Back to Annual Meeting

Back to Annual Meeting

APHA Scientific Session and Event Listing

Jose A. Suaya, MD, MPH, PhD¹, Donald S. Shepard, PhD¹, William B. Stason, MD, MSci¹, Sharon-Lise Normand, PhD², and Jeffrey Prottas, PhD³. (1) Schneider Institute for Health Policy, Brandeis University, 415 South Street, Waltham, MA 02454, 781 736 3904, Suaya@brandeis.edu, (2) Department of Health Care Policy, Harvard Medical School, 180 Longwood Ave., Boston, MA 02115, (3) Schneider Institute for Health Policy, Heller School, Brandeis University, 415 South St,, Mailstop 035, Waltham, MA 02454

Cardiac rehabilitation (CR) is widely recommended for the management of coronary heart disease. Although Medicare reimburses for outpatient CR services, utilization of CR was low—only 12.2% of eligible beneficiaries used it in 1997. We studied the effect of CR on 5-year survival among elderly Medicare beneficiaries eligible for CR. Using Medicare claims linked to Census 2000 data, we identified a retrospective cohort of 601,099 patients with a hospitalization with a coronary primary diagnosis in 1997 who survived at least 30 days after discharge and had analyzable data. All members of this cohort were considered candidates for CR.

For the propensity score approach, we first estimated a binary logistic regression model of survival with generalized estimating equations (which accounted for within-hospital clustering). Independent variables included patient characteristics (demographics, dual eligibility with Medicaid, coronary diagnosis and treatment, comorbid conditions), hospital characteristics (size, school affiliation, availability of catheterization, and Census division), zip code socio-economic characteristics of the patient's residence (e.g. % with college education and median income), and distance from patient's residence to the nearest CR facility. Based on the estimated probability of receiving CR services (propensity score), patients who received CR services were matched with patients who did not receive CR services but had equivalent propensity scores as well as similar gender, race, age-group, coronary diagnosis, coronary procedure, congestive heart failure, Medicaid enrollment, and Census division origin. We obtained 70,040 matched pairs of CR and non-CR users.

Because we were concerned about potential unmeasured selection factors, we used the ratio of the number of CR facilities per person aged 65 plus in the state and the distance from patient's residence to nearest CR facility as instrumental variables (IV).

In the matched paired analysis, the five-year mortality was 16.3% (95% confidence interval, CI: 16.1% to16.6%) among CR users and 24.6% (95% CI: 24.3% to 25.0%) among non-CR users. The absolute reduction was 8.3 percentage points (p<.0001) and the proportional reduction was 33.7%. With IV, both the absolute and proportional reductions (6.0 percentage points with p<.0001 and 24.4%, respectively) were moderately lower than benefits calculated using propensity scores. The underutilization of CR and important survival benefits underscore the importance of policy recommendations emphasizing widespread use of this highly effective service in eligible elders.

Learning Objectives:

• 1. Understand some of the selection issues in an observational study of cardiac rehabilitation (CR) and the need to control for them.
• 2. Understand how propensity scores and instrumental variables can each control for selection effects.
• 3. Understand the benefits of CR on 5-year survival using these two approaches, and reasons for minor differences between their results.

Presenting author's disclosure statement:

Not Answered