Machine learning-based detection of medical service anomalies: Kazakhstan’s health insurance data

Abstract

Background. With the exponential growth of medical data and limited analytical resources, healthcare systems are increasingly adopting Artificial Intelligence (AI) and Machine Learning (ML) technologies to enhance their decision-making processes. This research aims to apply advanced ML algorithms to analyze data from the Republic of Kazakhstan’s Obligatory Health Insurance Fund (OHIF) and automatically detect anomalies in the structure of delivered medical services. Methods. An automated AI system was developed and tested using nine ML models, including XGBoost, Random Forest, Decision Tree, Gradient Boosting, etc. The dataset comprised 329,584 real records, including demographic and socio-economic parameters. Model performance was evaluated using accuracy, precision, recall, F1-score, and the area under the curve (AUC). ROC and PR curves were used for visual validation. Results. Among the tested models, XGBoost and XGB_grid_model achieved the highest performance, with an accuracy of 93.2% and 93.6%, respectively. Precision: 91.8% and 92.1%, Recall: 90.4% and 91.3%, F1-score: 91.1% and 91.7%, AUC: 0.874 and 0.882. These models reliably detected irregularities such as billing duplications, out-of-pattern service provision, and inconsistencies with demographic profiles. Conclusion. The results demonstrate the feasibility of using ML for automated medical billing control. This approach can significantly enhance the transparency, accuracy, and accountability of healthcare financing in Kazakhstan, laying the groundwork for broader AI integration in national health systems.

Keywords

• artificial intelligence
• machine learning
• data science
• health insurance
• predictive modeling
• healthcare analytics
• kazakhstan