Improving Healthcare System Policy Library Efficiency: Policy Redundancy Analysis Using K-means Clustering

This capstone project addresses the challenge of policy library redundancy within healthcare settings, highlighting the need for more research and solutions in policy library management. With the increasing volume of policies at healthcare institutions, managing and eliminating redundant policies is crucial for operational efficiency and compliance. Methods: The project utilized artificial intelligence (AI) techniques, specifically sentence embedding and k-means clustering with large language models, to analyze and identify redundant policies, aiming for a 5% reduction in the policy library. A Python-script prototype was developed and tested in collaboration with the accreditation team to automate the labor-intensive process of identifying redundant policies. Results: The application of AI-assisted tools to an initial subset of 79 policies, used to test the clustering process, achieved a potential 13% reduction in policies. When applied to the entire library of 4,933 policies, the methods attained a potential 4% reduction. These results were produced with freely available services and software in under an hour, achieving outcomes comparable to those of a manual review, which required four people 225 hours over six months. Conclusions: The project shows that AI-assisted methodologies can significantly enhance efficiency through automation and reduce labor-intensive processes, demonstrating their potential and opportunities for further optimization. This research contributes to the emerging field of AI in healthcare administration, offering a scalable proof of concept for improving policy library management.