External Validation of the HIV In-Hospital Mortality Prediction (HIV-IMP) Risk Score

Abstract

Introduction: The HIV In-Hospital Mortality Prediction (HIV-IMP) score is a novel tool designed to predict in-hospital mortality among HIV-positive patients presenting to the emergency department. This study aimed to externally validate the HIV-IMP score and assess its generalizability. Methods: A prospective external validation study was conducted at the Chris Hani Baragwanath Academic Hospital in South Africa. A total of 1,153 HIV-positive patients were enrolled between May and October 2023. The primary outcome was in-hospital mortality. The HIV-IMP score was calculated using eight clinical and laboratory variables, and its performance was evaluated using discrimination (AUROC) and calibration (Hosmer– Lemeshow test). Results: In-hospital mortality was 15.2% (n=175). The HIV-IMP score demonstrated good discrimination (AUROC = 0.78, 95% CI: 0.74–0.81) and calibration (Hosmer–Lemeshow χ² = 5.16, p = 0.741). Mortality rates increased with higher scores, ranging from 1.1% for a score of 0 to 100% for a score of 10. Glasgow Coma Scale <15 and lactate >2 mmol/L emerged as the strongest predictors, consistent with the derivation cohort. The score performed particularly well at the extremes of predicted probabilities, with marginal over- or under-fit in intermediate deciles. Conclusion: The external validation confirmed the reliability of the HIV-IMP score for predicting in-hospital mortality among HIV-positive patients in a resource-limited setting. Its robust performance and practical design support its use as a valuable tool for clinical decision-making and resource allocation. Further research should explore its application in diverse settings and its impact on patient outcomes.