Validation of Predictive Score of 30-Day Hospital Readmission or Death among Patients with Heart Failure

Abstract

Existing prediction algorithms for the identification of heart failure (HF) patients at high risk of readmission or death after hospital discharge are only modestly effective. We sought to validate a recently developed predictive model of 30-day readmission or death in HF using an Australia-wide sample of patients. This study used data from 1046 HF patients at teaching hospitals in five Australian capital cities to validate a predictive model of 30-day readmission or death in HF. Besides standard clinical and administrative data, we collected data on individual socio-demographic and socio-economic status, mental health (PHQ-9 and GAD-7 score), cognitive function (MoCA score), and 2D echocardiograms. The original sample used to develop the predictive model and the validation sample had similar proportions of patients with an adverse event within 30 days (30% vs 29%, p=0.35) and 90 days (52% vs 49%, p=0.36). Applying the predicted risk score to the validation sample provided very good discriminatory power (C-statistic=0.77) in prediction of 30-day readmission or death. This discrimination was greater for predicting 30-day death (C-statistic=0.85) than for predicting 30-day readmission (C-statistic=0.73). There was little difference in the performance of the predictive model among patients with either LVEF<40% or LVEF≥40%, but an attenuation in discrimination when used to predict longer-term adverse outcomes. In conclusion, our findings confirm the generalizability of the predictive model that may be a powerful tool for targeting high-risk HF patients for intensive management.