AI-Driven Model for Fake Medical Providers Prediction: Enhanced Security for Hospitals

Abstract

 The growing senior population necessitates greater medical needs and incurs associated costs. Medicare is a healthcare program in the United States that offers insurance coverage mostly to persons aged 65 and above, aiming to alleviate some of the financial strain related to medical expenses. Nevertheless, healthcare expenses remain elevated and persistently rise. Fraud significantly contributes to the escalating healthcare costs. The predominant approach for carrying out the latter is manually scrutinizing claims data, which is a laborious and costly procedure. Machine learning models can significantly reduce auditing expenses by automatically examining incoming claims and identifying those that are considered suspect, meaning they may be wrong, for further manual auditing. This paper offers an extensive analysis utilizing machine learning techniques to identify fraudulent Medicare providers. This study utilizes publicly accessible Medicare data and provider exclusions for fraud categorizations to construct and evaluate three distinct machine learning models. In order to mitigate the effects of class imbalance, this framework utilizes Logistic Regression to establish two class distributions, considering the limited number of actual fraud labels available. Evidence indicates that the remaining algorithms exhibit inferior performance in comparison to Logistic Regression. Learners exhibit superior ability in detecting fraud, especially when dealing with class distributions of 80:20. They achieve high average AUC scores and demonstrate low false negative rates. This study effectively showcases the effectiveness of utilizing machine learning algorithms to identify instances of Medicare fraud.