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 7, 2023

In today’s Duke AI Health Friday Roundup: Coalition for Health AI releases blueprint for trustworthy AI; stark differences in US-UK life expectancy; US youth mortality rises, and COVID is not the only driver;  socioeconomic stigma among PhD students; open letter calls for “pause” on AI development; brain research adds to understanding of Alzheimer’s among Black patients; transgender persons may face increased healthcare costs, reduced access; taking stock of brain-computer interfaces; much more:


Closeup photograph of a blueprint construction diagram with a sharp pencil and ruler lying on top. Image credit: Sven Mieke/Unsplash
Image credit: Sven Mieke/Unsplash
  • The Coalition for Health AI – a group comprising experts in AI and healthcare drawn from both industry and academia, together with observers from government and regulatory agencies, has published the first iteration of its “Blueprint for Trustworthy AI in Healthcare.” Conceived as a living document that will change and grow over time, the Blueprint “addresses the quickly evolving landscape of health AI tools by outlining specific recommendations to increase trustworthiness within the healthcare community, ensure high-quality care, and meet healthcare needs.”
  • “By analysing where the authors’ affiliated institutions were located, the researchers found that 82% of authors were based in high-income countries, whereas just 0.5% were based in low-income countries. Articles led by researchers based in high-income countries were cited more than four times as much as those led by researchers in low-income countries.” Nature’s Carissa Wong reports on a recent study by Alberto and colleagues, available as a preprint from MedRxiv, that highlights the extent to which research on AI fairness in healthcare reflects an authorship that is primarily white, male, and from relatively affluent countries.
  • “…we call on all AI labs to immediately pause for at least 6 months the training of AI systems more powerful than GPT-4. This pause should be public and verifiable, and include all key actors. If such a pause cannot be enacted quickly, governments should step in and institute a moratorium….AI labs and independent experts should use this pause to jointly develop and implement a set of shared safety protocols for advanced AI design and development that are rigorously audited and overseen by independent outside experts.” An open letter signed by a number of prominent figures in the world of AI development calls for a “pause” on developing AI systems any further – a pause to be enforced by government action if voluntary measures are not sufficient. Opinion on the merits of the letter is divided in the AI community; here, a Twitter thread by Arvind Narayanan encapsulates some arguments against the letter’s recommendations.
  • “Overall, our early results show the immense promise as well as the dangers of using the system without further refinement of the methods – such as providing uncertainty estimates for low-confidence answers. Given their great promise, we need to conduct rigorous evaluations before we can rely routinely on these new technologies.” A web article at Stanford’s Human-Centered AI (HAI) institute by Dash, Horvitz, and Shah, offers a preview of a study (soon to be available from arXiv) that evaluates the use of large language models for answering clinical questions posed by healthcare providers.


Photograph of a child in a darkened room, with face only partly illuminated, peers through the gap between drawn blinds and the window, which allows a little light from outside through. Image credit: Alexander Grey/Unsplash
Image credit: Alexander Grey/Unsplash
  • “The increases in fatal injuries that preceded and may have been exacerbated by the COVID-19 pandemic mark a tragic reversal to years of progress in lowering pediatric mortality rates through advances in injury prevention (eg, safer automobiles, occupant restraints, bicycle helmets, smoke detectors) and the prevention and treatment of lethal pediatric diseases (eg, prematurity, neoplasms, congenital disorders). These advances have reduced pediatric deaths, but the recent increase in all-cause mortality means that these gains are now being entirely offset by injuries, primarily those involving violence, self-harm, and drug misuse.” A viewpoint article published in JAMA by Woolf, Wolf, and Rivara examines a concerning increase in mortality rates among US children and adolescents.
  • “…at *every* point on the income distribution, Americans live shorter lives than the English. It’s actually worse than that chart made out, because at most points on the income distribution, Americans earn much more than Brits. If we plot the same data by actual income instead of percentile, the US deficit is vast.” A set of data analyses (encapsulated here in a tweet thread, but also available as an article [subscription required) by Financial Times columnist John Burn-Murdoch shines a spotlight on the severity of recent declines in US life expectancy by comparing life expectancy across income distributions for both the US and UK.
  • “…because there’s been scant research in Black populations, and barely any brains from Black people without dementia to study, understanding these critical differences — and potential treatments they may lead to — remains a distant goal. Barnes isn’t convinced that a widely cited statistic, that Black Americans are twice as likely to get the disease, is true, because there’s a dearth of data from brains of Black people…” STAT News’ Usha Lee McFarling reports on efforts to improve the understanding of Alzheimer disease and cognitive decline among Black patients, who until recently have been excluded or significantly underrepresented in studies of the disease.
  • “The year 2023 promises to be significant for BCIs [brain-computer interfaces]. Multiple non-invasive systems are likely to enter the hands of researchers and clinicians, and several key clinical trials of implanted systems are scheduled to begin. To succeed, BCIs must establish themselves as safe and truly helpful to people with serious medical conditions.” A news feature at Nature Electronics by Liam Drew walks readers through a summary of recent progress in developing brain-computer interfaces.

COMMUNICATION, Health Equity & Policy

A hand holding a ball-point pen is poised over an open notebook of graph-ruled paper as if about to begin writing. Image credit: lilartsy/Unsplash
Image credit: lilartsy/Unsplash
  • “One day I’d had enough of writing the same sentence over and over. I decided to write the first draft of a manuscript as quickly as possible. That’s when something magical happened: The sentences seemed to flow out of me, and in a few hours I had a rough draft. I had to go back and edit my work later, but it was a relief to have text to work with—and to know that my draft could only get better.” In a perspective for Science, doctoral candidate Yunhe Tong writes about the struggle of writing in a language not one’s own, and a sudden epiphany that helped break through the impasse.
  • “Using longitudinal data from incoming PhD students at three North American universities, we aimed to better understand the role of socioeconomic status in doctoral students’ feelings of social integration in academic settings and interpersonal experiences inside and outside of academia. Over and above other demographic factors (i.e., international, URM [underrepresented minority] status and gender), we found that PhD students starting their graduate careers with relatively low SES experienced a lower sense of social integration into academia (i.e., higher perceived social isolation, lower belonging in graduate school, and less perceived support).” A research article by Park and colleagues published in the Journal of Social Issues explores stigma related to socioeconomic status among PhD candidates.
  • “…the goals of increasing diversity in clinical trial participation include earning and building trust, promoting fairness, and generating biomedical knowledge…These three goals are each valuable on their own, and accomplishing one goal might result in progress toward another. New initiatives designed to diversify clinical trials may not promote all three goals equally, however. We believe that earning trust and promoting fairness are the most important of these goals and provide the clearest opportunities for realizing value from increasing diversity and representativeness in trials.” A perspective article by Schwartz and colleagues, published in the New England Journal of Medicine, lays out an approach for increasing the diversity of clinical trial participants.
  • “Clinicians expect access to gender-affirming care to immediately decline with the expected requirement for more in-person examinations, according to Ducar. Requiring more in-person care means trans people will face the additional hurdle of spending even more money to access that care. In a recent survey from the Movement Advancement Project, nearly half of respondents who received gender-affirming care said they had spent $5,000 or more out-of-pocket, and one-third spent at least $10,000.” At STAT News, Theresa Gaffney reports on the potential fallout for the healthcare of transgender persons as COVID-era emergency policies that facilitated care access via telehealth wind down.