National Trends in Hospital Readmission Rates among Medicare Fee-for-Service Survivors of Mitral Valve Surgery, 1999–2010

Background Older patients who undergo mitral valve surgery (MVS) have high 1-year survival rates, but little is known about the experience of survivors. Our objective was to determine trends in 1-year hospital readmission rates and length of stay (LOS) in these individuals. Methods We included 100% of Medicare Fee-for-Service patients ≥65 years of age who underwent MVS between 1999–2010 and survived to 1 year (N = 146,877). We used proportional hazards regression to analyze the post-MVS 1-year readmission rate in each year, mean hospital LOS (after index admission), and readmission rates by subgroups (age, sex, race). Results The 1-year survival rate among patients undergoing MVS was 81.3%. Among survivors, 49.1% experienced a hospital readmission within 1 year. The post-MVS 1-year readmission rate declined from 1999–2010 (49.5% to 46.9%, P<0.01), and mean hospital LOS decreased from 6.2 to 5.3 (P<0.01). Readmission rates were highest in oldest patients, but declined in all age subgroups (65–74: 47.4% to 44.4%; 75–84: 51.4% to 49.2%, ≥85: 56.4% to 50.0%, all P<0.01). There were declines in women and men (women: 51.7% to 50.8%, P<0.01; men: 46.9% to 43.0%, P<0.01), and in whites and patients of other race, but not in blacks (whites: 49.0% to 46.2%, P<0.01; other: 55.0% to 48.9%, P<0.01; blacks: 58.1% to 59.0%, P = 0.18). Conclusions Among older adults surviving MVS to 1 year, slightly fewer than half experience a hospital readmission. There has been a modest decline in both the readmission rate and LOS over time, with worse outcomes in women and blacks.


Introduction
Mitral valve disease (most often regurgitation) is common in older adults [1]. In the subset of patients with severe symptomatic mitral regurgitation, definitive therapy involves repair or replacement of the mitral valve. Our group [2] and others [3] have shown a significant decline in U.S. mortality rates after mitral valve surgery (MVS) over time; for example, in Medicare beneficiaries 65 years undergoing isolated MVS over a 10-year period, risk-adjusted mortality declined by 40%, and by 2008 seven out of every eight patients undergoing MVS were alive at 1 year [2].
While this trend is encouraging, as the vast majority of MVS patients now survive, mortality statistics reveal little about the heterogeneity of experience among survivors. Survival is a successful outcome, but freedom from acute events that lead to hospitalizations is also important [4][5][6]. Do patients who survive after MVS tend to have repeated hospitalizations? How many days do they spend in the hospital? What are the causes of their hospitalizations? Data on these events in MVS survivors could be helpful in informing operative candidates about their risks after surgery and assist physicians and patients in making informed decisions and developing interventions to improve outcomes.
Accordingly, we sought to describe the overall hospital readmission rate and mean post-discharge length of stay (LOS) among U.S. older adults who survived MVS beyond 1 year, as well as trends in these outcomes over time, by analyzing data from 100% of Medicare fee-for-service (FFS) beneficiaries over a 12-year time period. The inclusion of all FFS beneficiaries allows for the establishment of a national benchmark for these outcomes, and also provides sufficient sample size to identify meaningful trends in potentially high-risk subgroups of age, sex, and race. To further characterize the experience among MVS survivors we analyzed discharge disposition, reasons for readmission, and outcomes by type of surgery.

Data Source and Coding
We used Medicare inpatient claims from the Centers for Medicare and Medicaid Services (CMS) to identify all Medicare Fee-for-Service patients discharged from an acute care hospital in the United States between January 1, 1999 and December 31, 2010, for a MVS procedure defined by the International Classification of Diseases, Ninth Revision, Clinical Modification (ICD-9-CM) codes of 35.12 (mitral vale repair), 35.23 (bioprosthetic mitral valve replacement), or 35.24 (mechanical mitral valve replacement). We further defined a subgroup of patients undergoing "isolated mitral valve surgery" by excluding patients with the following coexisting codes: CABG (36. 10 Institutional review board approval was obtained through the Yale University Human Investigation Committee, and the requirement for informed consent was waived based on the nature of the study. Beneficiary confidentiality was protected through a data use agreement with CMS. Patient information was anonymized but not de-identified prior to analysis. However, all analyses were conducted in a restricted office that meets CMS patient information and data security requirements.

Patient Characteristics and Comorbidities
Patient demographic information includes age, sex, and race (white, black, other). Comorbidities were drawn from those used to profile hospitals by the CMS 30-day mortality measures for acute myocardial infarction [7] heart failure [8], and pneumonia [9]. They were identified from secondary diagnosis codes recorded at time of discharge from hospitalization for mitral valve surgery (which did not represent a potential complication) as well as primary or secondary diagnosis codes of all inpatient hospitalizations up to 1 year before hospitalization for surgery. Data from 1998 were used for patients hospitalized for MVS in 1999.

Outcomes
The primary outcome was 1-year all-cause hospital readmission, defined as patients who had at least one rehospitalization within 1 year after surgery among patients who survived to 1 year. The "time zero" for a 1-year all-cause readmission was the date of the index MVS discharge. Secondary outcomes were the aggregate length of stay (LOS) (days) for all readmissions following index hospitalization for MVS, and total Medicare payments, within 1-year post-MVS. Post-MVS rehospitalization LOS represented the number of days spent in the hospital after the index MVS admission, and was calculated as the cumulative LOSs occurring after the index MVS hospitalization per MVS survivor. We report the mean LOS among those who were readmitted. The 1-year post MVS Medicare payments represented the inpatient Medicare expenditures (Medicare Part A) per MVS patient within the first year of MVS. We reported unadjusted dollar amounts and then adjusted using the annual Consumer Price Index inflation rate reported by the Bureau of Labor Statistics, United Stated Department of Labor (http:// www.bls.gov/data/inflation_calculator.htm).

Statistical Analysis
We examined patient demographic and clinical characteristics across years. The 1-year readmission rate was expressed as percentage; Medicare expenditures and length of stay as means (standard deviation [SD]). We used the Cochran-Armitage trend test to determine if changes over time in the outcomes were statistically significant. Data for each year of observation were included in this test; for purposes of simplicity, we report rates for first (1999) and last (2010) years in the text of the Results. We fitted a mixed model with a logit link function to estimate the changes in the 1-year rehospitalization rate after MVS over time, adjusting for patient demographics and comorbidities. We estimated the model with hospital-specific random intercepts to account for within-hospital and between-hospital sources of variation. We included an ordinal time variable, ranging from 0 to 10, corresponding to years 1999 (time = 0) to 2010 (time = 11), in the model. The odds ratios of this time variable represent the adjusted annual change in the 1-year readmission rate. We repeated the model for age and gender-race groups.

Patient characteristics
Over the aggregate study period, patients with mechanical MVR were the most likely to experience readmission (54.8%), followed by bioprosthetic MVR (51.8%) then MV repair (45.7%). Readmission rates over time stratified by valve subtype are shown in Table B  . The above findings did not change substantially after accounting for patient characteristics and within and between-hospital sources of variation, except for black patients, which after adjustment showed a statistically significant decline. Overall, the adjusted annual decline was 2.3% (95% CI 1.6 to 3.0). Fig 2 shows the adjusted annual decline in 1-year readmission by age-gender-race subgroup and overall.
In the isolated MVS cohort, 1-year readmission rates were generally lower in each year (

Trends in CMS payments
Based on the number of patients who were hospitalized in the year following MVS in 2010 (5,151/10,983), we calculated the increase in total cost compared with the equivalent number of patients hospitalized in 1999. The average unadjusted CMS payment per patient for all

Discussion
Using a 100% sample of Medicare fee-for-service beneficiaries we found that slightly fewer than half of older adults who underwent MVS and survived to 1 year were readmitted to the hospital. Over time there was a modest decline in the readmission rate, despite findings of an increasing age and comorbidity burden among the population. In addition, over the same period we found that fewer days on average (following the surgical admission) were spent in the hospital; by 2010 the mean number of days had declined to 5.3 (compared with 6.2 in 1999). Furthermore, there was a decline in the proportion of patients with extended hospital stays; in 2010, fewer than 1 in 25 patients spent more than 30 days in the hospital. Reasons for readmission were heterogeneous; in both 1999 and 2010 the top two diagnostic codes were heart failure and arrhythmia, although together they only accounted for 28.7% and 32.4% of the total readmissions in each year, respectively.
Several factors may explain the modest decline in readmissions. We found that over time, fewer patients were discharged directly to home without home care, and more were discharged either home with home care (relative increase 39%) or to skilled nursing or rehabilitation facilities (relative increase 86%). Similar trends have previously been described in Medicare patients undergoing CABG [10] and those hospitalized for heart failure [11]. It is possible that for patients undergoing MVS, the increased post-discharge attention provided in these settings facilitated interventions aimed at addressing preventable or unnecessary readmissions. Further study would be necessary to support this relationship; for other conditions such as heart failure and acute myocardial infarction, data on the potential for skilled nursing facility referral to reduce readmissions have been mixed [12,13]. Another factor in the reduced readmission rate among 1-year survivors may be the increasing focus of cardiac surgical teams in reducing sternal wound infections [14] and pneumonia [15] among patients undergoing open heart surgery, which may have reduced readmissions related to hospital-acquired infections over time. Technological advances in operative techniques, particularly an increasing adoption of mitral valve repair over time (obviating the need for long-term oral anticoagulation), may also have contributed. It is also possible that patients referred for surgery were healthier; while we found that age and comorbidities increased, which parallels findings of others [3], it has also been reported that preoperative New York Heart Association (NYHA) functional class among MVS candidates has improved over time [16], suggesting a trend towards earlier intervention.
There were important differences among subgroups that persisted over the study period. Over all years, black patients had the highest rates of readmission. This parallels findings of higher readmission rates for black patients after other cardiac procedures including CABG  [17] and percutaneous coronary intervention [18]. Given the administrative nature of our data we are limited in determining causal factors, though prior studies among patients undergoing MVS have reported black patients having higher rates of postoperative prolonged mechanical ventilation and renal failure [19] which may contribute to an increased vulnerability to readmission after discharge. Other system-level factors may also have played a role such as site of hospital care [20] or poor access to early outpatient follow-up care after discharge [21], as reported for other conditions. We also found that the oldest patients (85 years) had a higher rate of 1-year readmissions than younger age groups. While this may seem intuitive due to the general prevalence of comorbidities and frailty among this population, other studies have found no relationship between age and readmission after hospitalizations for heart failure [22] or general surgical procedures [23]. Finally, patients who underwent mechanical MVR were more likely to experience readmission than patients with MV repair or bioprosthetic MVR, which may be related to bleeding complications associated with mandatory anticoagulant therapy, or alternately due to other patient characteristics such as advanced heart disease (both technically prohibiting MV repair and leading to increased vulnerability for hospital readmission). Further studies with more granular data would be required to confirm such a relationship.
Our findings must be interpreted in the context of our study design. First, our analysis was limited to Medicare FFS beneficiaries and we therefore do not have data on patients enrolled in Medicare managed care programs, where enrollment has increased over time [24]. Similarly, we do not have data on patients younger than age 65 undergoing MVS. Second, our study sample was restricted to MVS survivors, and since mortality rates have decreased over time, the nature of patients surviving MVS (and at-risk for readmission) may have changed. However, the absolute decrease in mortality over time was relatively small. Third, since we relied on administrative data it is possible that coding of comorbidities changed over time (e.g. upcoding or down-coding), which could have affected our ability to adjust for potential confounders. With our administrative dataset we were also unable to account for certain selection attributes that may have influenced the choice of patients for surgery, and we could not adjust for preoperative data such as ejection fraction and symptom severity which may have influenced outcomes. Finally, beyond categories of age, sex, and race, our study was not designed to determine risk factors for post-MVS readmission. Readmission events are likely to represent a complex interaction of patient-level factors (including specific vulnerabilities to acute medical illness) and system-level characteristics (such as the quality of in-hospital and post-hospital care). Elucidating the determinants of post-hospital risk remains an important direction for future investigation.
In conclusion, we found that slightly fewer than half of MVS 1-year survivors experienced a hospital readmission. While there was a modest decline in this rate over time, there is still considerable room for interventions that may further reduce preventable events. Our finding of considerable variation among subgroups requires further study and suggests targeted areas for quality improvement.
Supporting Information S1 File. This file contains Tables A-C. Trends in readmission rate among mitral valve surgery survivors at 1 year, 1999-2010 (Table A). Trends in 1-year readmission by surgical subtype among 1 year MVS survivors, 1999-2010 (Table B). Trends in mean total length of stay (within 1 year of discharge) among mitral valve surgery survivors at one year, 1999-2010 (Table C). (PDF)