The use of AI technologies to diagnose conditions tops nonprofit ECRI’s annual list of patient safety concerns, highlighting growing challenges in AI diagnostic patient safety. While such programs have potential to improve diagnostic speed and accuracy, their performance so far has been inconsistent.
Reduced access to rural health care, and increasing rates of preventable acute diseases, rank as the second- and third-highest concerns in this year’s report, released earlier in March. ECRI produces the list using insights from senior executives at integrated health systems, children’s hospitals, rural community health centers and national associations.
The list gives journalists a primer on current patient safety issues. It also helps identify emerging trends and generate story ideas or questions when reporting on patient safety and AI or other health information technologies. Each item includes suggested strategies that health care organizations can use to improve patient outcomes and system performance.
More on AI and diagnostics
Many health care organizations are turning to AI technologies in an effort to improve diagnostic efficacy and precision and to reduce the risk of incorrect, missed or delayed diagnoses, the authors wrote. AI has been successfully adopted in certain diagnostic radiology procedures for years, they noted.
However, “AI systems are only as good as the algorithms they use and the data on which they are trained, and the potential for errors remains a significant concern,” they cautioned. For example, some machine learning models failed to recognize 66% of critical or deteriorating health conditions in simulated cases, and certain cancers and rare diseases may be harder for AI to detect in radiology studies. Placing too much trust in AI models to diagnose patients, without factoring in clinician experience, can lead to misdiagnosis — the very problem AI was intended to solve, the authors said.
Recommendations for safe AI use
For AI to be used effectively in diagnosis, clinicians must view it as a tool to support but not replace clinical expertise, they said. The authors recommended several actions for hospitals and health systems, including establishing AI usage policies and guidelines, ensuring staff are trained on proper use of AI systems, disclosing the use of AI to patients and allowing them to opt out, and fostering a culture where staff can speak up if issues with AI technologies occur.
One area where AI could play a significant role as tools are refined is in helping clinicians fill in knowledge gaps, said Divvy Upadhyay, M.D., M.P.H., diagnostic safety program leader at ECRI, during a March 20 webinar about the report. “The truth about medicine is it’s not black and white — it’s very gray, so clinicians and systems have to deal with a lot of uncertainty” about diagnoses, he said. “That’s where clinicians need a lot of help… It’s just impossible for any one human being to know everything about everything in medicine.”
Additional patient safety concerns
Reduced access to rural health care and the potential to increase health risks and disparities was No. 2 on the list. More than 80% of U.S. counties lack proper access to health care services such as pharmacies, primary care providers, hospitals, trauma centers, and low-cost health centers, the authors wrote. These health care deserts disproportionately affect rural communities as financial strains lead to widespread closures of health care facilities and diminished services. This puts rural residents at higher risk for worse health outcomes.
The No. 3 item this year was increasing rates of preventable acute diseases in communities and health care settings. Once largely controlled, illnesses such as measles are resurging due to factors such as declining immunization coverage, vaccine hesitancy, misinformation on evidence-based treatment and funding cuts to public health programs.
Rounding out the list
- Federal funding cuts hinder health care operations and safety (#4). Legislation passed by Congress in July 2025 included significant reductions to federal health care funding including a $1.2 trillion reduction in funding for Medicaid over the next 10 years and an estimated $491 billion reduction in funding for Medicare from 2026-2034, in addition to termination of National Institutes of Health grants to hospitals and medical schools worth several billion dollars. These actions are leading to discontinuation of services, potentially more people without insurance and layoffs.
- Lack of recognition and reporting of harm events (#5). Patient harm is one of the top causes of morbidity and mortality worldwide, according to the World Health Organization. However, health care systems only capture about half of harm events that occur and act on fewer, making it difficult to respond to, learn from, or mitigate the risk of adverse events, authors noted.
- Inadequate pain management for women (#6). Women face challenges in receiving adequate pain management due to structural and systemic barriers like implicit bias and inconsistent guidelines, the authors noted. Even when clinicians intend to remain impartial, they said, gender bias results in women’s pain being more frequently attributed to psychological or hormonal factors than to physical causes. Women also are less likely to receive analgesics than men.
- Persistent workforce shortages (#7). Staffing shortages are being seen across the industry, as a result of burnout, poor worker well-being, funding cuts and financial instability, the authors said. Shortages in behavioral health, obstetrics, long-term care and the Veterans Health Administration are of particular concern.
- Culture of blame hinders learning and improvement (#8). Health care workers say fear of being shamed or disciplined for reporting errors or unsafe practices is a key barrier to creating these reports. Organizations therefore lose opportunities to identify and rectify systemic issues, the authors said, putting patients at risk and creating a toxic work environment that makes continuous improvement impossible.
- Emergency Department boarding contributes to worse patient outcomes (#9). Boarding is a situation in which a patient should be admitted to the hospital for care but remains in the ED due to limited bed availability or staffing shortages. High boarding rates are an indicator of broader hospital- and system-wide patient flow, capacity and financial issues, the authors said. One analysis found that 40% of patients boarded in an ED for more than four hours, and 6.3% boarded for more than 24 hours. The experience contributes to issues such as increased readmissions, decreased patient satisfaction and perceptions of patient discrimination.
- Gaps in manufacturer packaging and labeling undermine medication safety (#10). Labels for medications produced by the same manufacturer may look alike at a quick glance. Look-alike products, strengths and concentrations of medication labels have contributed to up to 29% of events reported to the Institute for Safe Medication Practices’ Medication Errors Reporting Program. Confusion caused by packaging and labeling design has resulted in adverse events and patient deaths in products such as anesthetics and cardiac medications.
Resources
- Top 10 Patient Safety Concerns 2026 – a link to download the report and to view short videos featuring ECRI experts
- AI use in diagnostic care, rural care access, and surge in preventable diseases top annual report of patient safety concerns – an ECRI news release about the report
- Navigating the AI Diagnostic Dilemma – an ECRI webinar about the top patient safety concern
- Medical gaslighting tops list of highest patient safety risks – an AHCJ blog post focused on the 2025 report
