Adoption of Machine Learning in US Hospital Electronic Health Record Systems: Retrospective Observational Study

Comprehensive Summary

This retrospective observational study examines the way machine learning (ML) has been integrated into electronic health record (EHR) systems across U.S. hospitals and what types of hospitals are most likely to adopt these tools. Huang et al. analyzed data from 2,562 general acute care hospitals using linked American Hospital Association surveys collected between 2022 and 2024 to explore the relationship between hospital characteristics and ML adoption. They applied statistical models in order to identify how organizational and environmental factors influenced ML use, while also adjusting for non-responsive hospitals to reduce potential bias. The study found that 75% of U.S. general acute care hospitals had implemented a form of ML within their EHR systems by 2024. Most hospitals that adopted ML used it for both clinical purposes, such as predicting patient risks, and operational tasks, like scheduling or billing. Larger hospitals in metropolitan areas were significantly more likely to adopt ML. Additionally, hospitals working with major EHR vendors showed much higher adoption rates, suggesting that vendor support plays a major role. The discussion emphasizes that ML integration is strongly tied to available resources, which raises concerns about inequities between well and under-resourced hospitals. Huang et al. also highlight the limited evaluation and oversight of ML tools, emphasizing the need for stronger regulation and clearer standards to ensure these technologies are used safely and effectively in health care settings.

Outcomes and Implications

This research is significant as ML tools are increasingly becoming prevalent in clinical decision-making and hospital settings, yet not all hospitals have equal access to these advancements. ML integration into EHRs has the potential to enhance early disease detection, identify patient risk factors, and support follow-up care, all of which can directly improve patient outcomes. However, uneven application of ML resources means these benefits may not reach patients in smaller or rural hospitals. As ML use continues to expand over the next several years, targeted policy and public health efforts are necessary to ensure that these tools are both efficient and equitably distributed across health care settings.