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 19, 2024

In this week’s Duke AI Health Friday Roundup: TRIPOD-AI statement covers best practices for reporting on AI prediction models; taking stock of real-world effectiveness of RSV vaccination; totting up the balance sheet for generative AI; the importance of inclusivity in design decisions; imputing missing covariate data; AI hunts down source of metastatic cells; multiple nations gear up to battle smoking and vaping; much more:


Multiple black tripods with cameras on top are clustered together in front of a desert rock formation, with blue skies in the background. Image credit: Richard Lee/Unsplash
Image credit: Richard Lee/Unsplash
  • “AI is changing research. To ensure that it has value and importantly does not harm people or exacerbate existing inequities, we need transparency. The TRIPOD+AI reporting guidelines have been developed to help guide the writing of research ensuring that AI is fit for purpose and the findings presented in a usable format.” In an editorial published in the BMJ, Gary S. Collins hails the publication (in that same issue) of the TRIPOD+AI statement that offers guidance to authors for reporting on clinical prediction models that employ regression or AI.
  • “If enthusiasm for GenAI dwindles and market valuations plummet, AI won’t disappear, and LLMs won’t disappear; they will still have their place as tools for statistical approximation. But that place may be smaller; it is entirely possible that LLMs on their own will never live up to last year’s wild expectations. Reliable, trustworthy AI is surely achievable, but we may need to go back to the drawing board to get there.” At his Substack page, Gary Marcus questions whether the era of rapid progress with large language models may have run its course, or nearly so.
  • “Some stealthy cancers remain undetected until they have spread from their source to distant organs. Now scientists have developed an artificial intelligence (AI) tool that outperforms pathologists at identifying the origins of metastatic cancer cells that circulate in the body. The proof-of-concept model could help doctors to improve the diagnosis and treatment of late-stage cancer, and extend people’s lives.” Nature’s Smriti Mallapaty reports on a new AI application designed to track down the sources of metastatic cancer cells.
  • “We examine deterministic imputation (i.e. single imputation with fixed values) and stochastic imputation (i.e. single or multiple imputation with random values) methods and their implications for estimating the relationship between the imputed covariate and the outcome. We mathematically demonstrate that including the outcome variable in imputation models is not just a recommendation but a requirement to achieve unbiased results when using stochastic imputation methods.” A research article by D’Agostino McGowan and colleagues, published in the journal Statistical Methods in Medical Research, examines best practices in imputing missing covariate data in deterministic and stochastic models.


A man wearing glasses and a polo shirt is coughing, covering his mouth with a fist while reaching out his other palm in a “stop” gesture. Image credit: Towfiqu barbhuiya/Unsplash
Image credit: Towfiqu barbhuiya/Unsplash
  • “The W.H.O.’s previous stance was that only a handful of pathogens — those that travel in small droplets and spread across long distances, like tuberculosis — could be considered airborne. But the new report suggests broader categories that do not rely on droplet size or distance spread. Such changes were contentious because they raised the prospect that more diseases might now demand costly control measures, such as hospital isolation rooms and protective gear.” In an article for the New York Times, Carl Zimmer reports on the World Health Organization’s reassessment of the definition of “airborne” pathogens.
  • “The early introduction phase of newly licensed RSV immunizations in the US poses multiple challenges to the assessment of real-world effectiveness and impact. Low immunization uptake among the recommended groups during the 2023-2024 season may hinder assessments of immunization effectiveness using traditional observational study designs….In addition, there may be differential uptake within groups recommended for immunization, with some subgroups more likely to be immunized than others.” A perspective article published in JAMA by authors from the Centers for Disease Control and Prevention makes a case for real-world studies to more accurately determine the effectiveness of RSV vaccination.
  • “On 16 April, UK lawmakers backed one of the world’s most ambitious plans — to create by 2040 a ‘smoke-free’ generation of people who will never be able to legally buy tobacco. The proposal is now a step closer to becoming law. The UK, Australian and French governments are also clamping down on vaping with e-cigarettes. These countries’ bold policies are currently in the minority, say researchers, but such measures would almost certainly prevent diseases, as well as save lives and billions of dollars in health-care costs.” Nature’s Carissa Wong examines a flurry of international activity aimed at curbing rates of smoking and vaping.

COMMUNICATION, Health Equity & Policy

Pixelated images of pregnant women running on a race track, dynamically posed but distorted due to limited data availability. They are holding cards that read “error cannot generate,” symbolizing the struggle to accurately represent such dynamic figures due to inadequate online resources. Image credit: Amritha R Warrier & AI4Media / Better Images of AI / error cannot generate / CC-BY 4.0
Image credit: Amritha R Warrier & AI4Media / Better Images of AI / error cannot generate / CC-BY 4.0
  • “…design is not a standalone subject. Designers bring themselves and their lived experience to the drawing board. And they can’t draw what they don’t know. Design is an iterative process of failures followed by incremental successes that lead to the final product, so siloed training and lack of diversity can dangerously limit engineers’ inquiries, data, and visions. This has been true for decades and, unfortunately, continues to be the case…” An opinion article in STAT News by Catherine M. Klapperich argues forcefully for “pulling out a chair” for people who will be affected by research and design decisions – but who have often not been granted a seat at the decision-making table.
  • “I, like many others who have experimented with or adopted these products, have found that these tools actually can be pretty useful for some tasks. Though AI companies are prone to making overblown promises that the tools will shortly be able to replace your content writing team or generate feature-length films or develop a video game from scratch, the reality is far more mundane: they are handy in the same way that it might occasionally be useful to delegate some tasks to an inexperienced and sometimes sloppy intern.” In an essay at her Citation Needed blog, Molly White turns a critical eye on the cost/benefit balance of generative AI.
  • “Speakers invited by the committee discussed the current state of racial and ethnic health care disparities in the U.S., highlighted major drivers of health care disparities, provided insight into successful and unsuccessful interventions, identified gaps in the evidence base and proposed strategies to close those gaps, and considered ways to scale and spread effective interventions to reduce racial and ethnic inequities in health care. This workshop series is part of an ongoing consensus study examining the current state of racial and ethnic health care disparities in the U.S.” A new report from the National Academies Press highlights meeting proceedings that examined the ongoing problem of racial and ethnic disparities in healthcare.