Predicting the Risk of in-Hospital Mortality in Patients with HIV-Associated Talaromyces Marneffei Infection

Abstract

Disseminated Talaromyces marneffei infection (formerly termed penicilliosis) is the third most common microbiologically confirmed opportunistic infection in Southeast Asia, with mortality of up to 30% despite antifungal therapy. There are restrictive clinical algorithms to predict treatment outcomes. A total of 513 patients with microbiology-confirmed HIV-associated talaromycosis were included in the analysis. Poor outcome was observed in 143/513 patients (27.9%). In the univariate logistic regression analysis, hepatomegaly and splenomegaly were protective factors. Shorter duration of illness, higher respiratory rates, dyspnea, AIDS-associated central nervous system syndromes, platelet counts <50,000 cells/mL, aspartate transaminase (AST) >300 U/L, alanine transaminase (ALT) >150 U/L, serum creatinine >110 µmol/L were predictors of poor outcome. In the multivariate logistic regression analysis, shorter days of illness, higher respiratory rates, platelet counts <50,000 cells/mL, AST >300 U/L and serum creatinine >110 µmol/L, active tuberculosis (TB) and/or ongoing TB induction treatment and AIDS-associated central nervous system syndromes were independent predictors of poor outcome. The prognostic scores ranged from 0 to 19, corresponding to a mortality risk of 0% to 100%. The internal validation showed acceptable discrimination (AUC=0.68) and calibration slope (0.93). The Brier score for model performance was 0.14. We developed a simple scoring system that can predict the risk of death in patients with HIV-associated talaromycosis based on routinely measured characteristics on admission. The scoring system will be further externally validated using other cohorts in the region.