| dc.description.abstract | Despite the availability and implementation of well-known efficacious interventions for TB prevention and treatment by the Ministry of Health, Uganda (MoH), treatment non-adherence persists. Moreover, identifying a specific patient at risk of non-adherence is still a challenge. Thus, this study set out to utilize machine learning modeling to explore individual risk factors predictive of treatment non-adherence in the Mukono district.
This was a retrospective study based on a record review of 838 TB patients enrolled in six health facilities (3 government, 3 private-not-for-profit) in the Mukono district. We developed and evaluated five (5) machine learning algorithms (Logistic regression (LR), artificial neural networks (ANN), support vector machines (SVM), Random Forest (RF), and AdaBoost) to explore the individual risk factors for tuberculosis treatment non-adherence. Finally, we developed a web-based interface prototype utilizing the best-performing model to aid clinicians in identifying a specific patient at high risk of non-adherence.
Of the five developed and evaluated models, SVM performed the best with an accuracy of 91.28 % compared to AdaBoost (91.05%), RF (89.97%), LR (88.30%), and ANN (88.30%) respectively. Individual risk factors predictive of non-adherence included; TB type, GeneXpert results, sub-country, ART status, contacts below 5 years, health facility ownership, sputum test results at 2 months, treatment supporter, CPT Dapson status, risk group, patient age, gender, MUAC, referral, positive sputum test at 5 months and 6 months.
This study shows that classification machine learning techniques can identify patient factors predictive of treatment non-adherence and accurately differentiate between adherent and non-adherent patients. Thus, tuberculosis program management should consider adopting machine learning techniques evaluated in this study as a screening tool to both identify and target-suited interventions for these patients. | en_US |
