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 17, 2022
In today’s Duke AI Health Friday Roundup: the mathematics of randomized trials; medical debt affects huge proportion of Americans; k-safety properties help keep machine learning models on track; COVID worsens peer-review crisis; new analysis reveals US health disparities; human activity overwhelms animal senses; one girl’s illness yields new insights into lupus; FDA advisory panels give thumbs-up for COVID vaccines in small children; US still lags in clinical trial diversity; learning to measure what matters to patients; much more:
AI, STATISTICS & DATA SCIENCE
- “…we show the wide applicability of k-safety properties for machine-learning models and present the first specification language for expressing them. We also operationalize the language in a framework for automatically validating such properties using metamorphic testing. Our experiments show that our framework is effective in identifying property violations, and that detected bugs could be used to train better models.” In a preprint available from arXiv, Christakis and colleagues describe “k-Safety Properties” – a means for ensuring the “functional correctness” and fairness of machine-learning models (H/T @arXiv_Daily).
- “Most recently, Spielman has turned his attention to the mathematics behind randomized controlled trials, which underpin modern medical studies. The organizers of these trials try to randomly split study subjects between a test group, which receives an experimental treatment, and a control group, which doesn’t. However, finite-size groups always end up with an imbalance in some category, such as age or weight or blood pressure. Along with his research group, Spielman has worked to find algorithms for achieving a better balance.” Quanta’s Mordechai Rorvig interviews Yale mathematician Daniel Spielman about his foray into the world of clinical research, among other topics.
- “A sustained price at around $24K could shrink the Bitcoin network’s global energy use to around 170 TWh annually, according to de Vries. That might sound like an incremental change, but it would add up to a significant drop in electricity use and related greenhouse gas emissions. If you compare it to the annualized energy use de Vries estimated Bitcoin was responsible for throughout much of 2022, it would be like shaving off the amount of electricity the country of Ireland uses in a year.” The Verge’s Justine Calma reports on the recent and sudden collapse in value of cryptocurrencies – and what the environmental implications might be if investors desert the energy-hungry assets.
- “The transition to virtual delivery of health and care services will continue even after the Covid-19 pandemic ends. With the principles of equity, inclusivity, and transparency guiding that transition, virtual health and care can provide a vital complement to in-person care and make the highest level of health attainable for all.” In an editorial for STAT News, WHO Director-General Tedros Adhanom Ghebreyesus and Novartis Foundation head Ann Aerts call for a concerted global effort to consolidate gains in digital access to healthcare that emerged during the COVID pandemic and ensure that all can benefit from virtual care.
BASIC SCIENCE, CLINICAL RESEARCH & PUBLIC HEALTH
- “…we have also filled the silence with noise and the night with light. This often ignored phenomenon is called sensory pollution—human-made stimuli that interfere with the senses of other species… We have distracted them from what they actually need to sense, drowned out the cues they depend upon, and lured them into sensory traps. All of this is capable of doing catastrophic damage.” At The Atlantic, science writer Ed Yong explores the toll that byproducts of human activity – namely, light and noise pollution – is taking on the natural world.
- “The experts all expressed support for vaccinating children in this youngest population to better prevent against Covid-related hospitalizations and deaths that have risen in this group during the latest wave of the Omicron variant.” Endpoints News’ Zachary Brennan reports that the FDA’s advisory committees have voted unanimously in support of making both the Pfizer and Moderna COVID-19 vaccines available to younger (6 months to 6 years) patients.
- “The original Kika is a stuffed elephant that a family friend gave Piqueras during one of her long stays at a Madrid hospital. The family dog is also named Kika. But perhaps the most significant Kikas are the mice with Piqueras’s same kind of lupus, added into their cells with gene-editing technology.” STAT News’ Isabella Cueto has the story of Gabriela Piqueras, a Guatemalan girl whose idiosyncratic case of lupus is shedding new light on the autoimmune disease.
- “There is little to no change in most markers of cardiac status among heart failure patients with implanted ICD or CRT-D devices in NYC and MSP during the COVID-19 pandemic, and this is consistent for all age and sex subgroups. All the changes in medical care, the stress of the pandemic, and the change in behaviors did not seem to have any major effect on this patient population. It is plausible that telemedicine mitigated some of the effects of a decrease in inpatient care.” A paper by Lu and colleagues published in the Journal of Cardiac Failure used data gathered remotely from implanted cardioverter/defibrillators to assess the effects of COVID lockdowns on patients’ cardiac health.
- “Disparities in life expectancy among racial–ethnic groups are widespread and enduring. Local-level data are crucial to address the root causes of poor health and early death among disadvantaged groups in the USA, eliminate health disparities, and increase longevity for all.” A new analysis by a group of researchers with the Lancet Global Disease Burden Project have published a sweeping, 20-year, county-level analysis of racial and ethnic disparities in US life expectancy.
- “The current environment, characterized by a haphazard approach to deciding what to measure for which diseases, results in confusion, missed opportunities, inefficiencies, and frustrations. Patients and clinicians lack data relevant for decision making; health technology assessors hear from patients about what matters but find little data they can use; and systematic reviewers identify few studies using the same measures and endpoints. The key question is: What should be measured and monitored?” An article in Health Affairs Forefront by Perfetto and colleagues advances a framework for developing sets of criteria for developing assessments of health impact that matter most to patients.
- “I t was certainly interesting going through the Covid experience, when there was absolutely no tolerance from me or anyone else for that kind of delay. It was pretty impressive to see NIH rise to the occasion, with the development of vaccines in 11 months, the formation of the ACTIV public-private partnership [on Covid therapeutics] and the RADx initiative to develop home testing, which now everybody takes for granted.” STAT News’ Lev Facher sits down with Francis Collins, who recently stepped down from leading the National Institutes of Health only to be immediately called back to serve as White House interim science advisor.
- “Bačaq’s teeth, as well as those of another woman buried nearby, have now yielded genomic evidence of what researchers suggest is the ancestral strain of the Yersinia pestis bacterium responsible for the 14th-century Black Death pandemic, according to a study published on Wednesday in Nature. The paper also points to this region as the source of that notorious plague, which killed at least an estimated 30 to 60 percent of Europe’s population in a handful of years.” Scientific American’s Jen Pinkowski reports on archaeological research, recently published in Nature, that may shed new light on the origins of the 14th -century plague known as the Black Death.
COMMUNICATION, Health Equity & Policy
- “The investigation reveals a problem that, despite new attention from the White House and Congress, is far more pervasive than previously reported. That is because much of the debt that patients accrue is hidden as credit card balances, loans from family, or payment plans to hospitals and other medical providers.” A joint NPR/Kaiser Health News report by Noam Levey investigates the crushing load of medical debt burdening millions of Americans.
- “Improving diversity in clinical trials is one of the most important steps needed to achieve health equity. The reasons for this underrepresentation are myriad and complex. Obstacles range from individuals not having physical access to clinical trial sites, to general distrust of the medical institutions running the research, to a lack of proper community outreach. How can researchers overcome these challenges and champion the importance of increasing diversity in clinical trials to the medical community at large?” In the latest installment of the Color Code podcast, Nicholas St. Fleur examines the persistent and “woeful” lack of diversity among US clinical trial participants.
- “By contrast, the greater anonymity in peer-reviewed journal reviews seems to lead to less professional reviews. In my experience, it is not uncommon to receive reviews of one’s manuscript for a peer-review journal that are unprofessional, unintelligent, hostile, and completely off the mark. So, even when I do not agree with the reviews I may get from an NIH panel, they are rarely outside the bounds of intelligence and professionalism.” An interview with Indiana University – Bloomington nutrition researcher David Allison for the Good Science Project contains some candid and thought-provoking observations on the process of peer review, both for journal publication and grant review.
- “During COVID-19, journalists shifted their professional norms and practices to more readily include preprints. Our findings suggest this departure is likely to continue post-pandemic and expand beyond the use of preprints related to health and biomedical sciences.” A qualitative analysis published in preprint form at BioRxiv by Fleerackers and colleagues explores how journalists have reacted to the preprint revolution in medical publishing – particularly amid the COVID pandemic.
- “…academic publishing has long been a delicate system that operates—tenuously—on goodwill, in the form of comprehensive, unpaid article analyses from expert volunteers. But the pandemic has pushed this system to breaking, or close to it. With academics’ professional and personal lives disrupted in so many ways for years now, this kind of labor is increasingly harder to source: journal editors across fields say scholars are significantly less likely to accept article-review requests, if they respond at all, and…they are more likely to return reviews that are late or even rushed.” At Inside Higher Ed, Colleen Flaherty addresses the “peer-review crisis” and the additional stresses that the COVID pandemic has put on an overburdened cornerstone of academic research.