Development and validation of a prognostic score to predict tuberculosis mortality

Area of Science:

• Public Health
• Epidemiology
• Medical Informatics

Background:

• Tuberculosis (TB) remains a significant global health challenge.
• Accurate prediction of mortality risk in TB patients is crucial for effective management.
• Existing prognostic tools may lack simplicity or broad applicability.

Purpose of the Study:

• To develop and validate a straightforward prognostic scoring system.
• To predict mortality risk in tuberculosis patients during treatment.
• To identify high-risk individuals for optimized resource allocation.

Main Methods:

• Utilized data from CDC's Tuberculosis Genotyping Information Management System (TIGIS).
• Developed and validated a logistic regression-based prognostic mortality scoring system.
• Included 9 demographic and clinical characteristics available at initial patient visits.

Main Results:

• The scoring system demonstrated excellent discrimination and calibration in both developmental (AUC 0.82) and validation (AUC 0.80) cohorts.
• Prognostic scores were categorized into low (<15), medium (15-18), and high (>18) risk groups.
• The model showed strong predictive performance with non-significant Hosmer-Lemeshow tests.

Conclusions:

• A validated TB prognostic scoring system is available.
• The system uses readily available demographic and clinical data.
• It serves as a practical tool for identifying high-mortality-risk TB patients.