ICU Risk Stratification Models Feasible for Use in Sub-Saharan Africa Show Poor Discrimination in Malawi: A Prospective Cohort Study.

Abstract

Critical illness disproportionately affects people in low-income countries (LICs). Efforts to improve critical care in LICs must account for differences in demographics and infrastructure compared to high-income settings. Part of this effort includes the development and validation of intensive care unit (ICU) risk stratification models feasible for use in LICs. The purpose of this study was to validate and compare the performance of ICU mortality models developed for use in sub-Saharan Africa. This was a prospective, observational cohort study of ICU patients in a referral hospital in Malawi. Models were selected for comparison based on a Medline search for studies which developed ICU mortality models based on cohorts in sub-Saharan Africa. Model discrimination was evaluated using the area under the curve with 95% confidence intervals (CI). During the study, 499 patients were admitted to the study ICU, and after exclusions, there were 319 patients. The cohort was 62% female, with the mean age 31 years (IQR: 23-41), and 74% had surgery preceding ICU admission. Discrimination for hospital mortality ranged from 0.54 (95% CI 0.48, 0.60) for the Universal Vital Assessment (UVA) to 0.72 (95% CI 0.66, 0.78) for the Malawi Intensive care Mortality Evaluation (MIME). After tenfold cross-validation, these results were unchanged. The MIME outperformed other models in this prospective study. Most ICU models developed for LICs had poor to modest discrimination for hospital mortality. Future research may contribute to a better risk stratification model for LICs by refining and enhancing the MIME.