An assessment of race and gender-based biases among readmission predicting tools (HOSPITAL, LACE, and RAHF) in heart failure population

Abstract

The objective of our study is to retrospectively investigate if the HOSPITAL score, LACE index, and RAHF scale exhibit any bias based on gender and race in heart failure readmissions. This is a retrospective cohort study with all adult medical patients discharged with congestive heart failure from 2016 to 2018 from Southern Illinois University School of Medicine Hospitalist service. The receiver operating characteristic (ROC) curve was constructed comparing prediction tools (HOSPITAL score, LACE index, and RAHF scale) performance based on gender and race by measuring the area under the curve (AUC). Absolute Between-ROC Area (ABROCA) values were calculated. All statistical analyses were performed using R version 3.6.2. The performance of the HOSPITAL score in the majority and minority population showed a statistically significant difference between AUCs (0.714 and 0.633, p = 0.029) and an ABROCA of 0.081 indicating superior performance in predicting hospital readmissions in the majority group vs. the minority. The performance of RAHF score in females and males showed statistically significant differences between AUCs (0.567 and 0.527, p = 0.04) and an ABROCA of 0.04 indicating the superior performance of the RAHF score in females compared with males. Our study demonstrated that the HOSPITAL score and the RAHF scale showed significant differences in predicting 30-day readmissions risk based on race and gender, respectively, in heart failure patients, whereas the LACE index did not show any significant difference.