Abstract
BACKGROUND: Prognostic models, such as the Seattle Heart Failure Model (SHFM), have been developed to predict patient survival. The extent to which they predict medical resource use and costs has not been explored. In this study, we evaluated relationships between baseline SHFM scores and 1-year resource use and costs using data from a clinical trial. METHODS AND RESULTS: We applied generalized linear models to examine the relative impact of a 1-unit increase in SHFM scores on counts of medical resource use and direct medical costs at 1 year of follow-up. Of 2331 randomized patients, 2288 (98%) had a rounded integer SHFM score between -1 and 2, consistent with predicted 1-year survival of 98% and 74%, respectively. At baseline, median age was 59 years, 28% of patients were women, and nearly two-thirds of the cohort had New York Heart Association class II heart failure and one-third had class III heart failure. Higher SHFM scores were associated with more hospitalizations (rate ratio per 1-unit increase, 1.86; P < .001), more inpatient days (2.30; P < .001), and higher inpatient costs (2.28; P < .001), outpatient costs (1.54; P < .001), and total medical costs (2.13; P < .001). CONCLUSION: Although developed to predict all-cause mortality, SHFM scores also predict medical resource use and costs.