Integrating Clinical Trial Findings into Practice through Risk Stratification: The Case of Heart Failure Management

Abstract

Heart failure case management programs have been shown in clinical trials to be highly effective at preventing future hospitalizations. But the absolute benefits of these programs depend on the baseline risk of outcome in the treated population. Because baseline risks of hospitalization in trials are often higher than community-based samples, translating trial results to the community setting may be misleading. One solution is to identify subgroups for intervention that have sufficiently high baseline risk. Using estimates of hospitalizations averted from a previously published systematic review of heart failure management, we estimated a program's efficiency based on level of predicted risk. Medical history and demographic data on heart failure patients from a large integrated US health plan were used to build a logistic regression-based prognostic risk score for cardiovascular-related hospitalization over 1 year. We calculated the crude rate of hospitalizations for comparison with trial data. We also calculated the program's potential dollar savings from averting hospitalizations. The average risk of hospitalization in the systematic review's trials was 45%; our population's average observed risk was 18% and the risk among the highest risk patients was 33%. After accounting for the assumed annual intervention cost of $700, the base-case analysis (at $6000 per hospitalization) shows a savings of $122/patient at highest risk; failing to intervene according to predicted risk (no targeting) would actually cost $211/patient. Our findings illustrate how clinical trial findings can be efficiently integrated into community settings by using a prognostic risk score to focus attention on high-risk subgroups.