AI Health

Friday Roundup

The AI Health Friday Roundup highlights the week’s news and publications related to artificial intelligence, data science, public health, clinical research, health policy, and more.

December 10, 2021

Deep Breaths

In today’s Roundup: introducing simulation intelligence; replication project for cancer studies has hard time getting data; DeepMind makes splash with compact language transformer; survey bias overestimated vaccine uptake; pandemic takes toll on nation’s blood pressure; synthetic embryos raise thorny questions; study finds no benefit from Medicare Advantage bonus program; what ethnography can tell us about the reproducibility crisis; much more:

“Photographic self-portraits have existed for as long as cameras have been in human hands. But what about selfies in space? On Twitter last year, NASA astronaut Edwin “Buzz” Aldrin, who famously became the second man to walk on the moon in July 1969, laid claim to a spaceflight first: taking the first selfie in space during the Gemini XII mission in 1966.” National Geographic’s Jenny Howard explores the obscure history of the outer-space selfie.
“Ginsberg was commissioned to write an accompanying essay about the history of Poppaea and about the different ways she has been depicted by historians and playwrights. She was also invited to Vienna to see the opera debut and to take part in a question and answer panel on opening night.” At Duke Today, Eric Ferreri has a profile of Lauren Ginsburg, a Duke professor of classical studies who provided some important background material for a new opera about the life of Poppaea, the wife of the notorious Roman emperor Nero.

AI, STATISTICS & DATA SCIENCE

“Existing 3D measurement techniques draw on specialized hardware, such as motion capture or depth cameras, as well as deep multi-view and monocular computer vision. Continued advances at the intersection of deep learning and computer vision will facilitate 3D tracking across more anatomical features, with less training data, in additional species, and within more natural, occlusive environments. 3D behavioral measurement enables unique applications in phenotyping, investigating the neural basis of behavior, and designing artificial agents capable of imitating animal behavior.” A preprint recently posted to arXiv by Marshall and colleagues describes recent progress in capturing three-dimensional kinematics of animal movement. (You can read more about related work in this story posted at the Duke AI Health websit)
“We present the ‘Nine Motifs of Simulation Intelligence’, a roadmap for the development and integration of the essential algorithms necessary for a merger of scientific computing, scientific simulation, and artificial intelligence. We call this merger simulation intelligence (SI), for short. We argue the motifs of simulation intelligence are interconnected and interdependent, much like the components within the layers of an operating system.” A preprint by Lavin and colleagues available at arXiv presents a scientific manifesto for a synthesis of algorithms to support “simulation intelligence.”
“…if something is doing something consequential, I need to be able to ask why?… And the answer to that question is where explainable AI comes in.” A recent episode of The Gradient’s podcast features an in-depth interview exploring explainable and interpretable artificial intelligence with Upol Eshan (audio, transcript, and bibliography all available at link).
“Torres and his colleagues were looking for peptides that are naturally produced by people and that can fight microbes. To do it, they used an AI that scrutinized the chemical makeup of each and every one in the human proteome—the complete set of proteins our bodies can produce.” At Wired, Max G. Levy explores recent research findings that reported on an AI-guided effort to identify potential countermeasures, already extant in the human proteome, to “superbug” resistant microbial strains.
“Called RETRO (for “Retrieval-Enhanced Transformer”), the AI matches the performance of neural networks 25 times its size, cutting the time and cost needed to train very large models. The researchers also claim that the database makes it easier to analyze what the AI has learned, which could help with filtering out bias and toxic language.” At MIT Technology Review, Will Douglas Heaven reports on the recent release of several papers by the DeepMind AI project that describe a new approach to AI language models that promises improved performance with a much smaller model.
“…many original papers failed to report key descriptive and inferential statistics: the data needed to compute effect sizes and conduct power analyses was publicly accessible for just 4 of 193 experiments. Moreover, despite contacting the authors of the original papers, we were unable to obtain these data for 68% of the experiments. Second, none of the 193 experiments were described in sufficient detail in the original paper to enable us to design protocols to repeat the experiments, so we had to seek clarifications from the original authors.” A paper recently published in eLife by a team seeking to replicate key findings in the oncology literature reports a high degree of difficulty in obtaining the basic materials needed to replicate findings in the first place.
“Good intentions don’t excuse covert surveillance and evading outside scrutiny of decisions that affect millions of individuals’ health and billions of dollars in expenditures. For reasons both ethical and economic, it’s time to open up health care’s secret analytics and demand a transparency that so far has been markedly absent.” An opinion article at STAT News by Michael Millenson argues for greater transparency across the health analytics industry.

BASIC SCIENCE, CLINICAL RESEARCH & PUBLIC HEALTH

“The report cited significant increases in self-reports of depression and anxiety along with more emergency room visits for mental health issues. In the United States, emergency room visits for suicide attempts rose 51 percent for adolescent girls in early 2021 as compared to the same period in 2019. The figure rose 4 percent for boys.” The New York Times’ Matt Richtel reports on US Surgeon General Vivek Murthy’s recent warning that the country is facing an unprecedented mental health crisis among children and adolescents.
“Our central message is that data quality matters more than data quantity, and that compensating the former with the latter is a mathematically provable losing proposition.” An article published in Nature by Bradley and colleagues on Wednesday reports that a pair of large surveys conducted in the US during the first half of 2021 may have substantially overestimated the extant of vaccine uptake due to sampling bias (H/T @shirokuriwaki).
“To be sure, the growing availability of home tests is good news for a country that stumbled through more than a year of the pandemic with inadequate testing resources. Still, as the U.S. moves into a second pandemic holiday season with the Omicron variant looming, state and local health departments are increasingly relying on incomplete data and educated guesses to capture ups and downs in the infection rate and to guide decision-making.” STAT News’ Kathleen McLaughlin explores the paradoxical downside to an otherwise positive development, as growing availability of home COVID testing is working to obscure true case counts from health agencies.
“The test catastrophe documented by the HHS investigation has led to unprecedented demands from former federal science officials to overhaul the CDC in order to avert another such disaster. This is a move that would essentially transform the country’s public health system.” Buzzfeed News reporter Dan Vergano offers a remarkably detailed look at systemic problems within the CDC that led to the failure of the COVID test initially developed by the agency during the early phases of the pandemic’s spread in the US.
“Black patients have a higher incidence of heart failure (HF) and worse outcomes than white patients. Guidelines recommend palliative care for patients with advanced HF, but no studies have examined outcomes in a black patient cohort.” A brief report by Tobin and colleagues published in the Journal of Palliative Care explores racial disparities in palliative care for heart failure.
“After a primary focus on vaccination, the White House announced on Thursday that private health insurers would soon have to reimburse patients for such tests. It also said it would make 50 million free tests available for uninsured Americans, to be distributed through health clinics and other sites in rural and underserved communities….But for consumers who do have insurance, the White House approach will require some legwork.” At the New York Times’ Upshot blog, Sarah Kliff and Reed Abelson unpack the Biden Administration’s recent announcement of coverage-via-reimbursement designed to increase access to rapid testing for COVID.
“Last year was a tough one. Americans grappled with a global pandemic, the loss of loved ones, lockdowns that splintered social networks, stress, unemployment and depression…It is probably no surprise that the nation’s blood pressure shot up. On Monday, scientists reported that blood pressure measurements of nearly a half-million adults showed a significant rise last year, compared with the previous year.” The New York Times’ Roni Caryn Rabin reports that COVID may be having an impact on your health, even if you haven’t contracted it.

COMMUNICATIONS & DIGITAL SOCIETY

“Regardless of the true rate, however, it is clear that far more papers should be retracted than are being retracted.” An article published in the European Heart Journal by Oransky and colleagues points out that identifying critical problems in the scientific literature is often not sufficient to ensure a timely retraction of the offending paper.
“Ask most advocates of rigorous science why this is, and they will answer with two words: perverse incentives. Scientists are rewarded for getting things published, not for getting things right, and so they tend to favour speed and ease over robustness. But as an ethnographer, this explanation has never sat well with me. I’ve spent more than 15 years studying biomedical research cultures, and scientists’ behaviours are rarely so transactional.” A perspective article at Nature by Nicole C. Nelson makes the case for a rigorous ethnographic approach to examining the reasons why scientific findings are so often irreproducible.
“…The Markup has learned, however, that the app is selling data on kids’ and families’ whereabouts to approximately a dozen data brokers who have sold data to virtually anyone who wants to buy it….the app acts as a firehose of data for a controversial industry that has operated in the shadows with few safeguards to prevent the misuse of this sensitive information.” The Markup’s Jon Keegan and Alfred Ng report that a widely used “family safety” app has been selling highly granular (and potentially identifiable) location data to data brokers and government agencies.
“What I offer here is a plea for humility, not merely as a stance of modesty vis-à-vis the powerful and still poorly understood forces of nature and society, but also as a practice of reasoning and policy that accepts uncertainty as its foundation and harm mitigation as its goal. If we mean to do better against the inevitable next time, how should we diagnose what went wrong in the preparations the world had for so long been cultivating, and how should we plan more effectively for the next crisis that will surely come?” An essay in the Boston Review by Sheila Jasanoff advocates for a philosophical stance that embraces humility rather than “preparedness” as a better approach for meeting the next inevitable public health challenges to follow the COVID pandemic. (Also at Boston Review, Zenynep Pamuk provides a rejoinder that challenges this framing.)

POLICY

“At a time when many Americans are willing to endorse conspiracies around the virus and top health officials are blaming misinformation for fueling vaccine refusal, there’s a growing demand for medical boards to make sure the doctors they license aren’t contributing to coronavirus misinformation. While only a small minority of doctors are actively spreading disinformation, experts argue that irresponsible doctors can have an outsize impact.” The Washington Post’s Alexandra Ellerbeck reports on efforts to rein in or sanction credentialed physicians who have been implicated in spreading medical misinformation, particularly in the context of the COVID pandemic.
“As the science of synthetic embryology gets more sophisticated, the models become more useful. But each advance raises a new round of ethical questions about where embryo models end and embryos begin. If it divides, organizes, and develops like an embryo, does it matter how it was made? Should an embryo derived from stem cells get the same legal and ethical rights as one produced when sperm met egg?” At STAT News, Megan Molteni reports on the host of practical and ethical questions that have followed a recent paper by Kagawa and colleagues that describes a successful attempt to develop quasi-blastocysts in the lab.
“We observed no consistent differential improvement in quality for MA [Medicare Advantage] versus commercial enrollees under the quality bonus program. Program participation was associated with significant quality improvements among MA beneficiaries on four measures, significant declines on four other measures, and no significant change in overall quality performance (+0.6 percentage points). Together, these results suggest that the quality bonus program did not produce the intended improvement in overall quality performance of MA plans.” A research article by Markovitz and colleagues published in this month’s issue of Health Affairs evaluates the track record of the Medicare Advantage Quality Bonus Program.