Two new clinical diagnostic models for Talaromycosis marneffei in AIDS patients within 48 h: evaluation of performance and comparison with previous models.

backgroundTalaromycosis marneffei (TSM) is one of the leading causes of death among patients with acquired immunodeficiency syndrome (AIDS). Currently, culture-based testing methods for TSM take 3 to 7 days, leading to diagnostic delays. Early and accurate methods are urgently needed to enhances prognosis. This study aimed to develop and validate two clinical diagnostic models for the early and rapid identification of Talaromycosis marneffei (TSM) among AIDS inpatients within 48 h.

methodWe included 600 AIDS inpatients with or without TSM, who were randomly assigned to a training cohort (N = 420) and an internal test cohort (N = 180). An additional 180 patients were enrolled as a temporal testing cohort. LASSO and logistic regression (LR) analyses were conducted to identify and test potential predictors for TSM in AIDS inpatients.

resultsEight variables (rash, lymphadenopathy or hepatosplenomegaly, CD4 + T cell counts, adenosine deaminase levels, aspartate aminotransferase levels, white blood cell counts, procalcitonin levels, and galactomannan [GM] levels), which can be obtained from 24 to 48 h, were identified as significant predictors of TSM. Then a Basic model was established using the most basic clinical symptoms and biochemical markers, and a GM model was constructed by incorporating GM testing into the Basic model. Both models exhibited excellent discrimination ability with an area under the curve (AUC) of 0.88 (95% CI: 0.85–0.91) and 0.93(95% CI: 0.90–0.95) in the training cohort respectively. In the internal test cohort, the Basic model and the GM model yielded AUCs of 0.88 (95% CI: 0.83–0.93) and 0.92 (95% CI: 0.87–0.96), respectively. In the temporal test cohort, the Basic model and the GM model yielded an AUC of 0.81 (95% CI: 0.75–0.87) and 0.86 (95% CI: 0.80–0.92), respectively. Moreover, the Basic model demonstrated sensitivity and specificity comparable to those of previous models according to the DeLong test. The GM model outperformed the Basic model as well as the Mp1p and GM tests, with AUC, sensitivity, and specificity (p-values < 0.01).

conclusionsThis study successfully established and tested two TSM clinical diagnostic models suitable for regions with different levels of healthcare resources. The basic model has broader applicability and is suitable for environments with relatively limited medical resources. It can effectively improve the diagnostic performance for TSM in underdeveloped healthcare regions. Adding GM testing to the basic model enhances the diagnostic efficacy, offering a straightforward, practical, and quantitative approach for healthcare institutions capable of detecting GM.