geriatric syndromes, and phase 2 included a cross‐sectional analysis of the prevalence of these medications in a sample of patients discharged to SNFs.

Phase 1: Development of the MAGS List

Figure 1 depicts the underlying conceptual model and approach that was used in phase 1. The interaction between the patient factors and medication leads to polypharmacy that contributes to geriatric syndromes and additional adverse outcomes. As a starting point for mitigating geriatric syndromes, we used an iterative analytical process to identify a list of medications associated with the following geriatric syndromes that were documented to be highly prevalent in patients discharged to SNFs: cognitive impairment, delirium, falls, unintentional weight and/or appetite loss, urinary incontinence, and depression.[8] To be inclusive and sensitive, our approach differed from traditional systematic reviews, and in fact was meant to bring together much of the established systematic literature about disparate geriatric syndromes in 1 place, because patients often do not experience a geriatric syndrome in isolation, but rather experience multiple geriatric syndromes.[8] The MAGS list had 3 main inclusion criteria (Figure 1): (1) evidence in the published literature (systematic reviews, cohort studies, randomized clinical trials) that the medication is related to the syndrome, (2) expert panel opinion, and (3) drug databases (Lexicomp Online database[22] and/or US Food and Drug Administration [FDA]approved package inserts).[23] We generated an initial list of medications based on these 3 main criteria to identify medications with significant associations to each geriatric syndrome. The list was further expanded and vetted using an iterative review of each medication list as it related to each geriatric syndrome through a series of group meetings focused around each geriatric syndrome. Following further discussion, we obtained agreement among all team members for medications included in the final list. For each geriatric syndrome, we excluded medications from consideration if they were used to treat the same geriatric syndrome (eg, ‐adrenergic blockers used to treat incontinence in men were listed as associated with incontinence only in women). We classified medications according to the Established Pharmacologic Class available at the FDA website. We also compared our final MAGS list with the 2015 AGS Beer's list[20] by identifying medications that were related to the 6 geriatric syndromes. This included Beers[20]‐cited rationale of anticholinergic, extrapyramidal symptoms, orthostatic hypotension (eg, falls), high‐risk adverse central nervous system effects, sedating, cognitive decline (eg, antipsychotics), delirium, falls, fractures, incontinence, and gastrointestinal (eg, nausea, vomiting). Specifically, we assessed whether the medications were included as inappropriate by the AGS Beers 2015[20] list and also whether they documented the syndrome association for that medication.

Figure 1

RESULTS