|Home | About | Journals | Submit | Contact Us | Français|
Acquisition of data: Krumholz, Lichtman
Analysis or interpretation of data: Bueno, Curtis, Drye, Keenan, Kosiborod, Krumholz, Ross, Vidan, Wang, Normand
Drafting the manuscript: Bueno, Krumholz, Ross
Revising the manuscript for important intellectual content: All authors
Statistical analysis: Bueno, Krumholz, Normand, Wang
Obtaining funding: Krumholz
Administrative, technical, material support: Krumholz
For more than a decade, hospitals have focused on decreasing length of stay but this focus may have had unanticipated effects on patient care.
To describe the temporal changes in length of stay, discharge disposition, and short-term outcomes among older patients hospitalized for heart failure.
Observational study of Medicare fee-for-service hospitalizations for heart failure between 1993 and 2008, with a 30-day follow-up.
Length of hospital stay, in-patient and 30-day mortality, and 30-day readmission rates.
Between 1993 and 2008, 7,799,788 hospitalizations were studied. Length of stay decreased from 8.8 (95% CI, 8.8–8.8) to 6.3 (95% CI, 6.3–6.3) days. In-hospital mortality decreased by half, from 8.5% (95%CI, 8.4–8.6) in 1993 to 4.2% (95%CI, 4.2–4.3%) in 2008, whereas 30-day mortality decreased by 17%, from 12.8% (95%CI, 12.8–12.9%) to 10.9% (95%CI, 10.8–10.9%). Rates of discharge to home decreased 10% (from 74.4% to 66.9%), while discharges to skilled nursing facilities increased 65%, from 13% to 21.3%. Thirty-day readmission rates increased 27%, from 17.2% (95%CI, 17.1–17.3%) relative to 21.9% (95%CI, 21.8–22.0%) (all p values <0.001). Consistent with our unadjusted analyses, 2007–2008 risk-adjusted 30-day mortality risk was 0.86 (95% CI, 0.86–0.87) when compared with 1993–1994, while the 30-day readmission risk ratio was 1.19 (95%CI, 1.18–1.21).
For patients admitted with heart failure over the past 16 years, we observed reductions in length of stay and in-hospital mortality, less marked reductions in 30-day mortality, and increases in 30-day readmission rates and use of skilled nursing facilities after discharge.
Despite the therapeutic advances in treatment during the last decades, heart failure is the leading cause of hospitalization among Medicare beneficiaries.1–4 In the last years significant advances in the treatment of heart failure have been developed including drugs such as angiotensin-converting enzyme inhibitors, beta-blockers or aldosterone antagonists, and cardiac device-related therapies such as implantable defibrillators or resynchronization therapies. However, as most of the benefits produced by these treatments are seen after months or years of therapy, no parallel progress in the acute treatment of patients with heart failure has occurred. Despite this fact, there has been a substantial change in hospital length of stay for this population. As part of the effort to decrease hospital costs, the Health Care Financing Administration, now the Centers for Medicare & Medicaid Services, introduced the Prospective Payment System in 1982 while managed care organizations began to incentivize hospitals to reduce inpatient length of stay. Neither studies nor guidelines suggest criteria for determining an optimal length of stay for patients with heart failure, of whom the vast majority are 65 years and older and have substantial comorbidity.5–7 Unaccompanied by clinical evidence or a national surveillance system to determine the effect of this initiative on patients, the system nevertheless translated into a marked decrease in hospital length of stay.7–12
Large-scale changes in the way that care is delivered may be accompanied by unintended consequences. To understand the effects of these changes for payers and patients, it is necessary to examine care patterns and outcomes during and after the hospitalization. We studied patient outcomes in the hospital and peri-hospital period for Medicare patients hospitalized with heart failure. Using data from 1993 through 2008, we assessed changes in length of stay; discharge disposition; in-hospital, post-discharge and 30-day mortality rates; and 30-day readmission rates.
We obtained the Medicare Provider Analysis and Review (MEDPAR) files and the Denominator file from the Centers for Medicare & Medicaid Services for the years 1993 through 2008. The Denominator file includes Medicare beneficiary enrollment and mortality information from administrative enrollment records. It is an abbreviated version of the Enrollment Database that contains detailed data on all beneficiaries entitled to Medicare. The MEDPAR data contain hospital discharge abstracts for the acute care hospitalizations of all Medicare recipients covered by the hospital care program (Part A). Only patients covered by fee-for-service arrangements are included in the MEDPAR file. The study population included fee-for-service Medicare patients 65 years or older hospitalized with heart failure, as defined by a principal discharge diagnosis using International Classification of Diseases, Ninth Revision, Clinical Modification code (402.01, 402.11, 402.91, 404.01, 404.11, 404.91, 428, 404.03, 404.13, and 404.93). We excluded patients with incomplete information in the Medicare denominator file (e.g., health claim identification). For patients with multiple hospitalizations within a calendar year, only 1 randomly selected hospitalization was included in the sample. Hospitalizations in subsequent years beyond 30 days after the index hospital discharge were considered as potential index admissions.
Patient characteristics included demographics (age as a continuous variable, male gender, white, black, and other races), and history of cardiovascular and comorbid variables, the majority of which were utilized in the validated CMS HF 30-day all-cause hospital-specific mortality measure,Krumholz and were used to develop a statistical model that was clinically sensible and statistically sound built upon prior work.Jencks We created these variables based on the inpatient Part-A MEDPAR data (i.e., Part-A outpatient and carrier data were not included), using the information identified from claims submitted in the year before the index hospitalization and from claims found in the index admission for those conditions that could not represent a complication of the admission.
Outcomes included length of stay, defined as the difference between discharge and admission dates plus 1; all-cause 30-day readmission, defined as any re-hospitalization to any acute care hospital within 30 days from index discharge; and all-cause 30-day mortality, defined as death from any cause 30 days following the index admission date. We chose 30-day mortality and readmission rates because they are more likely to reflect variations related to changes in length of stay compared with outcomes measured at longer follow up periods, and because they are standard measurements of quality of care.13, 14
In addition, we examined in-hospital mortality, defined as death during the index hospitalization, and post-discharge mortality, which was calculated as the difference between 30-day and in-hospital mortality. Patients who were transferred from the admitting hospital to another acute care hospital, and those who died during the index hospitalization, were excluded for the readmission analyses. Information on readmission and in-hospital mortality were obtained from MEDPAR data, and post-discharge mortality information was obtained from the Denominator file by linking unique patient identifiers.
We conducted bivariate analysis to quantify changes in patient characteristics and observed outcomes by biannual years using the Mantel-Haenszel chi-square test for categorical variables and Cuzick's nonparametric test for continuous variables.15 We used 2-year periods to ease our presentation given space limitations. We used a survival life table to calculate 30-day readmission; patients who neither died nor were readmitted within 30 days after discharged were right-censored. We constructed Cox proportional models to evaluate the changes in 30-day all-cause mortality and readmission rates over time adjusted for age, gender and comorbidities. We created dummy variables for each 2-year period, using 1993–1994 period as the reference. We plotted Kaplan-Meier curves to test the assumption of proportional hazards, and the curves were roughly parallel. To account for within-hospital correlation of patient outcomes, we used the Lin and Wei sandwich method to calculate robust estimates of standard errors.16 All statistical tests were 2-sided, at a significance level of 0.05, and was performed with SAS version 9.2 (SAS Institute Inc., Cary, NC) and STATA 9.0 (STATA Corporation, College Station, TX).
The Yale University Human Investigation Committee determined that the study did not require approval or waiver.
Between 1993 and 2008, we identified 7,799,788 hospitalizations based on the selection criteria. The changes in baseline characteristics throughout the study period are shown in Table 1. Secular trends for length of stay and short-term outcomes in 2-year intervals are reported in Table 2. In 1993, 498,500 hospitalizations for heart failure were studied with a mean length of stay of 8.81 days (95%CI 8.79–8.83), and an in-hospital mortality rate of 8.5% (95%CI, 8.4–8.6). Of those patients who survived and were not transferred to another acute care hospital, 74% (95%CI, 73.8–74.1) were discharged to home and 13% (95%CI, 12.9–13.1) were discharged to skilled nursing facilities. In contrast, during 2008 for which 412,614 hospitalizations for heart failure were analyzed, mean length of stay was 6.32 days (95% CI, 6.31–6.33), and in-hospital mortality rate 4.2% (95%CI 4.2–4.3%). The proportion of survivors who were discharged to home was 66.8% (95%CI 66.8–66.9), and to skilled nursing facilities 21.3% (95% CI 21.2–21.4).
During the 16-year study period, the mean length of stay decreased 2.4 days (28% relative reduction, p<0.001, Figure 1A) and the discharge disposition changed significantly (p<0.001), with a 65% relative increase in the proportion of discharges to skilled nursing facilities, and a 10% relative reduction in the proportion of discharges to home [Figure 1C]). The in-hospital mortality rate decreased 4.3%, yielding a 50% relative reduction between 1993 and 2008 [Figure 1D].
The unadjusted rate of 30-day mortality rate decreased 2.14% (p<0.001) from 12.8% (95%CI, 12.8–12.9) in 1993 to 10.9% (95%CI, 10.8–10.9) in 2008, a 16.7% relative reduction [Figure 1D]. In contrast, post-discharge mortality —from discharge to the 30th day after admission— rose 2.4% (from 4.0% in 1993 to 6.4% in 2006), a 61% relative increase, p<0.001. Due to this steady increase, post-discharge mortality exceeded in-hospital mortality from 2003 to the end of the study period [Figure 1D].
There were 6,753,926 patients who were discharged alive and not transferred to an acute care hospital during the study period. The number of patients readmitted for all causes during the next 30 days after discharge increased from 71,412 in 1993 to 89,560 in 2008, such that the crude readmission rate increased 4.7% (p<0.001), from 17.2% (95% CI, 17.1–17.3) in 1993 to 21.9% (95% CI, 21.8–22.0) in 2008, a 27% relative increase [Figure 1B]).
During the study period, the risk-adjusted 30-day mortality and readmission risks changed progressively and inversely [Figure 2]. Compared with years 1993–1994, the 30-day mortality risk ratio for years 2007–2008 was 0.86 (95% CI, 0.86–0.87) while the 30-day readmission risk ratio was 1.19 (95%CI, 1.18–1.21).
In this large, national study, we find that over a 16-year period of reduction in hospital length of stay and increased use of skilled nursing facilities after discharge for Medicare patients with heart failure, 30-day mortality was reduced but post-hospital readmission and mortality risk increased. From the patient perspective, it is not clear that care in 2008 is markedly better than it was in 1993. The outcome of patients hospitalized for heart failure, as measured by short-term mortality, has improved, which may be a result of better treatment. However, because length of stay has substantially decreased, improvement based on in-hospital mortality is misleading. In contrast, rates of readmission and discharge to skilled nursing facilities have also increased, suggesting that patient outcomes, while better, have not improved substantially.
The Medicare fee-for-service system provided an incentive for shortening length of stay without penalty for potential unfavorable later outcomes such as increased readmission or mortality rates. This policy has been considered responsible for the progressive reduction in length of stay observed in patients with heart failure in the United States.9 However, despite the main objective of reducing hospital costs, it is possible that when the hospital and short-term non-hospital post-discharge costs are considered, this policy failed to reduce healthcare expenses. It is unknown whether the increased use of non-acute settings or the increase in short-term readmission risk are the best options for the system, or whether they have been aligned with patient preferences and resulted in increased patient satisfaction.
Several studies have addressed the secular trends of various aspects related to the care and outcomes of patients with heart failure in the United States, such as incidence, hospitalization rate, hospital stay, therapy, mortality, discharge destination, and readmission risk.1, 6, 7, 9–12, 17–24, Ross J The present study is the most comprehensive examination of recent short-term post-hospital management of heart failure and its consequences in the United States.
Our study also indicates the importance of examining an episode of acute care over a standardized period of assessment rather than merely focusing on the hospitalization. The approach of using a standardized period of assessment, particularly 30-day mortality, and readmission rates was endorsed for performance measures by the American College of Cardiology and the American Heart Association.13 Current payments are recognized as a major limitation of the Medicare fee-for-service payment system contributing to poor coordination of care across settings, and the Medicare Payment Advisory Commission has recommended that Medicare adopt episode-based payments including care provided during and post-hospitalization. In this study, had we focused solely on the period of hospitalization, we would have reached different conclusions, perhaps finding that hospital stay could be shortened and that there was a remarkable reduction in mortality. Only with the study of the peri-hospitalization period are we able to see the full change in outcomes for patients.
We found that hospital deaths were, to some extent, being shifted outside the hospital. The marked reduction in in-hospital deaths was accompanied by an increase in early post-discharge deaths. We cannot determine if these deaths were expected or whether the place of death imposed a burden on patients and their families or was consistent with their preferences. We also cannot determine if the shorter hospital stay contributed to some of the deaths. Nevertheless, we did find that the period did not have the magnitude of improvement in mortality that was suggested by the in-hospital experience alone.
The most striking finding is that the period was associated with an increase in 30-day readmission rate. While we cannot demonstrate that the shortened hospital stay caused these changes, it is certainly plausible that the effort to discharge patients quickly has led to transfers to non-acute institutional settings and occasionally sent patients out of the hospital before they are fully treated. Moreover, there is a paucity of studies that test criteria for readiness for discharge, adding to uncertainty about what constitutes appropriate hospital treatment for the condition.
Our findings are consistent with those of most studies and at odds with some others. Our findings concur with previous studies regarding reduction in the length of stay7, 9–12 and in the proportion of patients discharged to home, as well as an improvement in in-hospital1, 6, 9, 11, 12, 20 and 30-day mortality rates.6, 9, 10, 18, 20, 21 During the 1990s, a progressive increase in 30-day readmission rates for patients with heart failure was reported in the United States7, 9 and Canada,25 while more recent reports did not find such an increase in the 2000s in the Medicare population.6, Ross Our study demonstrates that during an observation period of 16 years, spanning most of the time during which the previous studies were performed, there has been a slow but steady increase in 30-day all-cause readmissions.
Our study has several limitations. First, the use of administrative data precludes the consideration of some clinically relevant prognostic factors as well as the evaluation of the quality of care. For instance, there is no information about changes in treatment during the study period. However, the reported trends towards an increase in the use of angiotensin-converting enzyme inhibitors and beta-adrenergic blockers, and a decrease in the use of inotropic agents in the United States12, 19, 23 as well as the modest effect of these therapies on short-term outcomes make it unlikely that therapeutic changes accounted for the increase in post-discharge outcomes. The inability to evaluate the quality of hospital care precludes the evaluation of the rate of premature discharge, as well as the identification of length of stay reduction as the causal factor for the increase in early post-discharge outcomes. Potential alternative explanations for our findings deserve consideration. For instance, a decrease in the threshold for admission of sicker heart failure patients would likely increase the rate of early outcomes. In addition, Medicare payments for hospital and skilled nursing care have undergone further changes over the study period. Our study focused on length of stay, discharge status, readmission and mortality, not on other important dimensions of patient outcomes, such as functional status or quality of life. Also, we were limited in our ability to determine if patients switched into managed care before the index hospitalization. This period was used to determine the comorbidities, but the results were similar to the crude analysis and would not be expected to affect the results. Finally, the study of fee-for-service beneficiaries may be associated with the selection of sicker patients, and an underestimation in outcomes and their time changes.Shimada
In conclusion, the pattern of care during hospitalization and immediately after for older patients with heart failure has changed substantially in the United States during the last 16 years. On the positive side, the 30-day mortality rate has decreased. However, the increase in the readmission rates which paralleled the decrease in length of stay does raise concerns – as does the increase in the discharge to nursing home facilities. The current model of care for older patients with heart failure in the United States may benefit from more attention to the care and outcomes in the early transition period after hospital discharge.
Data access and responsibility. Dr. Wang had full access to all the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Funding/Support and role of sponsor. The analyses upon which this publication is based were performed under Contract Number HHSM-500-2005-CO001C, entitled “Utilization and Quality Control Quality Improvement Organization for the State (commonwealth) of Colorado,” funded by the Centers for Medicare & Medicaid Services, an agency of the U.S. Department of Health and Human Services. The content of this publication does not necessarily reflect the views or policies of the Department of Health and Human Services, nor does mention of trade names, commercial products, or organizations imply endorsement by the U.S. Government. The author assumes full responsibility for the accuracy and completeness of the ideas presented. This work was also funded by grant number K01 DP000085-05 from the Centers for Disease Control and Prevention. The contents of this paper are solely the responsibility of the authors and do not necessarily represent the official views of the Centers for Disease Control and Prevention.
The sponsors did not participate in the design and conduct of the study; collection, management, analysis, and interpretation of the data; or preparation of the manuscript. CMS reviewed and approved the use of its data for this work and approved submission of the manuscript; this approval is based on data use only, and does not represent a CMS endorsement of or comment on the manuscript content.
Drs. Héctor Bueno and María T. Vidán were supported in part by grants from the Fondo de Investigación Sanitaria del Instituto de Salud Carlos III, Spain (BA08/90010 and BA08/90012, respectively) while at the Center for Outcomes Research and Evaluation, Yale-New Haven Hospital, where they participated in the development of this study. Dr. Joseph Ross is supported by the National Institute on Aging (K08 AG032886) and by the American Federation for Aging Research through the Paul B. Beeson Career Development Award Program. Dr. Kosiborod is supported by the American Heart Association Career Development Award in Implementation Research.
Conflicts of interest. Dr. Bueno reports having received consulting fees from Almirall, Bayer, BMS, and Sanofi-Aventis, and research grants from the Fondo de Investigaciones Sanitarias (Instituto Carlos III, Ministry of Science, Spain), Astra-Zeneca, BMS, and Pfizer. Drs. Wang, Chen, Normand, Curtis, Drye, and Keenan report that they develop and maintain performance measures under contract with the Centers for Medicare & Medicaid Services. Dr. Kosiborod reports that he served on the advisory board of Sanofi-Aventis and received speaking honoraria from the Vascular Biology Working Group and DiaVed, Inc. Dr. Krumholz reports that he chairs a scientific advisory board for UnitedHealthcare.
Additional information. This work has not been previously presented and is not under consideration elsewhere. There is no information to report on independent statistical analysis, and there are no individuals who have made substantial contributions to the work who are not authors.