ADS Capstone Chronicles Revised

1

Uncovering Healthcare Inefficiencies: A Data-Driven Solution for Market Saturation and Fraud Jessica Hin Applied Data Science Master’s Program Shiley Marcos School of Engineering / University of SanDiego jessicahin@sandiego.edu Samantha Rivas Applied Data Science Master’s Program Shiley Marcos School of Engineering / University of SanDiego samantharivas@sandiego.e du AmyOu Applied Data Science Master’s Program Shiley Marcos School of Engineering / University of SanDiego aou@sandiego.edu ABSTRACT This study aims to detect fraud, waste, and abuse (FWA) within the Centers for identifying the features that were most important for the model, along with the location information of the potential FWA candidate. Identifying the potential

Medicare & Medicaid Services (CMS) data. Leveraging machine learning models, the analysis focuses on data obtained from the CMS website spanning the years 2019 to 2023. These models were trained on the dataset's features, facilitating the identification of critical patterns and the predictive models to uncover potential irregularities. Ten machine learning models were evaluated to identify the most suitable fit for the dataset, with logistic regression serving as the baseline model. The models assessed include Ridge classifier, Lasso classifier, XGBoost, neural networks, quadratic discriminant analysis, bagging classifier, stochastic gradient descent classifier, Adaboost, and support vector machines. These models’ hyperparameters were tuned to ensure that the best parameters were used to optimize each model’s performance. Various performance metrics were used to determine the optimal model for the data. The bagging classifier had the best metrics and predictive power,

candidates for FWA is essential, as flagged cases can prompt audits that may reduce financial losses and enhance patient care. KEYWORDS

Machine learning, fraud detection, Medicare, Medicaid, classification 1 Introduction

Healthcare systems worldwide face significant challenges in efficiently allocating resources to mitigate FWA. This project uses the Market Saturation and Utilization State-County dataset to assess healthcare provider saturation relative to the beneficiary population across various regions ( Centers for Medicare & Medicaid Services Data , 2024). By identifying patterns of over- and under-utilization of

services, this analysis aims to reveal inefficiencies and potential FWA.

Healthcare fraud is a significant issue with broad impacts on individuals and businesses.

125