The AMIA Annual Symposium is underway in Atlanta this week, offering a mix of serious discussions and lighthearted networking opportunities. A highlight so far was the workshop on ethical considerations in AI for consumer health, which addressed algorithmic bias and the challenges of using outdated training datasets. Discussions also covered the limitations of EHR data and the importance of informed consent. Attendees debated AI's role in replacing clinicians and the need for new training methods for physicians. The symposium serves as a platform for both seasoned professionals and newcomers to exchange ideas and experiences.

Advance patient safety, quality, and leadership with Harvard Medical School's Safety, Quality, Informatics, and Leadership certificate program. Gain tools to improve outcomes, drive system-wide change, and lead data-driven health care improvement.

A CU Anschutz study led by Matthew DeCamp explores how patients use health care chatbots like the UCHealth portal's Livi, revealing that while these tools are popular for administrative tasks, they are also used for sensitive topics due to perceived anonymity. The study, published in the Journal of the American Medical Informatics Association, highlights the importance of trust and transparency in chatbot interactions, noting that older users are more comfortable sharing information than younger ones. DeCamp emphasizes the need for patient-centered design and digital literacy to ensure ethical use of chatbots in health care.

"Health care is built on trust and is crucial between clinician and patient, " Matthew DeCamp said. "But trust is fragile. That means if we make a mistake, it could affect trust in the broader health care system."

Researchers have introduced a tool, based on the PREVENT risk calculator, that uses sex- and age-specific percentiles to estimate 30-year cardiovascular risk, aiming to improve communication and motivate early prevention. "When a patient sees they are in the 90th percentile, we hope that this will serve as a wake-up call that risk starts early," said Sadiya Khan of Northwestern University. "We don't want to wait until it is too late, and someone has had an event." Research findings were published in the Journal of the American College of Cardiology.

Sanford Health is embedding AI into its electronic health records to enhance patient care by automating evidence-based recommendations. This integration has led to significant improvements, including doubling the number of patients with diabetes who receive chronic kidney disease tests and tripling the rate of early diagnoses. The AI tools, including a chronic kidney disease module and a forthcoming colon cancer screening tool, help clinicians personalize care and intervene earlier, which is crucial for patients in rural areas who face challenges accessing treatment. This initiative highlights the potential of AI to transform health care by providing timely insights without increasing cognitive burdens on physicians.

US and international authorities, including the Cybersecurity and Infrastructure Security Agency, the FBI and HHS, have updated a joint advisory warning about the Akira ransomware group's ongoing threat to critical infrastructure. Since March 2023, Akira has targeted sectors such as health care and public health, manufacturing, and education, using tactics like exploiting vulnerabilities and spear phishing. The group is associated with other notorious threat actors and has amassed significant ransomware proceeds. Authorities urge organizations to strengthen their cybersecurity measures to mitigate this pervasive threat.

Researchers have found that using AI can reverse the Dunning-Kruger effect, leading to a universal tendency to overestimate abilities regardless of actual skill. The study in Computers in Human Behavior, involving logical reasoning tasks with and without AI assistance, reveals that AI users engage in "cognitive offloading," reducing critical thinking and metacognitive monitoring.

At the American Society for Dermatologic Surgery 2025 Annual Meeting, speakers discussed the potential and challenges of AI in dermatology. Cheryl Burgess highlighted biases in AI databases, particularly the underrepresentation of skin of color, which affects the accuracy of AI in identifying lesions. Vivian Bucay noted the subjectivity in aesthetic dermatology outcomes, emphasizing the need for well-annotated datasets to reduce variability. Jane Yoo shared her experiences with AI tools, pointing out translation errors in patient records and the need for real-world validation. Despite these challenges, AI continues to offer promising applications in skin analysis and patient engagement.

A watchdog report has found that the NIH did not ensure robust cybersecurity protocols for the All of Us research program, which contains sensitive health data from over a million participants. The HHS Office of Inspector General report criticized the NIH for not adequately restricting data access and for failing to classify genomic data as high risk. NIH has accepted the recommendations and is taking steps to improve data security, including revising access controls and addressing national security concerns.

About 23% of US adults received a COVID-19 vaccination last season, compared with 47% who got a flu shot, according to the CDC, and health experts are concerned that vaccination rates may decrease even more this year due to conflicting advice and new, narrower federal recommendations. "A lot of misinformation is going around regarding COVID," said Alein Haro-Ramos, an assistant professor at the University of California, Irvine. "Vaccine hesitancy is going to increase."