AI Health
Friday Roundup
The AI Health Friday Roundup highlights the week’s news and publications related to artificial intelligence, data science, public health, and clinical research.
June 20, 2025
In this week’s Duke AI Health Roundup: LLMs excel at differential diagnosis; mapping diabetes incidence with microplastics burden in coastal US; judge orders canceled NIH grants restored; fired ACIP members speak up; HHS documents on vaccine safety sharply criticized; humans put too much trust in LLM-generated output in healthcare scenarios; agentic AI tackles automation of systematic reviews; much more:
AI, STATISTICS & DATA SCIENCE
- “Our findings show that the most advanced LLMs are very accurate at diagnosing common medical problems, often achieving nearly perfect accuracy when provided with sufficient information. For complex cases, their accuracy in identifying the primary diagnosis improved as more detailed information became available, with the best models achieving over 80%. Importantly, we observed that these AI models are particularly adept at generating a comprehensive list of potential diagnoses (a differential diagnosis).” A research article by Dinc and colleagues published in JAMIA Open offers a comparative analysis of LLMs used to conduct clinical diagnosis in both common and more complex clinical cases.
- “…participants perceived the AI-generated responses as remarkably similar to those provided by doctors, rendering them unable to accurately differentiate between the advice given by the AI and that offered by a registered physician on the online health care platform HealthTap….In addition to participants’ inability to distinguish AI-generated responses from doctors’ responses, we found that participants evaluated AI-generated responses as almost equal to, if not better than, responses provided by doctors across all metrics.” An article published in NEJM AI by Shekar and colleagues examines the tendency of human users to repose excessive credulity in medical advice offered by AIs compared with human experts, regardless of actual accuracy.
- “This increase in traffic has been hard to anticipate because few respondents were actively tracking bot traffic prior to the bots triggering a crisis in their collection. Many respondents did not realize they were experiencing a growth in bot traffic until the traffic reached the point where it overwhelmed the service and knocked online collections offline.” A new report by the GLAM (Galleries, Libraries, Archives & Museums)-E Lab finds that the attentions of AI bots scouring public collections for data may be impeding access to them.
- “Systematic review workflows are often hindered by the time- and labor-intensive demands of screening and data extraction. In this study, we demonstrate that otto-SR, an end-to-end SR automation pipeline, powered by GPT 4.1 and o3-mini-high, can accelerate these steps without compromising performance.” A preprint by Cao and colleagues, available from medRxiv, presents findings from a study evaluating the use of a customized agentic AI designed to automate the arduous process of performing a systematic literature review.
BASIC SCIENCE, CLINICAL RESEARCH & PUBLIC HEALTH
- “In this analysis of 152 coastal counties, we found that coastal counties potentially exposed to very high marine microplastic levels based on adjacent ocean water samples collected from 2015 to 2020 around the United States were associated with a higher prevalence of T2D, CAD, and stroke after adjustment for the median age of county population, sex, race, ethnicity, physician access, and the baseline socioeconomic, infrastructure‐related and environmental vulnerabilities of the communities residing in these counties.” A study published in the Journal of the American Heart Association by Makwana and colleagues presents findings that show correlations between the burden of microplastics in coastal waters and the incidence of diabetes in coastal US counties (H/T @erictopol@bsky.social).
- “This report suggests that more permissive firearm ownership and firearm-related use laws were associated with increases in firearm mortality among US children. Increases occurred in both homicide and suicide, but were more concentrated among homicides, thereby increasing the usual predominance of homicide as the leading intent (ie, highest share) in firearm deaths among the nonadult population.” A research article by Faust and colleagues published in JAMA Pediatrics (and with additional background detail provided by Faust in a post at his Inside Medicine Substack) examines the correlation between pediatric deaths due to firearms and the existence of more or less restrictive state firearms laws.
- “Secretary Kennedy’s unilateral decision to rescind the COVID-19 vaccine recommendation for healthy children and healthy pregnant women is at variance with existing data. COVID-19 vaccines are safe and effective for young children…Children clearly benefit from a primary COVID-19 vaccine series. In addition, pregnant women with COVID-19 are more likely to be hospitalized and die than women of the same age with COVID-19 who were not pregnant.” In a viewpoint article for JAMA, vaccine experts Lawrence Gostin, Dorit Reiss, and Paul Offit sharply criticize the recent changes in recommendations for offering COVID-19 vaccines to children and pregnant women.
- “…cognitive, social, affective, and behavioral neuroscientific findings have a crucial role to play in elucidating the brain mechanisms behind why people are susceptible to misinformation, how people are conditioned online to share misinformation with others, and, finally, how neuroscience can help inform intervention efficacy by examining brain responses that predict real-world engagement with interventions and their targets.” A perspective article published in Cell Neuron by van der Linden and Cohen examines the neurological antecedents of acceptance of misinformation.
COMMUNICATION, Health Equity & Policy
- “District of Massachusetts Judge William Young told the attorneys that the case raises serious concerns about racial discrimination related to health and said some evidence points to potential discrimination against women’s health…The NIH cut nearly $3.8 billion in grants to U.S. institutions, with Massachusetts losing more than $1.2 billion, per estimates from the Association of American Medical Colleges.” Axios’ Steph Solis reports on a recent judicial ruling that orders the NIH to restore grant funding that was recently cut off on the basis of supporting research related to diversity, equity, and inclusion (DEI) issues in healthcare.
- “In two instances, the HHS memo makes claims that are actively refuted by the papers it cites to back them up. Both papers support the safety and effectiveness of covid vaccines for pregnant women….The HHS document says that another paper it cites found ‘an increase in placental blood clotting in pregnant mothers who took the vaccine.’ But the paper doesn’t contain any reference to placental blood clots or to pregnant women. If he were grading the HHS document, ‘I would give this an ‘F,’ Turrentine said. ‘This is not supported by anything and it’s not using medical evidence.’” A story by Jackie Fortier, jointly published by Kaiser Family Foundation News and NPR, describes reactions to a recent policy recommendations document supplied to Congress by DHHS Secretary Kennedy.
- “The abrupt dismissal of the entire membership of the ACIP, along with its executive secretary, on June 9, 2025, the appointment of 8 new ACIP members just 2 days later, and the recent reduction of CDC staff dedicated to immunizations have left the US vaccine program critically weakened….The termination of all members and its leadership in a single action undermines the committee’s capacity to operate effectively and efficiently, aside from raising questions about competence.” A group of authors comprising the recently dismissed CDC Advisory Committee on Immunization Practices (ACIP), responsible (along with a counterpart FDA committee) for advising the CDC on vaccines and vaccination, has published a viewpoint article in JAMA supporting the previous work of ACIP and criticizing recent changes imposed by the DHHS Secretary.
