Machine Learning Models for Predicting Latent Tuberculosis Infection Risk in Close Contacts of Patients with...

Area of Science:

• Public Health
• Infectious Disease Epidemiology
• Machine Learning in Healthcare

Background:

• Latent tuberculosis infection (LTBI) poses a significant global health challenge.
• Accurate risk assessment is crucial for effective LTBI management and prevention strategies.

Purpose of the Study:

• To identify risk factors associated with LTBI.
• To develop and validate a machine learning-based risk prediction model for LTBI.

Main Methods:

• A cohort of close contacts of active pulmonary tuberculosis (TB) patients in Henan Province, China, were assessed for LTBI.
• Epidemiological data were collected via questionnaires and tuberculin-purified protein derivative testing.
• Five machine learning models (LR, DT, RF, SVM, MLP) were trained and evaluated using metrics like MSE, AUC, and F1-scores.

Main Results:

• LTBI prevalence among close contacts was 50.5%.
• The Support Vector Machine (SVM) model demonstrated the best performance (MSE=0.121) and identified key predictors: contact type, key population status, residential area, group activity frequency, and etiological results.
• Internal validation showed high predictive accuracy (AUC=0.921), while external validation indicated moderate performance (AUC=0.752).

Conclusions:

• The developed SVM model effectively predicts LTBI risk using specific epidemiological factors.
• This model holds promise for targeted screening and management of individuals at high risk for LTBI.