Freakonomics, But for Medicine
Why marathons can be bad for those who don't run them, and how birthdate impacts opioid addiction.
Sometimes, events in the world — whether acts of Mother Nature or changes of policy — will subject a randomly-selected group of people to some condition, while leaving a control group for comparison. These so-called “natural experiments” can facilitate research at a scale and on topics that is otherwise impossible.
In a chapter I cut from The Sports Gene — but published online with an explanation for the cut — I wrote about a tragic natural experiment known as the “Dutch Hunger Winter.”
During World War II, the Nazis cut off food shipments to one northwest section of the Netherlands during the winter of 1944-45. Six decades later, scientists tracked down the children of women who were pregnant during the famine, as well as Dutch children born at the same time outside the famine zone. In adults whose mothers were exposed to famine specifically during their first trimester, the scientists found biological differences that might predispose them to diabetes and obesity. For obvious reasons, that’s not research that can occur in a lab.
My abiding interest in natural experiments meant I jumped at the chance to read an advance copy of Random Acts of Medicine (out today!) by Massachusetts General Hospital physicians Anupam (Bapu) Jena and Christopher Worsham. Jena is also an economist, and in reporting Range I corresponded with him a bit and cited his research on the overuse of certain medical procedures.
Random Acts of Medicine is basically a Freakonomics of medicine — chock full of fascinating natural experiments. Below is an edited chat I had with Jena about topics in his new book. As always, my Q&As are longer than other posts, so there’s a ⚡⚡LIGHTNING ROUND⚡⚡ at the bottom if you just want a few quick, interesting tidbits.
David Epstein: This question is going to be a little painful for me because I’m a big running nerd. I think people might expect marathons to have some impact on public health, but not the one you found, which is that mortality increases in the area around the marathon on race day. Can you explain what’s going on?
Anupam Jena: When a marathon occurs, it's extremely disruptive to a city. More than 26 miles of road are blocked. And that has implications for people who have life threatening medical conditions. If you have a heart attack, and the ambulance needs to get you to the hospital, it's very difficult to bisect the marathon route, so they might take longer to get you to the hospital. And I remember, David, giving a talk a few years ago, and a woman came up to me after the talk, and I was talking about marathons and mortality. And she said, ‘I've got this condition in my colon where I'm predisposed to colonic bleeds, and one day I was at work, and I had a massive gastrointestinal hemorrhage. And I needed to get to the ER, and the ambulance could not get to me because there was a marathon outside my house.’ And she said, ‘My 75-year-old grandmother who lives down the street drove down the road on the sidewalk, in a minivan, and got me and took me to the emergency room.’ And that woman had a cardiac arrest; her heart stopped when she got to the emergency room, so they had to resuscitate her. And she obviously lived to tell me the story. But I was like, wow, you know, I had this idea sort of randomly, but it's so interesting when what you see in the data is corroborated with someone's actual experience.
DE: In the book you go in detail through how increased transit time to the hospital around a marathon has an impact on death rates of people not involved in the marathon. But as with many natural experiments in the book, the effect is small. That said, it is often the case that those small effects impact a huge number of people, so they’re still important. What kind of effect size are we talking about here with marathons and mortality?
AJ: Just to give you a sense, the mortality rate for something like cardiac arrest or a heart attack goes up by about 15 to 20% on the day of a marathon. Now, most people are not having cardiac arrest or a heart attack, so the aggregate impact on a city might be limited. But I think if I were to talk to people about the Boston Marathon bombings, most people would say that was a horrific event. But more people die because of marathon-associated road closures every year in a given city with a large marathon than died in the Boston Marathon bombings. But the bombings, what they did are so salient to us. Deaths in these other channels, we don’t even think about that.
DE: This gets to another point I want to make about the book in general, which is that you often explicitly address the question of whether we can actually do anything about some of these unexpected findings. You’re challenging my personal calculus about the net benefits of marathons, but I do think, with this information, there are things we could do to mitigate the impact of traffic disruption. Maybe we plot courses differently. Or maybe, say, you have certain emergency intersections where, for non-elite runners, you’re ready to pause the race for a minute to let an ambulance through. Adding a minute here or there doesn’t really matter for most people running a marathon, so you could start that system at a certain time. My point is that while I don’t know what to do, I do think there’s actionable info from these natural experiments.
AJ: The routes of the marathons could be altered to ensure that access remains to the hospitals that are involved in the care for high risk patients. Another thing to think about is do we need more ambulances to be dispatched across both sides of the marathon route. The first delay is getting the ambulance to the house, or nursing home. The second delay is getting from the home to the hospital. That first delay can be attenuated if there are just more ambulances floating around. Beyond that, I think this natural experiment shows us that minutes matter for these kinds of cardiac conditions, which I think cardiologists might appreciate but the general public might not appreciate that. If you have chest pain, maybe it’s not something you want to wait around for 30 minutes to see if it goes away.
DE: So for me the topic of marathons is emotionally charged, but I want to jump to a topic that’s more ubiquitously charged: Covid. Specifically birthday parties and Covid. Can you explain how you used children's birthday parties to examine whether politics led to excess deaths?
AJ: Sure. So I was interested in two questions. The first question I was interested in is whether or not the way that people were gathering in the pandemic was a source for Covid-19 spread. There were these huge super-spreader events that people talked about, but then also people were gathering in small groups with people that they know and trust. And you might think that those would not be a significant contributor in the way that a large super-spreader event might be…but it's so hard to study, because you have to figure out who's gathering and where they got Covid-19. Both of those are huge data challenges. And then you would have to figure out is it the gathering that led them to get Covid-19, or the fact that they are traveling a lot, or not washing their hands? And so our insight was that people get together for birthdays, and we could look at insurance data where we knew whether someone had a birthday or not, and look at whether or not they got Covid-19 and compare them to an otherwise identical person in that same city, in that same week, who did not have a birthday. If we saw a difference in Covid-19 rates — and we did see that — that might suggest that people were: a) gathering around their birthdays, which makes total sense, and b) that those small gatherings with people that you know and trust, which we might have thought could be okay, turned out not to be okay. So that was the first question.
The second question we were interested in was: Does that birthday effect vary across places? Early in the pandemic, there was this huge divide between Republicans and Democrats in terms of what people were willing to accept, in terms of masking mandates, school closures, business closures, physical distancing, all of that stuff. And if you look at surveys, Democrats reported being more conservative in those measures in their own lives than Republicans. But surveys can be incredibly misleading, because they don't reflect actual behavior, they just reflect what people say they're doing. And this was a good example of that, because the birthday effect [on Covid spread] was identical in heavily blue areas, and heavily red areas.
DE: A little surprising, I think.
AJ: It’s surprising. And to me it says that when something matters to you, like your birthday, or in particular a child's birthday, what you're willing to do is probably pretty similar whether or not you are a Democrat or Republican. It's not driven by political lines. And so one of the thrusts of that chapter of the book is to say: when we look at the actual behaviors of people, they might be very different than what they said they were doing. And then the other point is that there was a lot of emphasis in the pandemic on how different people were, and how Republicans and Democrats had these very different views about how we should manage the pandemic. But I think we lost sight of the fact that the difficulties that people experienced, that shared experience was so much greater than the differences that people experienced.
DE: And just to be clear, the research you’re talking about showed excess deaths related to birthdays before a vaccine was available. What about post vaccine availability?
AJ: When the vaccine became available, we see marked differences in the mortality rates between Republicans and Democrats. We talk in the book about this really interesting study by researchers at Yale where they were able to identify the political affiliation of people and look at whether or not they died by linking that information to mortality records. So, if you're a registered Democrat or Republican, that information is public record in some places, and whether or not you live or die is public record in some places or can be obtained. And so they linked that together; they showed that the mortality rate pre-pandemic was pretty similar between Republicans and Democrats — not Republican or Democratic areas, but literal registered Republicans and registered Democrats. When the vaccine came out, there's a divergence. The mortality rate for Republicans goes up relative to Democrats. And that's probably related to differences in vaccine uptake, and views about the efficacy of the vaccine versus side effects, maybe questions about workplace mandates and resistance based on those factors. But it all changed after the vaccine came out. And it was a pretty large effect.
DE: One more question about your previous point. You mentioned that people behaved similarly for things that were important to them, such that, when it came to Covid spread around a child’s birthday, mortality was similar for registered Democrats and Republicans pre-vaccine. I suppose one other possibility would be that some behavior actually was different — say, registered Democrats were masking and distancing more — but if so then it clearly wasn’t working as far as mortality was concerned. So you could have behavior that’s different but ineffective, and it wouldn’t show up in the data.
AJ: Right, that's correct. You’re absolutely right, yes. We raise both of those possibilities in the book; either it's that those interventions were being done by registered Democrats versus Republicans and they weren’t working, or both groups were behaving pretty similarly. Either one of those things could be true.
DE: You write more about political polarization in the book, and the idea that people say we need to get politics out of medicine. I think you make a passionate argument that that’s basically nonsense — although you say it much more diplomatically — and that when it comes to public health, medicine has always involved politics. I'm wondering if you can explain that, because I think it was a much more interesting take than the boilerplate “let’s get politics out of medicine.”
AJ: In the book we talk about how the politics of hospitals, or the religious affiliation of hospitals, affects the reproductive or contraceptive services that they might provide patients. We also talk about things like end of life care, which historically have had this political alignment, where conservatives are more forceful about the sanctity of life and whether or not we should continue to treat aggressively at the end of life. And we show that Republican and Democratic doctors actually are very similar in terms of their end of life care treatment preferences, which I think is reassuring. The lesson for me is that sometimes it matters, and sometimes it doesn't matter.
DE: I thought there were a number of reassuring findings. For instance, you showed that a doctor’s political affiliation influenced how they viewed the health of a patient who had had previous abortions, even for conditions that had nothing to do with abortions. If I recall correctly, the finding was that a conservative doctor was more likely to view that patient’s other medical issues as more serious, compared to a liberal doctor’s perspective. So political leanings colored how they felt about unrelated aspects of a patient who had a previous abortion. But at the same time, as you note above, you mentioned that end of life treatment was pretty consistent. And I think — correct me if I'm wrong — but surgical performance was similar even when a doctor and patient had different political beliefs. It didn’t matter for the outcome if surgeons and patients weren’t politically aligned. So you write about some differences in perspective, but that doesn’t necessarily translate to differences in care.
AJ: I think politics and ideology probably matter quite a bit for some kinds of care, and they may not matter for others. The end of life care thing, I think, was probably the most convincing study, because there we actually know the political affiliation of the doctor. And we know what exactly they're doing for the patients, not just what they're saying that they would do.
DE: I find that encouraging.
AJ: I totally agree.
DE: But I don’t want to get too happy here so let’s turn to something less encouraging: opioids. I think you coauthored this study, in which the doctor that someone happens to encounter a single time in the ER can actually make a big difference on the chance of long-term opioid use. Can you explain that?
AJ: We used this natural experiment where we recognize that doctors in the emergency department have different preferences for prescribing pain medication. Some prescribe more, some prescribe less, and it's totally random which patients they see. But if you're a patient who happens to see, by chance, a high opioid-prescribing doctor, you are a more likely to walk out of that emergency room with an opioid. And we see that you are then more likely to become a long-term user of opioids, and there's a true causal effect.
DE: Maybe it’s intuitive in the sense that exposure to opioids in the first place makes someone more likely to get addicted, but it was both fascinating and frightening how you showed the enduring impact of single, chance encounters. And to continue on opioids, there’s a study in the book that looked at whether a patient was about to turn 18, or had just turned 18, and how that influenced the chance that they were prescribed opioids. What you found is that, even if it was just a matter of days, an 18-year-old was more likely to be prescribed opioids.
AJ: You wouldn't think that just because you turn 18 someone would think about you differently, but society does think about people who are 17 years old and 11 months differently than if they're 18 years old and one month. Certainly from a legal perspective, we think about these two individuals differently, but whether or not a doctor would view them differently is a different issue. Physiologically, there's absolutely no difference in what happens overnight when someone turns 18.
DE: Speak for yourself. I just grew muscles everywhere that night. It was wild.
AJ: Yeah exactly, I woke up with a mustache. … But doctors have this heuristic in the back of their mind where they see an 18-year-old and instinctively think of them as being a quote-unquote adult. Whereas for a 17-year-old who's about to turn 18, they see them and think, okay, this is a kid. You're going to be much more careful about prescribing a pain medication to a child than you would be to an adult. And if you look at kids who are 17 years old and 11 months, there is a difference between the likelihood of getting an opioid than if an otherwise similar child is 18 years old and one month.
DE: And this has long-term effects?
AJ: Yes, similar to the emergency room study. Here, the kids who are 18 years and one month are more likely to get an opioid because they happen to show up to the ER right after their 18th birthday, and they're more likely to become long-term opioid users as well.
DE: That's amazing. And I have one just sort of stupid question for my own personal curiosity. One of the thoughts I had — and this is probably silly — was that it might actually be more dangerous to prescribe opioids to an adult because, as a legally recognized adult, they have greater ability to satisfy an ongoing need. A child might have difficulty actually obtaining more opioids. So doctors may perceive prescribing opioids to a child as more dangerous, for obvious reasons, but could it actually be more dangerous to prescribe to an adult who has a better ability to pursue an addiction?
AJ: Good question. I think it could go both ways. The acute effect of prescribing an opioid to a child might be greater than to an adult, just physiologically. If you look in the next seven days at kids and adults, you might have more problems with kids than adults, which is why I think we're more cautious. Then there are the questions of likelihood of physiologic addiction — and I don’t have a good answer for whether that differs. And then access. Adults are going to have much easier access. So you're right, it could have been the case that adults are actually more dangerous to prescribe opioids to because that initial prescription is more likely to take hold, because they can do something about it.
DE: I have about 40 more questions here, and I’m going to bug you again, but for now let’s move to a lightning round.
DE: I’m just going to spit out some topics from your book, and you give me your first association, based on the research you write about. First up: E-ZPass and baby health.
AJ: E-ZPass makes babies healthier because pollution goes down because cars are not stalling on the highway.
DE: Wind direction and mortality.
AJ: Nothing is more random than the wind, and if you live downstream of pollution and the wind is blowing in that direction, you're more likely to have bad health outcomes.
DE: Golf skill and medical specialities.
AJ: Surgeons. Orthopedists are good at golf.
DE: Motor skills and focus! One of the most interesting people I know, Bill Mallon, was a pro golfer, and he’s also a surgeon and editor-in-chief of the Journal of Shoulder and Elbow Surgery. …I went through your footnotes for more of these oddball gems: medical speciality and extreme-speeding tickets.
AJ: Psychiatrists are most likely to get ticketed, but cardiologists who are ticketed are driving the nicest cars.
DE: Practicing medicine in another country prior to practicing in the US and patient outcomes.
AJ: Foreign medical graduates have better outcomes, meaning lower patient mortality, and we think it's because they're the cream of the crop of where they're coming from.
DE: Finally, just to share an interesting finding, you showed that a child’s birthdate has an impact on grandparent mortality. If a kid is born in the fall, the flu vaccine is more likely to be available when they happen to have their annual checkup. They are then more likely to get the vaccine, and their grandparent is less likely to catch the flu from them and die. I think the policy advice here is obvious: from now on, only fall babies.
Thanks to Anupam Jena for his time. Random Acts of Medicine is out today! Pick it up and startle your friends and family with counterintuitive medical research.
And thank you, as always, for reading. If a friend sent this to you, you can subscribe below. (All content is free. Paid subscriptions are entirely voluntary, and much appreciated.)
If you enjoyed this post, please share it.
Finally, thanks to the many readers who left comments on the last post, about pouring out lesser ideas to get to great ones. The thoughts and tips you all shared are fantastic! I’m still working my way through, trying to respond at least in very brief to all of them.
Until next time…