Machine Learning-Based Identification of Predictors Influencing Sub-Therapeutic Rifampicin Concentrations in HIV/TB Co-Infected Patients
Keywords: machine learning, TB/HIV
TL;DR: This study used machine learning to identify predictors of sub-therapeutic rifampicin concentrations among TB/HIV co-infected patients, highlighting the potential for personalized interventions to optimize treatment outcomes.
Abstract: Managing Tuberculosis (TB)/HIV co-infected patients poses challenges due to pill burden, compliance, and possible toxic effects. Identifying patients at risk of sub-therapeutic drug concentrations is crucial for guiding interventions. This study used machine learning to identify predictors of sub-therapeutic rifampicin concentrations in 268 TB/HIV co-infected patients from the SOUTH cohort profile. Two datasets were analyzed: the original and synthetic (2000 data points generated). The best-performing model, Random Forest Classifier, was fitted and evaluated through cross-validation. Most participants showed sub-optimal rifampicin concentrations. BMI, age, systolic blood pressure, and baseline culture result emerged as crucial predictors for both datasets. ML demonstrates potential in improving TB/HIV patient management, enabling personalized interventions like drug dosing adjustments and adherence monitoring to optimize treatment outcomes.
Submission Category: Machine learning algorithms
Submission Number: 31
Loading