Customizing national models for a medical center's population to rapidly identify patients at high risk of 30-day all-cause hospital readmission following a heart failure hospitalization

Abstract

Background: Nationally-derived models predicting 30-day readmissions following heart failure (HF) hospitalizations yield insufficient discrimination for institutional use.

Objective: Develop a customized readmission risk model from Medicare-employed and institutionally-customized risk factors and compare the performance against national models in a medical center.

Methods: Medicare patients age ≥ 65 years hospitalized for HF (n = 1,454) were studied in a derivation cohort and in a separate validation cohort (n = 243). All 30-day hospital readmissions were documented. The primary outcome was risk discrimination (c-statistic) compared to national models.

Results: A customized model demonstrated improved discrimination (c-statistic 0.72; 95% CI 0.69 - 0.74) compared to national models (c-statistics of 0.60 and 0.61) with a c-statistic of 0.63 in the validation cohort. Compared to national models, a customized model demonstrated superior readmission risk profiling by distinguishing a high-risk (38.3%) from a low-risk (9.4%) quartile.

Conclusions: A customized model improved readmission risk discrimination from HF hospitalizations compared to national models.

Keywords: Heart failure; Model; Prediction; Readmission.