Artificial Intelligence for Healthcare Workforce Support in the United States: A Systematic Review and Explainable Machine Learning Analysis of Physician Shortages and Healthcare Access

Abstract

The uneven distribution of healthcare professionals across the United States remains an important challenge for healthcare access, particularly in medically underserved and geographically isolated communities. At the same time, advances in artificial intelligence (AI), including machine learning, natural language processing, and generative AI, are creating new possibilities for supporting healthcare professionals and improving the efficiency of healthcare delivery. However, evidence connecting documented healthcare workforce shortages with the potential application of AI remains fragmented. This study aims to examine the potential role of AI in supporting the US healthcare workforce by integrating systematic evidence from the AI literature with real-world healthcare shortage and access data.A systematic review will be conducted to identify studies investigating AI applications relevant to healthcare workforce support, clinical workflow, administrative activities, healthcare access, and resource allocation. For the empirical component, publicly available Health Resources and Services Administration (HRSA) datasets will be analyzed, including Primary Care Health Professional Shortage Areas (PC-HPSA), Mental Health HPSA, Dental Health HPSA, and Medically Underserved Areas/Populations (MUA/P). The datasets contain information on HPSA designation and status, HPSA scores, provider-to-population ratios, estimated underserved populations, poverty levels, rurality, geographic characteristics, and medically underserved area indicators. These variables will be integrated at an appropriate geographic level to characterize patterns of healthcare workforce pressure and limited access to care across the United States.Machine-learning models will subsequently be developed to identify geographic and population-level patterns associated with healthcare shortage and underserved status. Multiple classification and ensemble learning algorithms will be compared using appropriate discrimination, calibration, and classification measures. Explainable artificial intelligence techniques, including SHAP-based interpretation, will be used to identify the factors contributing most strongly to model predictions. Findings from the systematic review will then be mapped to the empirical patterns to identify healthcare contexts in which AI-supported technologies, such as automated documentation, clinical text processing, scheduling, decision support, and other workflow applications, may warrant further evaluation.This integrated framework will distinguish documented effects of implemented AI systems from data-driven identification of healthcare settings with potential need for AI-supported workforce interventions. The study is expected to provide a reproducible, evidence-based approach for examining how AI can complement healthcare workforce strategies and potentially support access to care in areas experiencing persistent provider shortages and medical underservice.