Smart healthcare informatics and AI-based MHAMFD framework for fraud detection in the American health system

Abstract

The digital transformation of the American healthcare system is increasingly reliant on advanced artificial intelligence (AI) to address the twin challenges of predictive analysis and fraud detection. This review synthesizes literature from 2019 to 2025 to map the evolution of data-driven healthcare, from fundamental predictive analytics to advanced fraud detection. Early applications of Logistic Regression and Random Forests have evolved with the integration of deep learning tools, such as Convolutional Neural Networks (CNNs) and Reinforcement Learning (RL). This progression empowers more sophisticated, real-time decision-making, moving beyond simple forecasting to dynamic treatment protocols that imitate clinical precision and reduce mortality. Concurrently, the integration of AI with Internet of Medical Things (IoMT) devices provides remote patient monitoring, real-time diagnostics, and personalized care. For healthcare fraud detection, the field has progressed from rule-based systems to a new generation of AI models designed to tackle complex anomalies. Techniques like SMOTE and ROS have given way to more sophisticated approaches, including Graph-Based AI and advanced hybrid models (e.g., MHAMFD), which are better equipped to detect collusion, overutilization, and data tampering. However, the power of these models is constrained by standardization issues and data silos. The next frontier in this AI progression is defined by solutions that address these barriers. This review argues that federated learning, which allows models to be trained without centralizing sensitive patient data, and Explainable AI (XAI), which builds trust and transparency in black-box models, are essential for overcoming current limitations and ensuring an effective and equitable future for AI in healthcare.