I started writing this piece on New Year’s Day, as Chicago was getting its first big winter storm of the season, making it an especially good day to honor my New Year’s resolution to pick up the reins of my blog again. The previous installment was last March. Until just now I haven’t seen where pieces about the subtleties of artificial intelligence or about the depredations of electronic medical records would capture much attention from health professionals who were dealing day in and day out with a pandemic that has been overwhelming the healthcare system and everybody in it.
It’s not like the pandemic is over. In fact there are more cases than ever now, with over 100,000 hospitalized covid patients. Burnout is rampant among health professionals. Still, we know a lot more about this virus than we did two years ago. Today we’ve got vaccines and effective treatments, with new ones in the pipeline. The highly contagious omicron virus variant that we’re currently battling doesn’t appear to be as lethal as some of its predecessors. There’s reason for cautious optimism. Anyway, I’m tired of putting things on hold, including this blog.
In the spirit of cautious New Year’s optimism, I’m writing again. The blogs and reviews will come more regularly from here on out. In this one I summarize seven articles that touch on how information technology can affect/has affected the overall health endeavor, especially public health. Of course some of the pieces are about covid. How could they not be?
Thank you for sticking with me. I look forward to your responses.
- Marc Ringel, MD
Smartphone science: apps test and track infectious diseases
Nature 2021
A group in India has employed a simple, portable, inexpensive test that was developed thanks to CRISPR technology to detect tiny amounts of covid virus RNA on a nasal swab. A positive result manifests as a band that appears on a paper strip. The problem is that not every result is clearly positive or negative. It takes a more sophisticated instrument than the human eye--not widely available in India, whose population is still 70% rural--to reliably classify borderline results. As of January 2022 there were an estimated 439 million smart phone users in that country. Thanks to this nearly ubiquitous technology, people in the field can take a photo of a covid test strip, upload it to a site where the image is processed by AI-enabled software that spits out an immediate yes/no answer about the presence of covid-19 coronavirus in the sample.
In the UK logs about covid symptoms and testing kept by regular people on their personal electronic devices and then uploaded to public health institutions have taught the medical community a whole lot about the distribution and significance of symptoms, contagion hotspots, and disease course. Healthmap aggregates a crowdsourced set of data, uploaded by citizens, plus newsfeeds and official reports, to produce a map entitled, “Outbreaks Near Me.” The animated map includes everything from acenitobacter baumannii to zika, with covid-19 and influenza in between. Another citizen science epidemiology project, Mozzie Monitors, uses data acquired by smartphone to do large scale mosquito surveillance at 1/5 the cost of professional programs.
View Article
Evidence-based medicine: how COVID can drive positive change
Nature 2021
How COVID broke the evidence pipeline
Nature 2021
Every doctor’s education is saturated from the very first day of training with the ideal of practicing scientifically. Never mind that each case is unique and that the sum total of medical knowledge falls far short of answering a large share of clinical questions with a firm, scientifically validated answer. Nevertheless, when we practice, physicians are expected to seek out the best evidence and adhere to it as closely as we can.
There are all sorts of resources that summarize and present the evidence for best practice. Studies that are prospective--meaning that before they ever start collecting data the researchers have to decide what questions they want to ask and what data they’ll need to harvest to answer them--get higher marks for trustworthiness. Designing and running a prospective study is a slow process. Then it generally takes months to years to collect and analyze data and publish results and another long time, if ever, to evaluate the validity of a study’s findings and place them in the context of medical practice. Now, along comes the covid pandemic, blindsiding almost everybody. The epidemic changes day-to-day. New interventions are proposed weekly. The virus evolves month-to-month. Pinning down what is to be done is like trying to thread a needle on the deck of a ship in a stormy sea.
Early in the pandemic practitioners on the front lines cast about frantically for guidance. Public health institutions, research centers, and healthcare systems did a yeoperson’s job of supplying providers with the best answers they could. As you would expect of a new, unknown disease, these answers didn’t always agree with each other. At the beginning recommendations sometimes changed from day-to-day.
The most highly regarded synthesizers of the medical literature have come to understand that they have to change. They’ve teamed up with the World Health Organization to re-engineer the process for generating research based medical evidence so that actionable items not only are arrived at orders of magnitude quicker, but they’re disseminated rapidly and widely, giving patients themselves, especially from the most vulnerable groups, a better chance of actually receiving the best care that evidence can provide.
Big Data in Context: Addressing the Twin Perils of Data Absenteeism and Chauvinism in the Context of Health Disparities Research
Journal of Medical Internet Research 2020
Frontline clinicians, like schoolteachers and law enforcement officers, routinely come up against the brutal fact that the person we’re tasked to help is trapped by poverty in an unhealthy neighborhood or family. The things they most need to heal are way beyond our resources and scope of practice. Public health data can at least provide a place to start helping by supplying an accurate assessment of community and society-level problems.
Big data gives some hope of greater understanding of what’s going on out there. The tech company Oracle provides this definition of big data, “… larger, more complex data sets, especially from new data sources. These data sets are so voluminous that traditional data processing software just can’t manage them. But these massive volumes of data can be used to address business problems you wouldn’t have been able to tackle before.” Indeed, big data is used extensively to inform a host of business functions, from how to sell a product to where to invest.
No matter whether you’re doing business or public health you need, above all, good data. The old saw, “garbage in, garbage out,” is just as true for big data as for any other kind of data. There’s the further hazard with enormous datasets of not looking critically enough at quality of the individual datapoints themselves because there are so many of them. It is generally true in statistics that the bigger the denominator (that is the greater the number of subjects), the more reliable the conclusions. By definition, big data provide a huge denominator. When such data have systematic errors, such as underrepresenting a part of the population to be studied or encoding prejudices in how the data are collected or evaluated, you not only get garbage out, you may get toxic, harmful garbage that serves to further entrench bad attitudes and policies.
Modern statistics itself was codified in the last hundred years or so by eugenicists who applied their statistical machinery to consistently affirm their belief in the superiority of the white race, especially those of Northern European origin. (If you want to explore this subject further, check out the 2021 book, Bernoulli’s Fallacy by Aubrey Clayton.) This article will help you to cast a critical eye on public health findings generated by big data.
A Comprehensive Overview of the COVID-19 Literature: Machine Learning–Based Bibliometric Analysis
Journal of Medical Internet Research 2021
…speaking of big data…The authors of this article scoured several huge medical literature databases and retrieved 196,630 publications, published between January 1 to July 21 2020, that appeared on their face to be about the covid-19 pandemic. They winnowed the field down to 28,904 pertinent articles. Starting with 8 publications in week 1, the literature reached a peak of 2276 articles published in week 22. Then they sliced and diced these results to find that the most articles had originated in China, followed by the United States. Studies were classified by type (surveys, reviews, case reports, case-control, etc.). With a generous dose of artificial intelligence, the reports were classified into 19 areas of study, from public health response to effect on organ transplantation, with lots of basic science about the virus, the host immune response and clinical protocols in between. Once more I’d like to ask, how does a conscientious person keep up?
Patterns of Routes of Administration and Drug Tampering for Nonmedical Opioid Consumption: Data Mining and Content Analysis of Reddit Discussions
Journal of Medical Internet Research 2021
The authors of this study did some fancy filtering of the text of all of some 4.5 billion(!) messages logged between 2014 and 2018 on the publicly-available database of Reddit, a widely used social media platform. They found that over these five years comments about opioids increased from 387,000 to 919,000 per year. Then they estimated the evolution of products used, routes of administration (injection, inhalation, swallowing, rectal insertion), and ways of modifying the drugs (brew, concentrate, evaporate, extract, etc.). This information could be very useful to people who design and run drug abuse programs as well as to people working in the field.
One Digital Health: A Unified Framework for Future Health Ecosystems
Journal of Medical Internet Research 2021
Here’s a global proposal, complete with a multilevel pie chart and two radar charts, to link individual health care and well-being with population, society and the environment by integrating every kind of data that bears on health that the authors can think of. It’s a lovely idea. Like a lot of other, big lovely ideas, I don’t expect it to get very far, at least not in America where, pandemic or not, public health will continue to get short shrift so long as our system is organized around profit, not around health.
