Category: Medicine
Left Digit Bias in Medicine
From my review in the WSJ of Random Acts of Medicine by Jena and Worsham:
You have probably heard of left-digit bias—the idea that $7.99 seems cheaper than $8, even though $8 is only negligibly different than $8.01. Left-digit bias is widely observed in pricing but the effect is more general. A car with 39,990 miles on the odometer, for instance, sells for more than a car with 40,005 miles (so be smart and buy the car with 40,005 miles). Could left-digit bias show up in medicine?
People who end up in the emergency room complaining of chest pains a few weeks before their 40th birthday are very similar to people who end up in the emergency room with chest pains a few weeks after their 40th birthday. But on a chart, the former are 39 years old and the latter are 40.
The big 40 is a heuristic among physicians for potential heart attack. Looking at more than five million patient records, the economist Stephen Coussens found that patients who were slightly over the age of 40 were almost 10% more likely to be tested for a heart attack than those just under 40. The difference shows up as a discontinuity, a jump up in the probability of being tested as patients cross their
40th birthday.
Messrs. Jena and Worsham show that similar discontinuities appear throughout medicine. Heart-attack patients just under the age of 80, for instance, are more likely to be given coronary artery bypass surgery than those just over 80. Kidneys from patients who die at age 69, just short of their 70th birthday, are more likely to be used for transplant than kidneys from patients just over 70, even though by all objective measures the kidneys are equally viable and valuable. Perhaps most tellingly, “children” just under the age of 18 are less likely to be prescribed opioids than “adults” slightly over the age of 18, even though these groups are statistically indistinguishable.
The point of these studies isn’t to titter or sigh at the peculiarities of human reasoning but to use these natural experiments to estimate the effect of medical procedures. If the only reason that near-18 and 18-year-olds are prescribed opioids differently is the semantics of “child” and “adult,” then we can use the discontinuity in prescriptions as a natural experiment—it’s as if prescribing around the age of 18 were randomly assigned. The authors find, for example, that compared to the just-under-18s, the just-over-18s were 12.6% more likely to later be diagnosed for an opioid-related adverse event such as an overdose. The greater rate of overdose is valuable information—but imagine the difficulty of trying to convince an Institutional Review Board that it would be ethical to randomly prescribe opioids to young people.
The Amy Finkelstein and Liran Einav health care plan
I am away from my review copy, so I am pleased that Matt Yglesias has offered ($) a good “standing on one foot” summary of the plan, as outlined in the new book We’ve Got You Covered: Rebooting American Health Care, by Amy Finkelstein and Liran Einav:
They call for:
- A universal basic insurance system, covering both catastrophic and routine care but at a bare bones/no frills level of service.
- A global budget, set by Congress, to determine how much money the basic plan has to spend on meeting the public’s basic needs, paired with expert panels to decide which services to cover.
- An additive system of private top-up insurance that people could (and they anticipate mostly would) buy into to secure access to shorter wait times and more creature comforts.
The book offers a “think it through using first principles” approach, so perhaps the authors will be frustrated by my invocation of a “how has politics been going lately?” kind of response. Nonetheless I see that Obamacare cost the Democrats dearly in more than one election, it had to be defanged (the mandate) to survive, it was supposed to be the new comprehensive framework that actually could pass (it did), and the most influential Americans just love their employer-provided private health insurance.
Whether you think those facts are good or bad, I take them as my starting point for health care reform. This book does not.
I observe also that Obamacare passed, and American life expectancy fell. I do not blame Obamacare for that, but I do notice it. As a result, I have grown increasingly interested in “how can we boost biomedical scientific progress?” and increasingly less interested in “how can we reform health insurance coverage again?” All the more because we seem to be living in a biomedical progress of science golden age.
One of the Democratic Party frustrations with conservatives during the ACA debates was witnessing them tolerate or even support Romney’s Massachusetts plan, but oppose Obamacare. That I can understand. One of the conservative frustrations with ACA was the fear that it would just be the first step in a never-ending, upward-ratcheting series of efforts to spend ever more on health insurance coverage, which has positive but only marginal implications for health itself. After all, where exactly do the moral arguments for spending more on health insurance coverage stop?
Is there a politically feasible version of the Finkelstein and Einav plan that can spend less or the same? Is there a politically feasible version of the plan period? How much trust will there be in the promise that if I give up my private health insurance coverage, it will be replaced by something better? How much trust should there be?
But again, the authors here have a very different perspective on the sector and how to do health care policy.
Australia fact of the day
Health officials have “virtually” eliminated HIV transmission in parts of Sydney that were once the centre of the Australian Aids epidemic, raising hopes of conquering a disease that has killed more than 40mn people.
HIV diagnoses in inner Sydney plunged 88 per cent from the 2008-12 average to just 11 cases last year, a decline on a scale never before recorded in a former Aids hotspot.
The results add to evidence that existing prevention strategies, including testing and pre-exposure drugs, are highly effective when implemented correctly.
“Rapid progress towards ending Aids is possible. If trends continue, several countries in several global regions will reach the [UN] goal of a 90 per cent HIV incidence reduction by 2030,” researchers said.
Here is the full FT story. As I have been saying people, you are living in a new age of biomedical miracles.
Mental health and European economics departments
We study the mental health of graduate students and faculty at 14 Economics departments in Europe. Using clinically validated surveys sent out in the fall of 2021, we find that 34.7% of graduate students experience moderate to severe symptoms of depression or anxiety and 17.3% report suicidal or self-harm ideation in a two-week period. Only 19.2% of students with significant symptoms are in treatment. 15.8% of faculty members experience moderate to severe depression or anxiety symptoms, with prevalence higher among nontenure track (42.9%) and tenure track (31.4%) faculty than tenured (9.6%) faculty. We estimate that the COVID-19 pandemic accounts for about 74% of the higher prevalence of depression symptoms and 30% of the higher prevalence of anxiety symptoms in our European sample relative to a 2017 U.S. sample of economics graduate students. We also document issues in the work environment, including a high incidence of sexual harassment, and make recommendations for improvement.
That is from a new paper by Elisa Macchi, Clara Sievert, Valentin Bolotnyy, and Paul Barreira.
How the NSF Moved Faster than the NIH During COVID-19
The NSF is a much smaller organization than the NIH but during the pandemic it moved more quickly. Why? Maxwell Tabarrok explains:
The NSF relied on its special congressional authority to skip peer review to bootstrap its pandemic-related granting. Two pre-existing programs which use this authority enabled the NSF’s speedy response. The RAPID (Rapid Response Research) and EAGER (EArly-concept Grants for Exploratory Research) programs focus on “proposals having a severe urgency,” and “exploratory work in its early stages on untested, but potentially transformative, research ideas,” respectively. Both turn applications around quickly: while typical federal science grants take 9-12 months of review, RAPID and EAGER grants usually provide funding to researchers in less than a month.
…The NSF funded valuable research through its RAPID grants program, including the development of the first COVID-19 test to get FDA approval, the Johns Hopkins COVID-19 data dashboard, and both inhaled and micro-needle patch vaccines, the latter of which is currently being scaled up for use in HPV vaccines. These examples don’t conclusively show that the NSF avoided sacrificing quality control for speed, but they suggest that the NSF’s internal team of reviewers funded multiple effective projects that benefited from faster turnarounds. The benefits of speeding up these big successes when they were urgently needed outweighed the hypothetical costs of approving some below-average projects.
In crises generally, the success of a science funder is determined by its biggest wins, not by the average quality of the projects it approves. Science’s impact on the pandemic was dominated by a single technology: the mRNA vaccine. The next most important contributions, likely testing or pharmaceutical treatments, were less important than the vaccine, and the average COVID-19 research project may have had minimal impact. External peer review slows down the funding of all projects to make sure that low-quality research is not funded. This kind of bottom-end quality control is less important in a crisis environment. At crisis-response margins, it’s probably better for science funding agencies to anchor less on quality control and instead take more shots on goal.
The NIH, to be fair, also responded more rapidly than usual and it used some special “shark-tank” like programs to do so which also worked well.
Read the whole thing for more recommendations.
Against human-AI collaboration
From a new NBER working paper by Nikhil Agarwal, Alex Moehring, Pranav Rajpurkar, and Tobias Salz:
Radiologists do not fully capitalize on the potential gains from AI assistance because of large deviations from the benchmark Bayesian model with correct belief updating. The observed errors in belief updating can be explained by radiologists’ partially underweighting the AI’s information relative to their own and not accounting for the correlation between their own information and AI predictions. In light of these biases, we design a collaborative system between radiologists and AI. Our results demonstrate that, unless the documented mistakes can be corrected, the optimal solution involves assigning cases either to humans or to AI, but rarely to a human assisted by AI.
I am more optimistic in my views, noting there may well be contexts such as radiology where the collaborations fail. I collaborate with Google’s AI all the time, and I am pretty sure that joint effort does better than either myself or “Google with no human” unaided. Still, this is a cautionary note of some import, as many humans are not good enough to work well with AIs.
Intergenerational transmission of mental health problems
We estimate health associations across generations and dynasties using information on healthcare visits from administrative data for the entire Norwegian population. A parental mental health diagnosis is associated with a 9.3 percentage point (40%) higher probability of a mental health diagnosis of their adolescent child. Intensive margin physical and mental health associations are similar, and dynastic estimates account for about 40% of the intergenerational persistence. We also show that a policy targeting additional health resources for the young children of adults diagnosed with mental health conditions reduced the parent-child mental health association by about 40%.
That is from a new NBER working paper from Aline Bütikofer, Krzysztof Karbownik, and Fanny Landaud.
China estimate of the day
The paper’s title is “The Largest Insurance Expansion in History: Saving One Million Lives Per Year in China”:
The New Cooperative Medical Scheme (NCMS) rolled out in China from 2003-2008 provided insurance to 800 million rural Chinese. We combine aggregate mortality data with individual survey data, and identify the impact of the NCMS from program rollout and heterogeneity across areas in their rural share. We find that there was a significant decline in aggregate mortality, with the program saving more than one million lives per year at its peak, and explaining 78% of the entire increase in life expectancy in China over this period. We confirm these mortality effects using micro-data on mortality, other health outcomes, and utilization.
It is striking how few Westerns have even heard of this policy, one of the more important global events in recent years. I do however wish to ask if this estimate is in accord with other, more general estimates from the literature. The Amish, for instance, don’t see doctors so often and their life expectancy seems to be perfectly fine. The new paper is from Jonathan Gruber, Mengyun Li, and Junjian Yi.
The Birth-Weight Pollution Paradox
Maxim Massenkoff asks a very good question. If pollution reduces birth weight as much as the micro studies on pollution suggest, why aren’t birth weights very low in very polluted cities and countries? Figure 1, for example, shows birth weights in a variety of highly polluted world cities. The yellow dashed and blue lines show “predicted” birth weights extrapolated from the well-known Alexander and Schwandt “Volkswagen study” which looked at the effects of increased pollution in the United States. Despite the fact that every one of the highly-polluted cities is much more polluted than the most polluted US city, birth weight is not tremendously lower in these cities. Indeed, there is no obvious correlation between birth weight and pollution at all.
Similarly, US cities were more polluted in the past but were birth weights lower in the past? Figure 2 shows a number of US cities which were two to three times more polluted in 1972 (right side of diagram) than 2002 (left side of diagram). Yet, birth weights do not appear lower in the more polluted past and certainly do not follow the extrapolated birth weight-pollution predictions from the micro literature.
Massenkoff looks at a variety of possible explanations. One possibility, for example, is culling. Perhaps in highly polluted areas there are more miscarriages, still births or difficulty conceiving with the result that the observed sample of births is highly selected. There is some evidence that pollution increases miscarriages and stillbirths but these tend to be correlated with lower birth weight–a scarring effect rather than a culling effect. In addition, the effect of pollution on miscarriages and stillbirths also appears to be bigger on a micro level than on a macro level. That is, these rates aren’t massively higher in high pollution countries.
Another possibility is that pollution isn’t that bad and, in particular, not as bad as I have suggested. As a good Bayesian, I update, but for reasons I have given here, it’s not justifiable to update very much.
I assume, as I always do, that there are some overestimates in the micro literature for the usual reasons. But, more fundamentally, my best guess for the birth-weight pollution paradox is that weight is one of the easiest margins on which the body can adapt and compensate. Even in poor countries there are plenty of calories to go around and so it’s relatively easy for the body to adjust to higher pollution, on this margin. Indeed, weight is known as a variable that creates paradoxes!
Micro studies on weight and exercise, for example, show that exercise reduces weight. But looking across countries, societies, and time we don’t see big effects–indeed, calorie expenditure doesn’t vary much with exercise! Importantly, notice that the micro-estimates are correct. If you increase physical activity for the next 3 months, holding all else equal (which is possible for 3 months), you will lose weight. However, the micro estimates are difficult to extrapolate to permanent, long-run changes because there are complex, adaptive mechanisms governing weight, calorie consumption and energy expenditure.
The exercise paradox doesn’t mean that exercise isn’t good for you–the evidence on the benefits of exercise is extensive and credible. In the same way the birth-weight pollution paradox doesn’t mean that pollution isn’t harmful–the evidence on the costs of pollution is extensive and credible. In particular, it’s going to be much harder to adapt to pollution for heart disease, cancer, life expectancy and IQ than for weight.
I am always impressed with papers that present big, obviously-true facts that most people have simply missed. Massenkoff is becoming a leader in this field.
Is “Lab Leak” now proven?
The WSJ ran a widely discussed article a few days ago, and many people have concluded that the Lab Leak hypothesis is now confirmed. I’ve now read the piece, and I don’t see relevant new information in there. The New York Times ran a rebuttal of sorts, with this as one key paragraph:
Recent news reports have unearthed new information about researchers from the Wuhan Institute of Virology who became sick in 2019. The news reports suggested that one of them could be patient zero. The information about the sick workers was first discovered at the end of the Trump administration. By August 2022, however, intelligence analysts had dismissed the evidence, saying it was not relevant. Intelligence officials determined that the sick workers could not tell them anything about whether a lab leak or natural transmission was more likely. Intelligence agencies view the information about the cases neutrally, arguing that they do not buttress the case for the lab leak or for natural transmission, according to officials briefed on the intelligence.
I read the London Times report, and didn’t see fundamentally new information in there either.
To be clear, I think the chance of Lab Leak being true is reasonably high, due to the accumulation of a lot of circumstantial evidence. But I don’t think the new accounts are anything close to a slam-dunk, nor do they show that any of the researchers were “Patient Zero.” That may well change as further information comes out, but so far it is a mistake to conclude that Lab Leak has been demonstrated to be true.
Addendum: As a side note, I am a little worried by how many people seem to be happy that Lab Leak hypothesis is (supposedly) confirmed. I suppose it would mean you could feel vindicated in a certain kind of contempt for elites, both American and Chinese. But under most normal views, the world where Lab Leak is true is a worse world than the world where Lab Leak is false. So you should instead feel sad and upset if you think it is true, rather than happy or gleeful. If you feel vindicated, it is a sign of a partial cognitive and emotive defect.
Second Addendum: This new national intelligence report doesn’t seem to confirm the Lab Leak take (though it doesn’t refute it either). It pretty definitely downplays the import of the scientists getting sick. Again, it is fine to not trust this report, but still a likely mistake to think new information has been coming out. Here is a good WaPo look at where things stand. Here are comments from Scott Sumner.
Does Britain Have High or Low State Capacity?
Tim Harford writing at the FT covers the question “Is it even possible to prepare for a pandemic?” drawing on my paper with Tucker Omberg.
[I]n an unsettling study published late last year, the economists Robert Tucker Omberg and Alex Tabarrok took a more sophisticated look at this question and found that “almost no form of pandemic preparedness helped to ameliorate or shorten the pandemic”. This was true whether one looked at indicators of medical preparedness, or softer cultural factors such as levels of individualism or trust. Some countries responded much more effectively than others, of course — but there was no foretelling which ones would rise to the challenge by looking at indicators published in 2019. One response to this counter-intuitive finding is that the GHS Index doesn’t do a good job of measuring preparedness. Yet it seemed plausible at the time and it still looks reasonable now.
…perhaps we need to take the Omberg/Tabarrok study seriously: maybe conventional preparations really won’t help much. What follows? One conclusion is that we should prepare, but in a different way….Preparing a nimble system of testing and of compensating self-isolating people would not have figured in many 2019 pandemic plans. It will now. Another form of preparation which might yet pay off is sewage monitoring, which can cost-effectively spot the resurgence of old pathogens and the appearance of new ones, and may give enough warning to stop some future pandemics before they start. And, says Tabarrok, “Vaccines, vaccines, vaccines”. The faster our systems for making, testing and producing vaccines, the better our chances; all these things can be prepared.
One thing that did seem to matter, as Tim notes, was state capacity. In other words, it’s not so much being prepared as being prepared to act. And here I have a mild disagreement with Tim. He writes:
In an ill-prepared world, the UK is often thought to have been more ill-prepared than most, perhaps because of the strains caused by austerity and the distractions of the Brexit process.
My view is that the UK got three very important things right. The UK was the first stringent authority to approve a COVID vaccine. The UK switched to first doses first and the UK produced and ran the most important therapeutics trial, the Recovery trial. Each of these decisions and programs saved the lives of tens of thousands of Britons. The Recovery trial may have saved millions of lives worldwide.
I don’t claim that Britain did everything right, or that they did all that they could have done, but these three decisions were important, bold and correct. The coexistence of both high and low state capacity within the same nation can be surprising. The United States, for example, achieved an impressive feat with Operation Warp Speed, yet simultaneously, the Centers for Disease Control and Prevention (CDC) flailed and failed. Likewise, India maintains a commendable space program and an efficient electoral system, even while struggling with tasks that seem comparatively simpler, like issuing driver’s licenses.
Instead of painting countries with a broad brush of ‘high’ or ‘low’ state capacity, we should recognize multi-dimensionality and divergence. How do political will, resources, institutional robustness, culture, and history explain capacity divergence? If we understood the reasons for capacity divergence we might be able to improve state capacity more generally. Or we might better be able to assign tasks to state or market with perhaps very different assignments depending on the country.
Vaccination sentences to ponder
None of the incidences of myopericarditis pooled in the current study were higher than those after smallpox vaccinations and non-COVID-19 vaccinations, and all of them were significantly lower than those in adolescents aged 12–17 years after COVID-19 infection.
I would gladly see a refereed symposium on attempts to overturn this result. Individuals with rejected papers could publish those rejected works on a separate website, with the accompanying referee reports of course. One side will try to tell you that “the elites” are against debate. It is sooner the case that the level of rigor in a useful debate should correspond to the level of rigor a subject matter requires.
Via Megan McArdle.
UK NHS fact of the day
Germany has six for every thousand people, Belgium has five, we have two.
Here is an essay by Sam Freedman, “How Bad Does It Need to Get?”, via Nick Thornsby.
Free Formatting For All on First Submission!
Many years ago I was incredulous when my wife told me she had to format a paper to meet a journal’s guidelines before it was accepted! Who could favor such a dumb policy? In economics, the rule is you make your paper look good but you don’t have to fulfill all the journal’s guidelines until after the paper is accepted. Sensible!
A paper just published in BMC Medicine estimates that this obtrusive norm costs researchers in biomedical journals alone some $230 million a year in wasted time. That’s consistent with an earlier study which estimated that over a billion dollars worth of time was wasted reformatting papers in all scientific fields. Quoting from that earlier study:
Our data show that nearly 91% of authors spend greater than four hours and 65% spend over eight hours on reformatting adjustments before publication…Among the time-consuming processes involved are adjusting manuscript structure (e.g. altering abstract formats), changing figure formats, and complying with word counts that vary significantly depending on the journal. Beyond revising the manuscript itself, authors often have to adjust to specific journal and publisher online requirements (such as re-inputting data for all authors’ email, office addresses, and disclosures). Most authors reported spending “a great deal” of time on this reformatting task. Reformatting for these types of requirements reportedly caused three month or more delay in the publication of nearly one fifth of articles and one to three month delays for over a third of articles.
It’s all very depressing. If we can’t get rid of unproductive paper reformatting standards–which benefit no one–how can we expect to tackle monumental tasks that require navigating complex tradeoffs such as resolving global climate change or making the tax code more just and efficient?
Yet perhaps there is hope. The BMC Medicine paper was covered in Nature and the authors have started a petition to change the reformatting norm. Do your part. Sign the petition! Free formatting for all on first submission!
Emergent Ventures winners, 26th cohort
Winston Iskandar, 16, Manhattan Beach, CA, an app for children’s literacy and general career development. Winston also has had his piano debut at Carnegie Hall.
ComplyAI, Dheekshita Kumar and Neha Gaonkar, Chicago and NYC, to build an AI service to speed the process of permit application at local and state governments.
Avi Schiffman and InternetActivism, “leading the digital front of humanitarianism.” Avi is a repeat winner.
Jarett Cameron Dewbury, Ontario, and Cambridge MA, General career support, AI and biomedicine, including for the study of environmental enteric dysfunction. Here is his Twitter.
Ian Cheshire, Wallingford, Pennsylvania, high school sophomore, general career support, tech, start-ups, and also income-sharing agreements.
Beyzamur Arican Dinc, psychology Ph.D student at UCSB, regulation of emotional dyads in relationships and marriages, from Istanbul.
Ariana Pineda, Evanston, Illinois, Northwestern. To attend a biology conference in Prospera, Honduras.
Satvik Agnihotri, high school, NYC area, to visit the Bay Area for a summer, study logistics, and general career development.
Michael Loftus, Ann Arbor, for a neuro tech hacker house, connected to Myelin Group.
Keir Bradwell, Cambridge, UK, Political Thought and Intellectual History Masters student, to visit the U.S. to study Mancur Olson and Judith Shklar, and also to visit GMU.
Vaneeza Moosa, Ontario, incoming at University of Calgary, “Developing new therapies for malignant pleural mesothelioma using epigenetic regulators to enhance tumor growth and anti-tumor immunity with radiation therapy.”
Ashley Mehra, Yale Law School, background in classics, general career development and for eventual start-up plans.
An important project not yet ready to be announced, United Kingdom.
Jennifer Tsai, Waterloo, Ontario and Geneva (temporarily), molecular and computational neuroscience, to study in Gregoire Courtine’s lab.
Asher Parker Sartori, Belmont, Massachusetts, working with Nina Khera (previous EV winner), summer meet-up/conference for young bio people in Hanover, New Hampshire.
Nima Pourjafar, 17, starting this fall at Waterloo, Ontario. For general career development, interested in apps, programming, economics, solutions to social problems.
Karina, 17, sophomore in high school, neuroscience, optics, and light, Bellevue, Washington.
Sana Raisfirooz, Ontario, to study bioelectronics at Berkeley.
James Hill-Khurana (left off an earlier 2022 list by mistake), Waterloo, Ontario, “A new development environment for digital (chip) design, and accompanying machine learning models.”
Ukraine winners
Tetiana Shafran, Kyiv, piano, try this video or here are more. I was very impressed.
Volodymyr Lapin, London, Ukraine, general career development in venture capital for Ukraine.