PREDICTIVE MODELING OF MALARIA DRUG RESISTANCE CASE STUDY GURATOPP RAYFIELD PRIMARY HEALTHCARE, PLATEAU STATE, NIGERIA

Résumé

The primary approach for treating malaria has been Artemisinin-based Combination Therapy (ACT), yet emerging resistance threatens its effectiveness. This study applies a logistic regression framework to predict malaria drug resistance among patients at Guratopp Rayfield Primary Healthcare Center. Using a data set of 500 patient records, predictors included age, type of diagnostic test, test result, drug type and dosage administered. The study explores the relationship between patients and resistance to malaria treatment, as well as, building a binary classification model on drug resistance among patients. From the malaria tests conducted, results showed that drug dosage and positive malaria test outcomes, were positively associated with resistance, while test type and negative results, were significant predictors of non-resistance. Logistic regression and GridSearch cross-validation both produced an accuracy score of 74% in classifying resistance status, outperforming alternative models. The model coefficients estimated via a maximum likelihood, highlight the statistical relationship between patient characteristics and resistance outcomes. Findings emphasize the importance of statistical modeling in malaria research, providing a replicable framework for monitoring resistance patterns. This approach offers healthcare providers and policymakers data-driven insights for adapting treatment protocols and enhancing surveillance of drug resistance.