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.

April 4, 2025

In this week’s Duke AI Health Friday Roundup: teachers fret about AI impact on critical thinking; potential for coordinated multi-agent healthcare AI; the role of creative pursuits in shoring up mental health; avoiding condescension in the clinic; imaging AI shows disparate diagnostic performance; RCT evaluates therapeutic chatbot; correlations between wealth and health in US and Europe; rolling out an LLM-based documentation support tool; much more:

AI, STATISTICS & DATA SCIENCE

This image shows a blue xray image of a person's chest - it shows the ribs, the faint outline of a heart, and other organs. The image features yellow squares surrounding the organs (left lung, trachea, right lung, heart, diaphragm) - each of them feature labels such as: normal, midline, normal size. There is text in the right bottom corner in white which states 'no abnormalities detected'. Image credit: Elise Racine / https://betterimagesofai.org / https://creativecommons.org/licenses/by/4.0/
Image credit: Elise Racine / https://betterimagesofai.org / https://creativecommons.org/licenses/by/4.0
  • “Compared with the patients’ doctors, the AI model more often failed to detect the presence of disease in Black patients or women, as well in those 40 years or younger. When the researchers looked at race and sex combined, Black women fell to the bottom, with the AI not detecting disease in half of them for conditions such as cardiomegaly, or enlargement of the heart. These disparities persisted when the team tested CheXzero using four other public data sets of chest x-rays from other regions, including Spain and Vietnam.” An article by Science’s Rodrigo Pérez Ortega unpacks a recent study by Yang and colleagues published in Science Advances that found substantial differences in the performance of clinical imaging AI models based on the sex and race of patients.
  • “We would recommend that adopters of AI-supported tools engage in deliberate inquiry into quality and safety using multipronged strategies that are customized to the risk of the use case, with rapid-cycle feedback loops to integrate learnings into practice and executive deliberations backed by data.” A case study published in NEJM AI by Cain and colleagues describes Kaiser Permanente’s experience in rolling out an LLM-assisted clinical documentation tool.
  • “Multi-agent systems for healthcare are likely to be the next paradigm in medical AI. We expect that it is likely that within the next decade there will be AI agents that monitor sensor data, analyse medical histories, provide real-time insights and regularly collaborate with each other to provide care and to monitor procedures and outcomes. In some cases, the agents will communicate directly with the physician or the patient to offer guidance, answer questions or provide emotional support. A commentary article recently published in Nature Biomedical Engineering by Moritz, Topol, and Rajpurkar explores the potential for multiple coordinated healthcare AI agents to advance the potential for AI in healthcare.
  • “As the first RCT of its kind, our study supports the feasibility, acceptability, and effectiveness of a fine-tuned, fully GenAI–powered chatbot for treating mental health symptoms. Users demonstrated sustained engagement and rated their alliance with Therabot as comparable to human therapists during the 4-week trial….Overall, results from the Therabot RCT are highly promising. We found high engagement and acceptability of the intervention, as well as symptom decreases while maintaining a therapeutic alliance comparable to that of human therapists and their patients.” A research article published in NEJM AI by Heinz and colleagues report findings for a  randomized trial of a chatbot-based mental health intervention.

BASIC SCIENCE, CLINICAL RESEARCH & PUBLIC HEALTH

A paint-spattered artist’s palette, a disorderly pile of acrylic or oil paint tubes, and a heap of brushes and other art supplies lie on a canvas drop cloth. Image credit: Rebecca R/Unsplash
Image credit: Rebecca R/Unsplash
  • “In the end, the answer is yes, delving into a creative pursuit is profoundly helpful in coping with stress, as it brings with it creativity, mastery, and purpose, while silencing mind noise. However, there is no substitute for taking inventory of all five life domains, identifying those that could use a tune-up, and putting energy into bringing all of them into balance.” In a blog article for Psychology Today, Sherry Pagoto unpacks the potential mental health benefits of creative activity – but notes that it’s not a panacea.
  • “Importantly, the analysis revealed a significant reduction in new dementia diagnoses in those eligible for vaccination. After accounting for the fact that not all who qualified for the vaccine received it, the authors estimated that vaccination led to a 20% relative reduction in dementia risk.” A news article by Nature’s Anupam B. Jena explores recent research findings that suggest that the shingles (herpes zoster) vaccine may help prevent dementia.
  • “We found that wealth was associated with mortality across the United States and Europe and that the difference in mortality between the top and bottom quartiles of wealth appeared to be larger in the United States than in Europe. Mortality in the United States was higher than in Europe, even at higher wealth levels.” A research article by Machado and colleagues, published in the New England Journal of Medicine, examines correlations between wealth and health outcomes in the US ad Europe.
  • “In nearly one-fourth of the reviews, the primary study’s retraction occurred before the systematic reviews’ publication, which could have been avoided by verifying the article’s status on the publisher’s page or on Retraction Watch….On average, the impact of the retracted studies was modest, as recalculated effect estimates in most meta-analyses remained within the 95% CIs of the original estimates. However, including retracted studies in the analyses changed effect estimates, mainly in favor of the intervention. In addition, 68% of the 50 reviews reported erroneous effect estimates (95% CIs) in their abstracts and 14% would have meaningful changes that could affect interpretation after excluding the retracted study.” A research article published in JAMA Internal Medicine by Possamai and colleagues reports findings from a meta-analysis that scrutinized the frequency with which retracted studies appear in systematic reviews and meta-analyses.

COMMUNICATION, Health Equity & Policy

Students at computers with screens that include a representation of a retinal scanner with pixelation and binary data overlays and a brightly coloured datawave heatmap at the top. Kathryn Conrad / https://betterimagesofai.org / https://creativecommons.org/licenses/by/4.0/
Image credit: Kathryn Conrad / https://betterimagesofai.org / https://creativecommons.org/licenses/by/4.0
  • “Parnaby, who teaches AP Language and Composition, noted the AP test her students take emphasizes the “concept of a line of reasoning,” requiring students to demonstrate critical thinking by constructing logically flowing arguments in their essays. Relying on AI chatbots risks atrophying critical thinking muscles and not developing the ability to produce those kinds of argumentative essays.” An Axios summary by Ivana Saric explores educators’ concerns that rapidly spreading use of LLMs by students is affecting their ability to engage in critical reasoning and argument.
  • “Afterwards, I realised what the smile implied. It was something like: ‘You’re old and I’m young, so your knowledge is likely to be out of date and inferior to mine.’ It may also have meant: ‘You may be a doctor, but you’re not the doctor in this consultation.’ Either way, I’m afraid that its effect on me was to make the doctor’s effort at charm seem like a thin veneer for patronising me and disqualifying my lived and professional experience.” An opinion article published in BMJ by John Launer warns clinicians against inadvertently patronizing patients through subtle but important body-language cues.
  • “Even as AI specialists grapple with the conceptual and social consequences of building non-deterministic systems, their rhetoric about these systems’ role in society is trapped by the dominant paradigm of their industry. From the moment that ChatGPT was launched, technologists pronounced inevitable futures. Rather than being challenged, their deterministic prognostications were reinforced by journalists, companies, scholars, and policymakers, all of whom scrambled in response to the idea that ‘AI will change everything.’” A transcript of a recent talk by Danah Boyd, available at Tech Policy Press, invites the audience to question deterministic framings about the inevitable adoption of AI technologies (or particular forms of AI technologies).

Sites@Duke Express is powered by WordPress. Read the Sites@Duke Express policies and FAQs, or request help.