Machine Learning Models for Predictive Risk Assessment in Healthcare IT Projects

Abstract

This integrative review focuses on the role of machine learning (ML) in improving predictive project risk management in healthcare information technology (HIT) environments. With the ever-increasing complexity of HIT initiatives, e.g. electronic health record (EHR) implementations and tele-medicine systems, the older risk assessment practices are proving inadequate because they are retrospective. ML presents an anticipatory solution, allowing one to analyse and predict various risks of the project in real-time. To investigate these issues, the study employed an integrative review methodology, which enabled the inclusion of diverse scholarly and technical literature spanning empirical studies, theoretical frameworks, and conference proceedings. A systematic search across major academic databases from 2013 to 2025 facilitated the thematic synthesis of ML models (e.g., Gradient Boosting, Random Forest, and SVM) and types of risks covered (technical, operational, strategic, and clinical), risk domains, implementation outcomes, and adoption barriers in healthcare IT project risk management. It was found that ML can be used to increase accuracy in project delay, cost overrun, and compliance prediction, alongside resource allocation optimisation greatly. Nonetheless, its actual use is still restricted by a lack of infrastructure, transparency of algorithms, and moral issues. As the Technology Acceptance Model (TAM) guides the study, the key factors identified that led to adoption are perceived usefulness and ease of use. It suggests multidisciplinary working, transparent model development, and sound ethical constructs as preconditions of effective implementation. The reviewed paper can be added to the developing debate around the use of AI in healthcare by providing practical implications to be used by researchers, system developers, and policymakers interested in implementing ML in risk-resistant HIT project management.