|Home | About | Journals | Submit | Contact Us | Français|
For patients with coronary heart disease, exercise-based cardiac rehabilitation improves survival and has beneficial effects on risk factors for coronary artery disease. The relationship between the number of sessions attended and long-term outcomes is unknown.
In a national 5% sample of Medicare beneficiaries, we identified 30,161 elderly patients who attended at least 1 cardiac rehabilitation session between January 1, 2000, and December 31, 2005. We used a Cox proportional hazards model to estimate the relationship between the number of sessions attended and mortality and myocardial infarction (MI) at 4 years. The cumulative number of sessions was a time-dependent covariate. After adjustment for demographic characteristics, comorbid conditions, and subsequent hospitalization, patients who attended 36 sessions had a 14% lower risk of mortality (hazard ratio [HR], 0.86; 95% confidence interval [CI], 0.77–0.97) and a 12% lower risk of MI (HR, 0.88; 95% CI, 0.83–0.93) compared with 24 sessions; a 22% lower risk of mortality (HR, 0.78; 95% CI, 0.71–0.87) and a 23% lower risk of MI (HR, 0.77; 95% CI, 0.69–0.87) compared with 12 sessions; and a 47% lower risk of mortality (HR, 0.53; 95% CI, 0.48–0.59) and a 31% lower risk of MI (HR, 0.69; 95% CI, 0.58–0.81) compared with 1 session.
Among Medicare beneficiaries, there was a strong dose-response relationship between the number of cardiac rehabilitation sessions and long-term outcomes. Attending all 36 sessions reimbursed by Medicare was associated with lower risks of mortality and MI at 4 years compared to attending fewer sessions.
Exercise-based cardiac rehabilitation is an important part of the long-term management of patients with chronic stable angina1 and of patients who have had a myocardial infarction (MI)2 or undergone coronary artery bypass graft surgery (CABG).3 Medicare has long reimbursed for cardiac rehabilitation for these indications, and the list of eligible indications recently expanded to include percutaneous coronary intervention, heart transplantation, and cardiac valve replacement or repair.4
Among patients with coronary heart disease, participation in cardiac rehabilitation, compared with nonparticipation, has been reported to have beneficial effects. In a meta-analysis of randomized controlled trials, participation in exercise-based cardiac rehabilitation improved both survival and risk factors associated with coronary artery disease.5 An observational analysis of Medicare beneficiaries in 1997 also found a survival benefit associated with any participation in cardiac rehabilitation among patients hospitalized for coronary conditions.6 Among patients with heart failure, results from Heart Failure: A Controlled Trial Investigating Outcomes of Exercise Training (HF-ACTION) suggested that exercise was efficacious.7
However, among patients who participate in cardiac rehabilitation, evidence regarding the optimal “dose” of cardiac rehabilitation is limited.5,6,8 It is not known whether attending more cardiac rehabilitation sessions is better than attending fewer or if there is a threshold number of sessions above which patients obtain less benefit. In this study, we sought to characterize the dose-response relationship between the number of cardiac rehabilitation sessions attended and the long-term risks of mortality and MI in a cohort of Medicare beneficiaries participating in cardiac rehabilitation.
We obtained research-identifiable Medicare claims data from January 1, 1999, through December 31, 2006, for a national 5% sample of beneficiaries from the Centers for Medicare & Medicaid Services. The data include inpatient, outpatient, physician, and denominator files. The inpatient file contains claims for institutional inpatient services. The outpatient file contains claims for institutional outpatient services, including emergency department visits. The physician file contains claims for other noninstitutional physician services. The denominator file contains information on beneficiary eligibility and enrollment and demographic characteristics. The study population was limited to Medicare beneficiaries aged 65 years or older living in the United States. The analyses were limited to claims incurred during periods of fee-for-service coverage. The institutional review board of the Duke University Health System approved the study.
We identified all patients with at least 1 Medicare claim for early outpatient cardiac rehabilitation services with or without continuous electrocardiogram monitoring (Current Procedural Terminology codes 93797 and 93798) during the study period. The date of each patient’s first cardiac rehabilitation claim served as the index date. To allow for determination of comorbid conditions present on the index date and determination of subsequent accrual of cardiac rehabilitation sessions and outcomes, we limited the analysis to patients whose index date was between January 1, 2000, and December 31, 2005. In addition, we required patients to have been enrolled in fee-for-service Medicare for the full year before the index date.
The exposure of interest was the number of cardiac rehabilitation sessions reimbursed by Medicare for each patient. We searched outpatient and physician claims for the dates of up to 36 cardiac rehabilitation sessions per patient in the 36 weeks after the index date. Although the Medicare national coverage determination specifies coverage for 36 sessions over 18 weeks, Medicare contractors have the discretion to cover additional sessions over 36 weeks.4
The primary outcome for this study was all-cause mortality, which we determined based on death dates in the Medicare denominator files. We also examined subsequent MI, which we determined by searching for the first inpatient or emergency department claim with a primary diagnosis of a new MI (International Classification of Diseases, Ninth Revision, Clinical Modification [ICD-9-CM] codes 410.x1) after the index date. We examined both outcomes through December 31, 2006. The length of follow-up used in the statistical models was 4 years following the index date.
Patient age, sex, and race were available in the denominator files. Comorbid conditions were based on the presence of relevant ICD-9-CM diagnosis codes on any inpatient, outpatient, or physician claim in the year before the index date. We used previously validated coding algorithms9,10 to search for any diagnosis of cancer (140.x-172.x, 174.x-208.x, 238.6), cerebrovascular disease (362.34, 430.x-438.x), chronic obstructive pulmonary disease (416.8, 416.9, 490.x-505.x, 506.4, 508.1, 508.8), dementia (290.x, 294.1, 331.2), diabetes mellitus (250.x), heart failure (428.x, 398.91, 402.x1, 404.x1, 404.x3), hypertension (401.x-405.x 437.2), peptic ulcer disease (531.x-534.x), peripheral vascular disease (093.0, 437.3, 440.x, 441.x, 443.1-443.9, 557.1, 557.9, V43.4), and renal disease (403.x1, 404.x2, 404.x3, 582.x, 583.0-583.7, 585.x, 586.x, 588.x, V42.0, V45.1, V56.x).
We searched inpatient claims for the date of each patient’s first hospital admission, if any, after the index date. We searched all claims in the year before the index date for the indication that qualified the patient for cardiac rehabilitation coverage. Between 2000 and 2005, the covered indications included MI (ICD-9-CM diagnosis code 410.xx), CABG (procedure code 36.1x or diagnosis code V45.81), and stable angina (diagnosis code 413.x). For patients with multiple indications, we selected the qualifying indication using a hierarchy of CABG, MI, and angina, in that order.
To describe characteristics of the study population at the index date, we present categorical variables using percentages and continuous variables using medians and interquartile ranges. We use a histogram to summarize the total number of cardiac rehabilitation sessions reimbursed by Medicare within 36 weeks of the index date for each patient.
To describe the observed mortality of the study population, we plotted cumulative incidence estimates based on a single Kaplan-Meier curve for all patients from the index date through 36 weeks, while rehabilitation sessions accrued, and separate Kaplan-Meier curves, conditional on survival to 36 weeks, from 36 weeks through 4 years for patients grouped by the total number of cardiac rehabilitation sessions. We tested for differences between the separate curves using a log-rank test. To describe the observed rate of MI while accounting for the competing risk of mortality, we plotted a single cumulative incidence curve11 for all patients from the index date through 36 weeks and separate cumulative incidence curves, conditional on event-free survival to 36 weeks, from 36 weeks through 4 years for patients grouped by the total number of cardiac rehabilitation sessions. We tested for differences between the curves using Gray tests.12
We estimated the unadjusted and adjusted relationships between the number of cardiac rehabilitation sessions and each outcome using Cox proportional hazards models. Treating the cumulative number of sessions as a time-dependent covariate, we first modeled its effect using a restricted cubic spline with knots at 10, 16, 22, and 28 sessions. A restricted cubic spline function allowed us to estimate a smoothed nonlinear relationship between exposure and outcome in a regression model.13 We assessed the nonlinear spline terms as a group using the Wald test. If this group test indicated that the terms added significant explanatory value to the model at α = 0.10, we retained the terms in the final model. If not, we omitted the terms from the final model, leaving only the linear term. The adjusted model also included demographic characteristics, comorbid conditions, and a time-dependent indicator for any hospitalization during the first 36 weeks after the first rehabilitation session to account for stops or delays in cardiac rehabilitation due to serious illness. We ran the models for all patients and by indication. If the relationship between an outcome and the number of sessions was estimated to be nonlinear, we plotted the hazard ratios (HRs) and 95% confidence intervals (CIs) for 36 sessions compared with each number of fewer sessions.
We performed a few sensitivity analyses. First, we ran the models on the subset of patients who had any diagnosis of heart failure in the previous 12 months, because patients with heart failure are a population of interest for cardiac rehabilitation. Second, to address the bias that may arise when some patients drop out of cardiac rehabilitation early, we ran the models on patients who attended at least 6 sessions. Finally, if the nonlinear spline terms were deemed necessary, we altered the number and position of the knots to determine whether the specification of the spline function led to different results.
We also used a linear regression model to estimate the associations of demographic characteristics, comorbid conditions, qualifying indication, and subsequent hospitalization with the number of cardiac rehabilitation sessions patients attended. Only patients who survived until 36 weeks were included in this analysis.
There were 30,161 Medicare beneficiaries in the national 5% sample who started cardiac rehabilitation between January 1, 2000, and December 31, 2005. Table 1 shows the characteristics of the study population at the index date. Most patients were men (64%) and 5% were nonwhite. The average age was 74 years. Forty percent had a diagnosis of congestive heart failure in the previous year and 36% had diabetes mellitus. The most frequent qualifying indication was CABG (61%). The qualifying indication for 4% of the patients could not be determined.
Figure 1 shows the distribution of cardiac rehabilitation sessions for each patient within 36 weeks of the index date. The median number of sessions was 25 (interquartile range, 13–34). More than 40% of the patients attended at least 30 sessions, and 13% attended fewer than 6. We limited the number of cardiac rehabilitation sessions to 36 because not all Medicare carriers reimburse for additional sessions. This limitation affected 1225 patients (4%) who had more than 36 sessions recorded in the claims data. Nearly one third of the patients (n = 8560) were hospitalized at some point during the 36 weeks after the index date. Among those hospitalized, 2676 (31%) resumed attending cardiac rehabilitation sessions after discharge.
The observed incidence of mortality is shown in Figure 2. Overall mortality during the first 36 weeks was 2.0% (n = 605). From this point through the end of follow-up, mortality differed significantly by the total number of rehabilitation sessions attended (P < .001). Mortality was consistently highest among patients who attended fewer than 12 sessions and lowest among patients who attended 36 or more sessions.
Figure 3 shows a similar plot for incidence of new MI. The overall incidence of MI was 1.6% (n = 464) during the first 36 weeks after the index date, with no noticeable early spike in events. After 36 weeks, there were statistically significant differences in the incidence of MI by the total number of cardiac rehabilitation sessions (P = .002). Patients who attended 24 or more sessions had the lowest subsequent rate of MI, while patients who attended fewer than 12 sessions had the highest incidence of MI.
The initial model for mortality indicated that the nonlinear spline terms for the cumulative number of cardiac rehabilitation sessions were statistically significant (χ22 = 15.6; P < .001). Significant results were also obtained for all indications. Therefore, we retained these terms for all of the mortality models. The number and position of knots for the spline function in the mortality model did not change the results, so we used the original specifications. The nonlinear spline terms were not statistically significant in the initial MI model (χ22 = 1.0; P = .61). These terms were insignificant for the indications, as well. Therefore, all MI models included only a linear term for the cumulative number of cardiac rehabilitation sessions.
In the unadjusted mortality model, the spline function estimated a substantial and statistically significant relationship between the number of cardiac rehabilitation sessions and the risk of mortality. For example, attending 36 sessions was associated with an 18% lower risk of mortality compared with attending 24 sessions (HR, 0.82; 95% CI, 0.73–0.92), a 29% lower risk compared with attending 12 sessions (HR, 0.71; 95% CI, 0.64–0.78), and a 58% lower risk compared with attending 1 session (HR, 0.42; 95% CI, 0.38–0.47). Results of the adjusted mortality model are shown in Table 2. The HRs remained highly significant but were moderated slightly after adjustment for demographic characteristics, comorbid conditions, and subsequent hospitalization. The full spline function of the adjusted HR for mortality for 36 sessions compared with fewer sessions is shown in Figure 4. Attending 36 cardiac rehabilitation sessions was associated with significantly lower mortality compared with any other number of sessions, though the most substantial differences were seen in comparison with the first 12 to 18 sessions.
The unadjusted linear relationship between the cumulative number of cardiac rehabilitation sessions and MI was statistically significant. Attending 36 sessions was associated with a 15% lower risk of mortality compared with attending 24 sessions (HR, 0.85; 95% CI, 0.81–0.90), a 28% lower risk compared with attending 12 sessions (HR, 0.72; 95% CI, 0.64–0.80), and a 38% lower risk compared with attending 1 session (HR, 0.62; 95% CI, 0.52–0.73). Statistical significance remained after adjustment, though the HRs were slightly moderated (Table 2). After adjustment, each additional 6 cardiac rehabilitation sessions was associated with a 6% (HR, 0.94; P < .001) lower risk of MI.
Specific results from the adjusted mortality model by cardiac rehabilitation indication are shown in Table 3, and the full estimated spline functions based on those models are shown in Figure 4. For patients who underwent CABG, there appears to have been a plateau at 24 cardiac rehabilitation sessions, beyond which no difference in estimated risk was observed, compared to attending 36 sessions. Based on this model, attending 36 sessions was associated with significantly lower risk only compared with attending 15 sessions (HR, 0.86; 95% CI, 0.76–1.00) or fewer. For patients who had an MI, attending 36 sessions was associated with significantly lower risk of mortality compared with attending any lower number of sessions. For patients with a diagnosis of stable angina, there appears to have been a plateau at 18 sessions, beyond which no difference in estimated risk was observed, compared to attending 36 sessions. Attending 36 sessions compared with attending 9 sessions (HR, 0.76; 95% CI, 0.60–0.96) or fewer was statistically significant in this group.
Specific results from the adjusted MI model by cardiac rehabilitation indication are shown in Table 3. For each indication, attending 36 sessions was significantly associated with a lower risk of MI compared with any lower number of sessions. After adjustment, each additional 6 cardiac rehabilitation sessions was associated with a 5% (HR, 0.95; P = .02) lower risk of MI among patients who underwent CABG, a 6% (HR, 0.94; P = .03) lower risk among patients with a previous MI, and a 11% (HR, 0.89; P = .001) lower risk among patients with stable angina.
Among the 11,955 patients in the study population who had a diagnosis of heart failure, the distribution of qualifying indications was similar to the overall population. The adjusted relationship between the number of cardiac rehabilitation sessions and mortality was similar in shape to the relationship found among all patients, though statistical significance was not achieved for all comparisons with 36 sessions. The linear relationship between the number of cardiac rehabilitation sessions and MI was significant. Attending 36 sessions was associated with lower risk of MI compared with attending any lower number of sessions. Table 3 shows other comparisons by the number of cumulative cardiac rehabilitation sessions based on the adjusted mortality and MI models.
Among the 26,497 patients who attended 6 or more cardiac rehabilitation sessions, the estimated relationship between the number of sessions attended and both mortality and MI were not substantially different from the results reported above for all patients. Compared with attending 36 sessions, attending any number of fewer sessions was associated with an increased risk of both mortality and MI. Based on regression models estimated within this subgroup, specific comparisons by number of cumulative sessions are shown in Table 3.
Table 4 shows predictors of the number of cumulative sessions attended. On average, patients who underwent CABG attended 0.5 more sessions than patients with an MI and 1 more session than patients with stable angina. Subsequent hospitalization led to patients attending almost 3 fewer sessions. Men attended slightly more sessions than women. Most comorbid conditions had only modest associations with the number of sessions attended. Overall, the model explained little of the variation in number of sessions attended (R2 = 0.05).
In this analysis of elderly Medicare beneficiaries who participated in cardiac rehabilitation, we found significant dose-response relationships between the number of cardiac rehabilitation sessions attended and the risk of mortality and MI at 4 years. For both outcomes, attending the maximum of 36 sessions was associated with lower risk compared with any lower number of sessions. The relationship with survival was found to be nonlinear, with greater differences in survival associated with attending the first 18 sessions compared with later sessions. The relationship with MI was found to be linear.
Previous evidence on the dose-response relationship between cardiac rehabilitation and survival among patients with coronary heart disease is mixed. A review of 14 published randomized controlled trials did not find a dose-response relationship.5 Most of these trials were published before 2000. A recent analysis of Medicare beneficiaries hospitalized in 1997 did find a dose-response relationship. 6 Both studies relied on a dichotomous measure of dose. Our work updates and expands on these analyses. First, we use data on patients who attended their first cardiac rehabilitation session between 2000 and 2005. Second, instead of using a dichotomous measure of dose, we explored dose as a continuous measure, which gave us greater statistical power to detect a relationship between dose and long-term outcomes. Using a continuous measure of dose also allowed us to determine the shape of the relationship.
Patients with heart failure benefited from attending more cardiac rehabilitation sessions. The dose-response relationship for both mortality and MI among patients with heart failure was similar to the relationship observed among all patients. During the study period, cardiac rehabilitation was only reimbursed by Medicare for diagnoses or events related to ischemic heart disease. Therefore, it is not known whether patients with nonischemic heart failure would have the same benefit we observed. Recently reported results from HF-ACTION indicate that exercise-based cardiac rehabilitation may be beneficial in this group.7 An additional analysis of the HF-ACTION cohort participating in the exercise intervention identified a dose response of exercise intensity, in terms of metabolic equivalent hours per week, in predicting the primary outcome of the trial, all-cause mortality or all-cause hospitalization.8
Cardiac rehabilitation is underused in the indicated populations. Previous research suggests that less than 20% of eligible Medicare beneficiaries participate in cardiac rehabilitation.14 We found that even among patients who attended any rehabilitation sessions, only 18% attended the full 36 sessions that Medicare reimburses. Given our finding that 36 sessions of rehabilitation is significantly associated with lower long-term risk of mortality and MI compared with fewer sessions, it may be worthwhile to investigate the reasons for lack of adherence to a full course of cardiac rehabilitation. Demographic characteristics and comorbid conditions were not strong predictors of adherence in this population.
The effects of cardiac rehabilitation among elderly patients are not well characterized. In this study, we used a large national sample of elderly Medicare beneficiaries to demonstrate a dose-response relationship between cardiac rehabilitation and long-term outcomes. Similar to findings among patients of all ages,15 there did not appear to be a safety signal with respect to early MI in this population.
Our study has some limitations. First, because the study population consists of Medicare fee-for-service beneficiaries, the results may not be applicable to beneficiaries enrolled in a Medicare managed care program or to younger patients. Second, it is not possible to state definitively that more frequent attendance at cardiac rehabilitation sessions is the sole mechanism of reducing the risks of long-term mortality and morbidity in this population. Adherence to early cardiac rehabilitation may be a proxy for other factors, such as a patient’s general activity level, long-term exercise adherence, medication adherence, socioeconomic characteristics, and smoking status,16 all of which directly affect adverse outcomes. To the extent that patients who are more adherent to cardiac rehabilitation have a more favorable profile for these other factors, it is possible that we overestimated the observed relationships. We did not have the data to explore these possible explanations. Third, there is a potential for bias due to the unmeasured underlying health status of patients in this study. We based the analysis on patients who attended at least 1 cardiac rehabilitation session because these patients are generally healthier than patients who do not participate in cardiac rehabilitation.13 However, there may be residual confounding by health status even among these patients, though it is unclear in which direction this might have biased the results. Healthier patients may not feel the need to attend cardiac rehabilitation and may be more likely to attend fewer sessions. It is also possible that less healthy patients cannot tolerate cardiac rehabilitation and thus attend fewer sessions. We tried to address this concern in a sensitivity analysis by only including patients who attended 6 or more sessions. The significant dose-effect relationship remained in this subset of the study population.
Among Medicare beneficiaries, there is a significant dose-response relationship between the number of cardiac rehabilitation sessions attended and long-term outcomes. Attending all 36 sessions that Medicare reimburses was associated with a lower risk of mortality and MI in the 4 years after the initiation of cardiac rehabilitation compared with attending fewer sessions. However, only 18% of patients who attended any cardiac rehabilitation sessions attended all 36 sessions. Our findings suggest that physicians should promote greater use of cardiac rehabilitation services and should strive to understand the barriers that prevent patients from attending the sessions to which they are entitled.
We thank Damon M. Seils, MA, Duke University, for editorial assistance and manuscript preparation. Mr Seils did not receive compensation for his assistance apart from his employment at the institution where the study was conducted.
Source of Funding: This study was supported by grant R01AG026038 from the National Institute on Aging and by grant U01HL063747 from the National Heart, Lung, and Blood Institute.
Exercise-based cardiac rehabilitation is recognized as an important part of the long-term management of coronary artery disease. It is associated with improved survival and beneficial changes in risk factors associated with coronary artery disease. However, the relationship between the dose of early outpatient cardiac rehabilitation and long-term outcomes is unknown. Using a large national sample of Medicare beneficiaries who attended any cardiac rehabilitation, we found that there was a strong dose-response relationship between the number of sessions attended and long-term mortality and myocardial infarction (MI). Attending all 36 sessions reimbursed by Medicare was associated with substantially lower risks of mortality and MI at 4 years compared to attending any number of fewer sessions. However, only 18% of patients who attended any cardiac rehabilitation sessions attended all 36 sessions. Combined with the fact that a minority of eligible patients attends any cardiac rehabilitation in the first place, the proportion of patients who attend all of the sessions to which they are entitled is small. Our findings suggest that physicians should promote greater use of cardiac rehabilitation services and should strive to understand the barriers that prevent patients from completing a full course of early outpatient cardiac rehabilitation.
Publisher's Disclaimer: Disclaimer: The content of this manuscript is solely the responsibility of the authors and does not necessarily represent the official views of the National Institute on Aging, the National Heart, Lung, and Blood Institute, or the National Institutes of Health.
Previous Presentation: Presented at the American Heart Association Quality of Care and Outcomes Research in Cardiovascular Disease and Stroke Conference; May 1, 2008; Baltimore, Maryland.
Conflict of Interest Disclosures: No relevant disclosures were reported. Drs Curtis and Schulman have made available online detailed listings of financial disclosures (http://www.dcri.duke.edu/research/coi.jsp).