Author + information
- Robb D. Kociol, MD∗∗ (, )
- Bradley G. Hammill, MS†,
- Gregg C. Fonarow, MD‡,
- Paul A. Heidenreich, MD§,
- Alan S. Go, MD||,¶,
- Eric D. Peterson, MD, MPH∗,#,
- Lesley H. Curtis, PhD∗,# and
- Adrian F. Hernandez, MD, MHS∗,#
- ∗CardioVascular Institute, Beth Israel Deaconess Medical Center, Harvard Medical School, Boston, Massachusetts
- †Duke Clinical Research Institute, Duke University School of Medicine, Durham, North Carolina
- ‡Ahmanson-UCLA Cardiomyopathy Center, Los Angeles, California
- §Division of Cardiology, Department of Medicine, Palo Alto Veterans Affairs Medical Center, Stanford University School of Medicine, Palo Alto, California
- ||Division of Research, Kaiser Permanente Northern California, Oakland, California
- ¶Departments of Epidemiology and Biostatistics and Medicine, University of California, San Francisco
- #Department of Medicine, Duke University School of Medicine, Durham, North Carolina
- ↵∗Reprint requests and correspondence:
Dr. Robb D. Kociol, Beth Israel Deaconness Medical Center, 471 Brookline Avenue, Boston, Massachusetts 02215.
Objectives This study sought to examine the associations of hospitalist and cardiologist care of patients with heart failure with outcomes and adherence to quality measures.
Background The hospitalist model of inpatient care has grown nationally, but its associations with quality of care and outcomes of patients hospitalized with heart failure are not known.
Methods We analyzed data from the Get With the Guidelines-Heart Failure registry linked to Medicare claims for 2005 through 2008. For each hospital, we calculated the percentage of heart failure hospitalizations for which a hospitalist was the attending physician. We examined outcomes and care quality for patients stratified by rates of hospitalist use. Using multivariable models, we estimated associations between hospital-level use of hospitalists and cardiologists and 30-day risk-adjusted outcomes and adherence to measures of quality care.
Results The analysis included 31,505 Medicare beneficiaries in 166 hospitals. Across hospitals, the use of hospitalists varied from 0% to 83%. After multivariable adjustment, a 10% increase in the use of hospitalists was associated with a slight increase in mortality (risk ratio: 1.03; 95% confidence interval [CI]: 1.00 to 1.06) and decrease in length of stay (0.09 days; 95% CI: 0.02 to 0.16). There was no association with 30-day readmission. Increased use of hospitalists in hospitals with high use of cardiologists was associated with improved defect-free adherence to a composite of heart failure performance measures (risk ratio: 1.03; 95% CI: 1.01 to 1.06).
Conclusions Hospitalist care varied significantly across hospitals for heart failure admissions and was not associated with improved 30-day outcomes. Comanagement by hospitalists and cardiologists may help to improve adherence to some quality measures, but it remains unclear what care model improves 30-day clinical outcomes.
The fee-for-service Medicare program provides a financial incentive to hospitals to reduce patients’ length of stay for heart failure, without penalty for increases in readmission or mortality. In response, many hospitals have increased their reliance on hospitalists to provide inpatient care with the goal of decreasing hospital length of stay and costs while maintaining similar levels of reimbursement (1,2). Hospitalists care for a wide range of general medical patients and increasingly for patients traditionally cared for by subspecialists. The use of hospitalists has grown exponentially in the United States (3,4).
This model of care evolved in the mid-1990s in response to increasing cost pressures on hospitals and provider groups, increasing outpatient care responsibilities among primary care physicians, and decreasing inpatient volume along with greater inpatient acuity (2). The initial experience in the mid-1990s among large health systems involved transitioning to hospital coverage for 24 hours, 7 days a week (1,2,5). Most health systems now use a voluntary hospitalist system in which primary physicians can choose to admit to a hospitalist service or attend to their own patients (2).
The hospitalist model of care is directly aligned with the hospital or health care system. Increased specialization in inpatient medicine, improved in-house availability, and participation in hospital quality initiatives are hypothesized to drive reported associations with reduced length of stay and associated cost savings for conditions such as chest pain, pneumonia, and heart failure (1,2,6–8). Although previous studies have found that hospitalists have greater adherence to inpatient quality measures for acute myocardial infarction and pneumonia, results vary for heart failure (9,10). Cardiologist care, compared with generalist care, has been associated with improved adherence to evidence-based therapies and improved outcomes (11,12). It is unclear whether hospitalist care is associated with improvements in 30-day outcomes among patients with heart failure, who frequently have cardiac and noncardiac comorbid conditions (10–16). Moreover, it is unclear to what extent hospitals rely exclusively on hospitalists for the care of patients with heart failure and how the use of cardiologists in combination with hospitalists is associated with outcomes. Previous studies have compared hospitalist care to generalist care, but it is unknown whether the use of cardiologists influences associations between hospitalist care and patient outcomes (10,15).
We linked data from the American Heart Association’s Get With the Guidelines-Heart Failure registry to Medicare claims to examine associations of hospitalist and cardiologist care of patients with heart failure with outcomes and adherence to quality measures.
We obtained data on patient characteristics and admissions from the Get With the Guidelines-Heart Failure registry (17). The registry is a hospital-based quality-improvement program that collects data on patients with new or worsening heart failure who are admitted to participating hospitals. It is a continuation of the Organized Program to Initiate Lifesaving Treatment in Hospitalized Patients With Heart Failure registry, which has been described previously (18). Only data from registry sites that routinely reported the results of admission examinations and laboratory tests were used for the study. We considered discharge dates between January 1, 2005, and November 30, 2008, for the analysis.
Patient-level outcomes were from a 100% sample of Medicare inpatient claims and the associated denominator files. To identify registry patients in the Medicare files, we linked registry hospitalizations to Medicare inpatient claims using a method that identified records in both data sources using hospital service dates, patient sex, and patient date of birth (19). For the beneficiaries represented in these linked hospitalizations, we gathered all records from the 2005 to 2008 Medicare denominator and inpatient files. The inpatient files contain information from claims generated by hospitals for inpatient services. The denominator files contain information on Medicare enrollment and mortality.
We also used the Medicare inpatient files to describe the distribution of attending physicians at hospitals included in the study. We determined physician specialty using claims for professional services for a 5% random sample of Medicare beneficiaries between 2005 and 2008. Among other information, Medicare physician claim files include type of service, place of service, and self-reported specialty.
Finally, hospital characteristics were from the American Hospital Association annual survey files from 2005 through 2008. This survey captures information about specialty services available at each hospital in addition to descriptive information about each hospital.
The study cohort consisted of patients 65 years and older enrolled in fee-for-service Medicare whose registry hospitalizations were identified in the Medicare claims data. For patients with multiple linked hospitalizations, we chose the earliest for the analysis. We limited the cohort to patients who were discharged before November 30, 2008, to ensure that data for 30 days of follow-up were available. We further limited the cohort to patients from hospitals that had at least 25 Medicare-linked registry admissions to allow for stable estimates of the distribution of attending physicians.
Use of hospitalists
The main hospital-level exposure of interest was the use of hospitalists, which we defined as the percentage of heart failure admissions for which a hospitalist was the attending physician. To identify hospitalists, we first determined the specialty of all physicians in each hospital using all claims for inpatient evaluation and management services (Current Procedural Terminology codes 99221-99223, 99231-88233, and 99251-99255) and outpatient evaluation and management services (codes 99201-99205, 99211-99215, and 99241-99245) filed by physicians listed as an attending physician on any inpatient claim from a registry hospital. We then categorized physicians into 1 of 4 groups based on their practice patterns: hospitalists, cardiologists, nonhospitalist generalists, and other specialists. We defined cardiologists as physicians whose evaluation and management claims indicated a cardiology specialty. We defined hospitalists using the algorithm described and validated by Kuo et al. (4). Specifically, physicians were classified as hospitalists if they had 5 or more evaluation and management claims in the Medicare 5% sample for a given year (meaning at least 100 evaluation and management claims annually), if at least 90% of these evaluation and management claims were inpatient claims, and if the specialty on the claims was internal medicine, general medicine, or family medicine. We defined nonhospitalist generalists as physicians with evaluation and management claims indicating a specialty of internal medicine, general medicine, or family medicine but who were not classified as hospitalists. We classified all other physicians as other specialists.
After categorizing physicians according to specialty, we identified all heart failure hospitalization claims (primary ICD-9-CM diagnosis code of 428.x, 402.x1, 404.x1, or 404.x3) at the study hospitals and determined the specialty of the attending physician on each claim. We then summarized the distribution of attending physicians annually by hospital.
There were multiple patient-level outcomes of interest. The 3 main outcomes included 30-day mortality, 30-day readmission, and length of stay. The 30-day period for mortality started on the date of admission and ended on the death date in the Medicare denominator file. Patients discharged to hospice and patients discharged alive within 1 day of admission were not included in the mortality analysis. The 30-day period for readmission started on the discharge date, and the readmission date was determined from subsequent inpatient claims. The readmission analysis did not include patients admitted for rehabilitation, patients transferred out to other short-term hospitals, patients discharged to hospice, or patients who died in hospital. Length of stay was determined as the number of days from the admission date to the discharge date of the index hospital stay. We truncated length of stay at 21 days for very long stays to minimize the influence of outliers. Patients transferred to another short-term hospital were not included in the analysis of length of stay.
We were also interested in 2 defect-free composite scores of heart failure process and quality measures (20). One composite measure summarized the registry achievement measures: angiotensin-converting enzyme inhibitor or angiotensin receptor blocker for left ventricular systolic dysfunction (LVSD) at discharge; β-blocker for LVSD; discharge instructions; measurement of left ventricular function; and smoking cessation counseling. The other composite measure summarized the registry quality measures: aldosterone antagonist therapy for LVSD at discharge; anticoagulation therapy for atrial fibrillation at discharge; evidence-based β-blocker for LVSD at discharge; hydralazine and isosorbide dinitrate combination for African American patients at discharge; implantable cardioverter-defibrillator implanted or prescribed at discharge; and measurement of blood pressure at discharge. The definitions of and eligibility criteria for each of these measures are described in detail elsewhere (21). For both composite measures, patients were scored according to whether they received all measures for which they were eligible. Patients eligible for no measures in a particular set were not included in the analysis of that composite measure.
To describe the hospitals included in the analysis, we used the number of beds, the number of heart failure records in both the registry and the Medicare inpatient files, hospital membership in the Council of Teaching Hospitals, and available cardiac specialty services. Membership in the Council of Teaching Hospitals is granted by the American Association of Medical Colleges, which is affiliated with a Liaison Committee on Medical Education–accredited medical school and typically sponsors at least 4 residency programs (22). Specific services of interest included interventional cardiac catheterization, cardiac intensive care, adult cardiac surgery, and adult heart transplant.
We used patient characteristics from the registry to describe patients in the analysis and for risk adjustment. Specific patient-level information included demographic characteristics (age, sex, and race), comorbid conditions (anemia, atrial fibrillation, chronic obstructive pulmonary disorder, depression, diabetes mellitus, hyperlipidemia, hypertension, peripheral vascular disease, previous stroke, renal insufficiency, smoking status), devices in place at hospital admission (implantable cardioverter-defibrillator, cardiac resynchronization therapy, or permanent pacemaker), heart failure details (ischemic etiology, previous heart failure hospitalizations, and systolic function), and vitals and laboratory test results (hemoglobin, serum creatinine, serum sodium, systolic blood pressure).
We describe the characteristics of hospitals by quartile of hospitalist use. We present frequencies with percentages for categorical variables and medians with interquartile ranges for continuous variables. We tested for differences between quartiles using chi-square tests for categorical variables and Kruskal-Wallis tests for continuous variables.
We describe the characteristics and outcomes of patients by quartile of hospitalist use at the admitting hospital. We present frequencies with percentages for categorical variables and medians with interquartile ranges for continuous variables. We tested for differences between quartiles of hospitalist use using chi-square tests for categorical variables and Kruskal-Wallis tests for continuous variables. For each outcome, we present the number of patients included in the denominator, because eligibility was different for each outcome.
To measure associations between the use of hospitalists and patient-level outcomes, we estimated a linear mixed model for length of stay and a modified Poisson generalized linear mixed model (23) for all other dichotomous outcomes. We selected the Poisson model because the exponentiated parameter estimates approximate the true relative risk, even for common outcomes. All patient-level models included terms for the percentage of hospitalist attending physicians and the percentage of cardiologist attending physicians at the hospital and the year of admission. We tested for an interaction between these 2 terms and retained the interaction only if it was significant at the α = .05 level. We present results from an unadjusted model that contained only the terms for hospitalist and cardiologist attending physicians. We present results from 2 adjusted models. The first adjusted model controlled for the patient characteristics listed above. The second adjusted model also controlled for hospital characteristics. All models included a random intercept by hospital-year to account for patient clustering. We present the association of an absolute 10% increase in the percentage of hospitalist attending physicians for each outcome to aid interpretation of the results. For models in which there was a significant interaction term, we present the association of an absolute 10% increase in hospitalist use at both low and high levels of cardiologist use, and vice versa. Low and high levels of hospitalist and cardiologist use were based on the first and third quartiles of percentages of attending physicians across all hospitals.
We used SAS version 9.2 (SAS Institute Inc, Cary, North Carolina) for all analyses. The institutional review board of the Duke University Health System approved the study.
The study included 166 hospitals representing 326 hospital-years and 31,505 heart failure hospitalizations between 2005 and 2008. Table 1 shows hospital characteristics stratified by quartile of hospitalist use. Hospitals in the lowest quartile had fewer beds, were less often members of the Council of Teaching Hospitals, and had fewer annual median admissions of Medicare beneficiaries for heart failure. Hospitals in the middle 2 quartiles of hospitalist use were larger and had more heart failure admissions. Figure 1 shows the distribution of attending physician types for patients with heart failure. Rates of hospitalist use from the lowest to highest quartile were 0% (0.0% to 1.2%), 7.9% (6.0% to 9.6%), 18.3% (15.1% to 22.3%), and 37.0% (31.4% to 46.3%). The lowest rates of cardiologist use for patients with heart failure occurred in hospitals in the highest quartile of hospitalist use. There was a stepwise decrease in rates of generalist care for patients with heart failure with increasing quartiles of hospitalist use.
Table 2 shows patient characteristics by hospital-level quartiles of hospitalist use. Age and sex were similar across quartiles. A lower proportion of African American patients were admitted to hospitals with the highest rates of hospitalist use. Patients in hospitals with greater hospitalist use were similar with respect to comorbid conditions, with the exception of modestly higher rates of atrial fibrillation, previous stroke, and chronic renal insufficiency. Other differences in comorbid conditions were statistically significant, but the absolute differences were small.
Table 3 shows patient-level outcomes and adherence to established and emerging heart failure performance measures by quartile of hospitalist use. Online Table 1 shows the complete results of the adjusted model for mortality, readmission, and length of stay. Table 4 shows the unadjusted and adjusted associations between hospitalist and cardiologist use and patient outcomes. Rates of 30-day mortality across quartiles of hospitalist use were similar. After adjustment for patient and hospital characteristics, there was a small, statistically significant association between each 10% increase in hospitalist use and 30-day mortality (risk ratio: 1.03; 95% confidence interval [CI]: 1.00 to 1.06; p = 0.02). Hospitals in the highest quartile of hospitalist use had lower 30-day readmission rates (22.9% vs. 24.9%, 24.9%, and 24.4% in the first, second, and third quartiles, respectively; p = 0.02). However, after adjustment for patient and hospital characteristics, use of hospitalists was not associated with 30-day readmission. Every 10% increase in hospitalist use was associated with a reduction in median length of stay of 0.09 days (interquartile range: 0.02 to 0.16).
Adherence to the composite of established heart failure performance measures was lowest in hospitals with the lowest use of hospitalists (74.3% vs. 80.1%, 79.7%, and 76.3% in the second, third, and fourth quartiles, respectively; p < 0.01). In contrast, adherence to the composite of emerging heart failure measures was highest in hospitals with the lowest use of hospitalists (44.2% vs. 41.2%, 41.8%, and 40.7% in the second, third, and fourth quartiles, respectively; p < 0.01). After adjustment for patient and hospital characteristics, hospitalist use was associated with greater adherence to established heart failure performance measures; however, this association was modified by the use of cardiologists. In other words, the use of hospitalists was only associated with greater adherence to established performance measures in hospitals with greater use of cardiologists (risk ratio: 1.03; 95% CI: 1.01 to 1.06). The use of hospitalists was not associated with adherence to emerging heart failure quality measures. Online Table 2 shows associations with each heart failure measure.
Increasing use of cardiologists in the care of patients with heart failure was not associated with 30-day mortality, 30-day readmission, or length of stay after adjustment for patient and hospital characteristics. We observed a small association between greater use of cardiologists and adherence to established heart failure quality measures, but this association was only observed in hospitals with greater use of hospitalists (risk ratio: 1.03; 95% CI: 1.02 to 1.05).
Ours is among the largest studies of hospitalist care for patients with heart failure and its associations with quality of care and clinical outcomes. Overall, the rate of hospitalist use varied significantly across hospitals. The use of hospitalists was associated with shorter length of stay and better adherence to some quality measures. However, 30-day readmission was not associated with the use of hospitalists, and there was a small but statistically significant increase in 30-day mortality associated with greater use of hospitalists. Although these findings require further study, they suggest a need for alternative models of care for patients hospitalized for heart failure, such as comanagement with cardiologists.
Previous studies suggested that the use of hospitalists reduces length of stay and may reduce costs, compared with the use of general practitioners (12,15) A recent study by Goodrich et al. (16) found an association between hospitalist care and improved 30-day readmission rates among patients with heart failure. In contrast, our results do not suggest that hospitalist care is associated with lower rates of short-term mortality or readmission. We observed a modest association between greater hospitalist use and 30-day mortality after adjustment for case mix. Likewise, hospitalist use was associated with a slight decrease in length of stay that may not be meaningful at the health system level.
This study is also 1 of the largest to examine association between the use of hospitalists in heart failure with quality and outcomes while adjusting for the use of cardiologists in the same hospital. Hospitals with greater use of cardiologists in the care of patients with heart failure did not have better short-term outcomes or reduced length of stay. However, the use of hospitalists was associated with greater defect-free adherence to a composite of heart failure performance measures only in hospitals with a high prevalence of cardiologist attending physicians, suggesting a potential synergy between hospitalist and subspecialty care models.
Previous research found a relationship between hospitalist care and adherence to some quality measures, such as disease-specific evidence-based medications and counseling for common conditions (9,10). In contrast, Vasilevskis et al. (24) found that patients with heart failure in the Multicenter Hospitalist Study admitted to hospitalists received similar care quality compared with those admitted to nonhospitalists, although those admitted to hospitalists were more likely to have follow-up within 30 days. We found a relationship between hospitalist use and a composite of heart failure quality measures, but the associations were complex and were modified by cardiologist care. Hospitalist and cardiologist care were associated with greater defect-free adherence to a composite of heart failure performance measures (angiotensin-converting enzyme inhibitor/angiotensin receptor blocker for LVSD, β-blocker for LVSD, assessment of left ventricular function, discharge instructions, smoking cessation counseling). In contrast, neither hospitalist nor cardiologist care was associated with greater adherence to a defect-free composite of achievement measures of heart failure quality of care (20), such as prescription of aldosterone antagonists at discharge and use of specific evidence-based β-blockers.
Some of the observed differences in associations between hospitalist use and outcomes and adherence to quality measures may be related to heterogeneity in hospitalist physicians. Some hospitalists receive dedicated training in hospitalist medicine either during or after residency (2). Others complete an internal medicine or family practice residency and opt to work as hospitalists. Some hospitals hire highly specialized hospitalists, which may result in a discrepancy in the association between hospitalist use and outcomes by hospital type that we cannot fully account for in this study (25,26).
Our findings highlight the need for studies to determine optimal models of care for patients hospitalized with heart failure. The results suggest that comanagement by hospitalists and cardiologists may help to improve adherence to some quality measures, but it remains unclear what care model improves 30-day clinical outcomes. Moreover, the observed variability in the use of hospitalists and cardiologists suggests that considerable uncertainty exists about how best to care for hospitalized patients with heart failure. Ideally, a coordinated approach to hospital care with hospitalists and cardiologists who are well qualified to treat patients with heart failure would translate into better quality of care and improved outcomes.
Because this was a retrospective observational study, there may have been residual or unmeasured confounding, though we adjusted for patient and hospital characteristics, comorbid conditions, and other prognostic variables. In particular, sicker patients may have been cared for by cardiology subspecialists. Moreover, hospitals in the study were participating in the Get With the Guidelines-Heart Failure registry, which may limit the generalizability of the findings to other hospitals. Patients in the study were 65 years or older and enrolled in fee-for-service Medicare, which may also limit the generalizability of our results. We defined hospitalists indirectly on the basis of the percentage of Medicare Part B claims that were for inpatient services. This approach has been validated previously but might have led us to misclassify physician type. Another limitation of this study is that we could not identify the degree to which cardiologist input in the form of consultation mediated the observed outcomes. Further study will be needed to determine how formal cardiology consultation affects outcomes related to hospitalist care of patients with heart failure. Finally, just as there is potential heterogeneity in the skill set of different hospitalists, some cardiologists may or may not have specific formal training in the care of patients with heart failure. These characteristics were not captured in the present analysis.
Hospitals use a variety of physician types to care for patients with heart failure, and the relative mix of physician types varies significantly by hospital setting. Hospitalist care was not associated with improved 30-day outcomes and had a slight association with reduced length of stay. Hospitals with greater use of both hospitalists and cardiologists for patients with heart failure had greater adherence to some measures of quality. Further study is needed to identify the optimal inpatient care model to improve outcomes for patients with heart failure.
Damon M. Seils, MA, Duke University, assisted with manuscript preparation.
This work was supported by an award from the American Heart Association-Pharmaceutical Roundtable and David and Stevie Spina. Dr. Fonarow has a received research grants from AHRQ and NIH and is a consultant to Novartis, Medtronic, and Gambro. Dr. Curtis has received research grants from GlaxoSmithKline and Johnson & Johnson. Dr. Hernandez has received research grants from BMS, Portola Pharmaceuticals, Janssen, and Novartis. All other authors have reported that they have no relationships relevant to the contents of this paper to disclose. Clyde W. Yancy, MD, acted as Guest Editor for this article.
- Abbreviations and Acronyms
- angiotensin-converting enzyme
- angiotensin receptor blocker
- confidence interval
- Get With the Guidelines-Heart Failure
- implantable cardioverter defibrillator
- left ventricular systolic dysfunction
- Received June 13, 2013.
- Accepted July 17, 2013.
- American College of Cardiology Foundation
- Jong P.,
- Gong Y.,
- Liu P.P.,
- Austin P.C.,
- Lee D.S.,
- Tu J.V.
- Bonow R.O.,
- Ganiats T.G.,
- Beam C.T.,
- et al.
- Hernandez A.F.,
- Fonarow G.C.,
- Liang L.,
- Heidenreich P.A.,
- Yancy C.,
- Peterson E.D.
- ↵About COTH. Association of American Medical Colleges. Available at: https://www.aamc.org/members/coth/about/. Accessed June 12, 2013.
- Zou G.