A new federal report found that predictive AI is becoming mainstream in hospitals — but it also underscored gaps in how it is adopted, applied and overseen across different facilities.
Here are five takeaways from the report:
Seventy-one percent of U.S. hospitals used predictive AI integrated with their EHRs in 2024, according to a September 2025 report from the Assistant Secretary for Technology Policy. The findings are based on data from the American Hospital Association’s 2023 and 2024 IT Supplement.
AI adoption varied by affiliation. While 86% of system-affiliated hospitals reported using predictive models, only 37% of independent hospitals did the same.
Hospitals using the leading EHR vendor saw the highest adoption rates at 90%, compared to 50% among hospitals using other systems.
Inpatient risk prediction was the most common application of predictive AI, but administrative uses are growing quickly. Year over year, use of AI for billing increased by 25 percentage points and for scheduling by 16 points.
Eighty percent of hospitals used models from their EHR vendor, while 50% used self-developed tools and 52% used third-party models. Hospitals sourcing AI externally were more likely to apply it to tasks such as billing and outpatient risk prediction.
Evaluation practices also differed: 82% assessed for accuracy, 74% for bias and 79% conducted post-implementation monitoring. Three-quarters said oversight was shared across multiple entities.
The post Are hospitals actually using AI? 5 things to know appeared first on Becker’s Hospital Review | Healthcare News & Analysis.
Health IT
