ECRI names AI diagnostics as the top safety patient concern for 2026

The Emergency Care Research Institute (ECRI) publishes the Top 10 Patient Safety Concerns report each year. This list provides facilities with research-backed safety concerns facing healthcare systems across the nation. It also provides actionable steps to identify and improve safety culture where it is lacking.

Artificial intelligence (AI) continues to be an area of concern for patient safety. Last year’s concerns were around the lack of policies governing the use of AI: when it’s appropriate to use, what departments should use it, clinician verification of diagnoses and treatment plans, and the necessity of informed consent from patients. Many healthcare systems and facilities have still not developed guidance on these points. 

This year’s top safety concern is the use of AI in the diagnostic setting. In theory, AI has the potential to reduce diagnostic errors and the complications from missed, delayed, or bad diagnoses. But as the ECRI report states, “AI systems are only as good as the algorithms they use and the data on which they are trained.” Problems arise when the data used to train AI systems is incomplete, inaccurate, or lacking in diversity. It has been well documented that medical bias is a patient safety issue for marginalized patient populations, and AI systems trained with this biased data can arrive at inaccurate diagnoses that make healthcare disparities worse.

Another major concern is the impact of AI use on clinicians. According to the ECRI report, “Over time, overreliance on AI can erode people’s critical thinking skills.” Many are concerned that physicians who rely too heavily on AI to assist with diagnoses may not develop their own diagnostic skills, or may find their existing skills diminishing over time.

According to the report, AI use by physicians increased 74% from 2023 to 2024. It can be very easy for physicians to become overreliant on AI systems for diagnoses, especially as burnout in healthcare workers is still a major concern across the country. However, ECRI recommends a balanced approach: “Clinicians must view it as a tool designed to supplement and support clinical expertise—not replace it.”

Facilities and healthcare systems that are implementing AI should develop usage policies, guidelines, and procedures. As with any new tool, training is as important as ongoing assessment. Developing a plan for how to integrate AI systems will help ensure accurate and proper utilization.

Editor’s note: This article was excerpted from out Patient Safety Monitor Journal newsletter.