Multivariate predictive model for predicting in-hospital mortality in HIV-associated talaromycosis: a multicenter retrospective study.

BACKGROUND

Talaromycosis is an invasive fungal infection that predominantly affects immunocompromised individuals, with a particularly high incidence and mortality rate among HIV-infected patients. The purpose of this study was to develop and validate a novel nomogram model to predict mortality risk in HIV-associated talaromycosis (HTM) patients.

METHOD: The authors retrospectively analyzed HTM patients from January 2013 to December 2023 at three research centers. The research participants were randomly divided into the training and validation sets at a ratio of 7:3. To determine the crucial variables for establishment of the predictive model, the study sequentially applied univariate logistic regression, lasso regression, stepwise logistic regression. The validation set was used to assess the performance of the established prediction model, with its efficacy evaluated through receiver operating characteristics curve, clinical decision curves, and calibration curves.

RESULT: A total of 431 subjects were enrolled in the study with 55/431 (12.76%) patients dying during hospitalization. Statistical analysis shows that there was no difference between training set and validation set in the baseline demographic and clinical characteristics. Five factors including breathlessness, elevated TB, APRI, CRP and decreased Hb were identified as predictive factors for HTM mortality. A nomogram model was built and the area under the curve (AUC) for the nomogram in predicting death was 0.83 (95% CI: 0.76-0.90) in the training set and 0.81 (95% CI: 0.70-0.93) in the validation set. The H-L test and calibration curves showed a strong alignment between predicted and actual results in both sets. Additionally, the decision curve analysis (DCA) indicated that the model provided significant net benefits for patients experiencing poor outcomes.

CONCLUSION

The nomogram model developed in this study integrating easily accessible clinical indicators and symptoms is effective in predicting in-hospital mortality in patients with HTM, which will greatly assist clinicians in the individual management of HTM patients.