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.

July 12, 2024

In this week’s Duke AI Health Friday Roundup: what we mean when we talk about AI; links between race and air pollution mortality; concerns over lack of access to cardiologists, obstetrics; ChatGPT for literature reviews; an LLM tailored for mobile use; FTC turns critical eye on PBMs; coping with the mental health crisis in academic science; AI supremacy for certain tasks may be more vulnerable than thought; much more:

AI, STATISTICS & DATA SCIENCE

“AI is a catchall term for a set of technologies that make computers do things that are thought to require intelligence when done by people. Think of recognizing faces, understanding speech, driving cars, writing sentences, answering questions, creating pictures. But even that definition contains multitudes….And that right there is the problem. What does it mean for machines to understand speech or write a sentence? What kinds of tasks could we ask such machines to do? And how much should we trust the machines to do them?” In an article for MIT Technology Review, Will Douglas Heaven examines a basic but profound question: what, exactly, is AI, anyway?
“Essentially, as coding evolves, ChatGPT has not been exposed yet to new problems and solutions. It lacks the critical thinking skills of a human and can only address problems it has previously encountered. This could explain why it is so much better at addressing older coding problems than newer ones.” IEEE Spectrum’s Michelle Hampson explores a recent study that evaluates how well ChatGPT does at writing computer code.
“…research has revealed weaknesses in one of the most successful AI systems — a bot that plays the board game Go and can beat the world’s best human players — showing that such superiority can be fragile. The study raises questions about whether more general AI systems will suffer from vulnerabilities that could compromise their safety and reliability, and even their claim to be ‘superhuman’.” At Nature, Matthew Hutson describes recent work demonstrating that the supremacy that AI has demonstrated against humans in certain activities may be on shakier ground than previously thought.
“Contrary to prevailing belief emphasizing the pivotal role of data and parameter quantity in determining model quality, our investigation underscores the significance of model architecture for sub-billion scale LLMs….we establish a strong baseline network denoted as MobileLLM, which attains a remarkable 2.7%/4.3% accuracy boost over preceding 125M/350M state-of-the-art models….Moreover, MobileLLM model family shows significant improvements compared to previous sub-billion models on chat benchmarks, and demonstrates close correctness to LLaMA-v2 7B in API calling tasks, highlighting the capability of small models for common on-device use cases.” A preprint by Liu and colleagues, available from arXiv, describes the creation and testing of a high-performance large language model with fewer than a billion parameters, engineered for efficient use on mobile devices.
“Current literature is mostly opinion-driven, and there is limited published literature originating from the library and information profession. At present, limitations outweigh the strengths of ChatGPT for systematic literature searching, caution should be exercised, and human oversight is essential. More research is required, and information specialists and librarians are in a prime position to develop guidelines and share examples of best practice.” A review published by Parisi and Sutton in the Journal of the European Association for Health Information and Libraries weighs the evidence underpinning the desirability of using ChatGPT for systematic literature reviews.
“Rodman discusses the evolution of clinical reasoning, the importance of probabilistic models, the implications of AI in diagnostic processes, and details his work with large language models like GPT-4. He also reflects on the balance between the benefits and challenges of AI in medicine, emphasizing the necessity of collaboration between computer scientists and medical professionals.” NEJM AI’s Grand Rounds podcast interviews Harvard’s Adam Rodman.

BASIC SCIENCE, CLINICAL RESEARCH & PUBLIC HEALTH

“Almost one-half of U.S. counties do not have practicing cardiologists even though their residents experience greater prevalence of CVD risk factors and higher cardiovascular mortality. These counties with limited access to cardiologists tended to be rural and socioeconomically disadvantaged, with the mismatch between CVD burden and access to specialty care greatest in the South, highlighting substantial geographic inequity in the United States….Although cardiologists are not the only determinants of cardiovascular outcomes, the lack of access to cardiologists in areas with greater disease prevalence and mortality rates in almost one-half of U.S. counties is quite concerning.” A research letter published in the Journal of the American College of Cardiology by Kim and colleagues maps geographic patterns of access – or lack thereof – to cardiovascular care across the United States.
“Our results indicate the strong association of race/ethnicity with adverse environmental health outcomes; an association that is even stronger than for education, rurality or social vulnerability-related factors. This finding aligns with a growing body of evidence…demonstrating that racial/ethnic categories are not simply proxies for socioeconomic differences but are also proxy measures, while imperfect, for exposure to historical and contemporary discriminatory practices.” A research article by Geldsetzer and colleagues, published in Nature Medicine, investigates the associations between mortality attributable to air pollution and race/ethnicity.
“As participants listened to multiple short sentences containing a total of around 450 words, the scientists recorded which neurons fired and when. Williams says that around two or three distinct neurons lit up for each word, although he points out that the team recorded only the activity of a tiny fraction of the prefrontal cortex’s billions of neurons. The researchers then looked at the similarity between the words that activated the same neuronal activity.” Nature’s Sara Reardon reports on a neuron-level brain map that allows researchers to pinpoint the neurons that encode linguistic meaning.
“In the United States (US), access to maternity care in rural counties continues to decline. Overall, 49.0% of rural counties (969/1976) had hospital-based obstetrics in 2010, and 41.2% of rural counties (814/1976) had hospital-based obstetrics in 2022. By 2022, 58.8% (1162/1976) of rural counties had no hospital-based obstetric services.” An infographic created by Kozhimannil and colleagues and published by the University of Minnesota’s Rural Health Research Center maps the continuing loss of obstetric care among rural regions in the United States.

COMMUNICATION, Health Equity & Policy

“Federal antitrust regulators issued a report describing the PBM industry as highly concentrated among a handful of companies that use their market power to put pressure on clients and competitors….The 73-page Federal Trade Commission report, released Tuesday, doesn’t yet call for action such as breaking up pharmacy benefit managers from the insurers and drugstores they’ve merged with in recent years, or for putting limits on their business practices. But it does find that the investigation’s initial conclusions ‘urgently warrant further scrutiny and potential regulation.’” At Endpoints News, Drew Armstrong reports on the recently issued FTC report critically scrutinizing the practices of pharmacy benefit managers, or PBMs.
“We believe, however, that lawmakers ought to be wary of the risks posed by a policy strategy that encourages payer-led consolidation. One concern is that by controlling physicians and patient data in medical practices, conglomerates can maximize diagnosis coding and inflate risk-adjusted government payments.” A perspective article by Rooke-Ley and colleagues, published last week in the New England Journal of Medicine, examines the potential problems caused posed by consolidation and vertical integration of healthcare provision and coverage.
“The problems are particularly acute for students and early-career researchers, who are often paid meagre wages, have to uproot their lives every few years and have few long-term job prospects. But senior researchers face immense pressure as well. Many academics also experience harassment, discrimination, bullying and even sexual assault. The end result is that students and academics are much more likely to experience depression and anxiety than is the general population.” A Nature news feature by Shannon Hall examines the emerging response to a mental health crisis among academics in the sciences, from students to senior faculty.