Assessment of Health Care Compliance in Managing Pseudomonas Aeruginosa in Urinary Tract Infection Using Machine Learning Techniques

Pseudomonas aeruginosa poses a grave global threat to public health due to its escalating antibiotic resistance, resulting in an undue economic burden on vulnerable communities facing depreciated economies. This study assesses physician adherence to treatment protocols for urinary tract infections (UTIs) caused by Pseudomonas aeruginosa using machine learning algorithms. A retrospective dataset spanning January 2018 to December 2022 from a local Tripoli hospital in North Lebanon was analyzed. Non-compliance cases were classified based on factors impacting protocol adherence. The antimicrobial susceptibility test results were utilized to evaluate physicians' empirical treatment choices. Various algorithms, including Support vector machine, classifier chain, K-Neighbors classifier, Linear Regression, and Random Forest, were employed to build and train the machine learning model. The support vector machine algorithm exhibited the highest accuracy, accurately predicting antimicrobial susceptibility in 70% of cases. Results were categorized into three scenarios based on correlation with clinical protocol. Future research should explore additional features like antimicrobial usage patterns, resistance gene identification, susceptibility profiling, and environmental factors to enhance model performance and shed light on non-compliance factors.