Adapted from “Random Acts of Medicine: The Hidden Forces That Sway Doctors, Impact Patients, and Shape Our Health” by Anupam B. Jena, Joseph P. Newhouse Professor of Health Care Policy and associate professor of medicine and associate physician at Massachusetts General Hospital, and Christopher Worsham, pulmonologist and critical care physician at Massachusetts General Hospital, instructor of medicine at Harvard Medical School, and teaching associate at the HMS Department of Health Care Policy.
Chance occurrences change the course of our lives all the time. Imagine the couple who meets in an airport when a blizzard cancels their flights. Or think of the entrepreneurs who happen to sit next to each other at jury duty and who hatch a business plan to start a company in the six hours it takes to be dismissed. Imagine the woman who misses the bus because a meeting ran late and who, on her walk home, happens to pass an animal shelter and adopts her new best friend.
These are unpredictable moments — chance waltzing into our lives when we least expect it. We can all probably think of times in our lives when we were led down alternate paths, both good and bad, thanks to chance. These phenomena go by many familiar names: randomness, luck, serendipity, kismet, happenstance, accidents, or flukes.
Chance can even play a role in life and death. A retiree collapses in the supermarket from a heart attack and is rushed to the hospital as paramedics perform CPR, but the ambulance is delayed due to scheduled road closures. He dies from the heart attack two weeks later. If he had collapsed the day before when the roads were clear, might he have survived? A child doesn’t get a flu shot at his annual doctor’s visit, because the vaccine is not yet available in his doctor’s office. His parents never bring him back for a flu shot appointment. That winter, he catches influenza and spreads it to his grandmother, who ends up in the hospital. Had the flu shot been available at his annual visit, he likely would have received it. Would he and his grandmother still have gotten sick?
It’s frightening to think about how random occurrences contribute to our health, life, and death. We all like to think that if we do the right things — eat well, wear a seat belt, quit smoking, take the medications our doctor prescribes — we can control what happens with our bodies and our lives.
This is no less true for doctors. We too like to think that the decisions we make for our patients — whether to prescribe them a drug, perform surgery, order a diagnostic test — are based only on science and carefully considered data, not on simple chance. The reality, though, is that medicine can be messy, complicated, and uncertain. There are many opportunities for randomness to affect the medical care we give and receive.
Most people tend to think in terms of “good luck” and “bad luck”: the good luck of getting to the bus stop just as the bus arrives, or the bad luck of driving over a nail and getting a flat tire. But in everyday medicine, people are sent down paths of care by factors they may not think to consider — the doctor who happened to be working in the ER the day they sprained their ankle, or the patient they happened to share a waiting room with just prior to a routine doctor’s visit. It’s not “good luck” to sprain your ankle on a Tuesday or “bad luck” to sprain it on a Wednesday — it’s as random as a roll of the dice. Yet the day a person sprains their ankle may determine which doctor happens to treat them in the ER, and thus the likelihood they’re prescribed an opioid pain medication that could lead to long-term use. Because opioid-prescribing tendencies vary from doctor to doctor, as one important study demonstrated, the doctor you happen to see can have lasting repercussions. Similarly, it’s not inherently “good luck” or “bad luck” to share a doctor’s office waiting room with someone. But if that person happens to have a viral infection, a chance encounter could lead to a bout of the flu two weeks later, and especially for the very young and the elderly the flu is no laughing matter.
* * *
Economists, epidemiologists, and social scientists sometimes talk about “natural experiments.” Natural experiments are “natural” because they occur without the influence of any manipulating hand. One person grows up in one ZIP code; another person grows up across the street, which happens to be in a different ZIP code. One baby is born into a season of drought; another is born during record monsoons. There is no researcher designing a study, no patients signing up to participate, no new medical intervention being intentionally tested. These are conditions for accidental experiments, science occurring in the wild.
Natural experiments run in contrast to what we might think of when we use the word “experiment.” In medicine, randomized controlled trials — the gold standard of science, where researchers randomly assign subjects to either a treatment or a control group and then follow them into the future — are our most powerful and preferred tool for studying cause and effect. They are our best way of knowing whether an intervention really works. These are the studies that have been used for decades to prove the efficacy of the blood pressure drugs, cancer therapies, and vaccines we use today.
But randomized controlled trials aren’t perfect. They can be logistically difficult, expensive to perform, take unreasonable amounts of time, or be flat-out unethical. Imagine you were interested in studying the effect of air pollution on human health. A scientist couldn’t simply assign human test subjects to regions with different levels of air pollution and observe the results. Or imagine you wanted to study the long-term effects of screen time on children. Assuming you could surmount the complicated ethics of such a controlled study, you might still have to wait dozens of years to see results, at which point your study might have ceased to be relevant.
So researchers in some fields — economics in particular — have come to rely on natural experiments in their work. How so? Let’s circle back to air pollution. You can’t just purposefully expose people to air pollution, but what if scientists could isolate a naturally occurring event in which some defined groups of people were exposed to higher levels of air pollution than some other groups were — by nothing more than chance? Actionable conclusions could be drawn from those findings.
In one study, the Princeton economist Janet Currie and the UC Berkeley economist Reed Walker did precisely that. They showed that among families living near congested highway tollbooths in Pennsylvania and New Jersey, babies who happened to be born just before E-ZPass automatic payments were introduced were more likely to be born prematurely and with low birth weights than the babies born just after E-ZPass, when air pollution levels dropped alongside traffic congestion (since cars no longer needed to wait in long lines at tollbooths).
You might be unconvinced. What if there were other factors at play? What if the types of mothers living near the toll plazas were somehow different — older or younger, more or less healthy — before and after E-ZPass started? The researchers wondered the same thing. But their analysis didn’t yield any major differences in the two groups, and the overall result was unchanged even after statistical adjustments were made for small differences in smoking, teenage pregnancy, education, race, and birth order (moms’ second versus third baby). The researchers even considered whether the findings were driven by health-conscious prospective home buyers, who knew air pollution would be reduced in the area, being more likely to move into the neighborhood. If that actually occurred, the greater demand should have driven up housing prices. But they found no differences in housing prices near toll plazas before and after E-ZPass. They could only conclude that E-ZPass meant less air pollution, and thus improved birth outcomes in the vicinity.
Here’s another natural experiment in a similar spirit, this time by the University of Illinois Urbana-Champaign economists Tatyana Deryugina, Nolan Miller, David Molitor, and Julian Reif and the Georgia State University economist Garth Heutel, who were interested in the health effects of air pollution on elderly patients. The researchers looked at the death rates of certain patients on days when the wind happened to be blowing polluted air into a certain area compared with days when the wind blew away, directing pollution elsewhere. Is there any purer example of the role of chance than, literally, which way the wind blows? Sure enough, the economists found convincing evidence (“statistically significant” evidence, they would say) that days with greater incoming air pollution led to higher hospitalizations and death among the elderly of that region.
In both instances, health outcomes were affected by chance — the mother who happened to deliver her baby after E-ZPass was introduced near her home, the elderly adults whose health outcomes were affected by the wind direction. But in both cases, the role of chance was measurable. This is more than just interesting data; it helps us rigorously quantify the effect of air pollution on our health, something no randomized controlled study could ever ethically achieve.
* * *
The beauty of natural experiments that occur in medicine is not only that they have the power to let us uncover problems in our health-care system that cannot be readily answered with traditional research but that they point us toward potential solutions.
One of the first natural experiments ever studied was an influential study that many call the physician who wrote it, Dr. John Snow, the “father” of the field of epidemiology — the study of diseases within populations.
In 1854, London was suffering an outbreak of cholera, a diarrheal disease that would often lead to death from dehydration. At the time, it wasn’t known how cholera was spread. Because the disease caused gastrointestinal symptoms, Snow hypothesized that patients who contracted the illness had ingested something that caused disease. When an outbreak occurred among people living in a specific neighborhood, he began to investigate. There were dozens of deaths concentrated in the area, but he found it curious that some neighbors were completely unaffected while others became sick. He studied those who became sick or died, and he found that they had drunk from a specific water pump — how Londoners got their water at the time. Meanwhile, healthy neighbors who were otherwise similar — living in similar physical conditions, making a similar income, eating from similar food sources — happened to get their water from a different nearby pump, drawn from a different water supply. The water source could be the only culprit.
This was a natural experiment at work. The findings supported Snow’s hypothesis that whatever was causing cholera was being ingested by sick patients. For further confirmation, Snow had the handle removed from the offending pump, preventing locals from getting their water there. The result? Cases of cholera declined. As it turned out, the water supply at that pump had been contaminated by sewage from a household with an early infection. And because his experiment was “natural,” he was able to establish cause and effect without ever seeing bacteria under a microscope or doing any of the modern tests we would rely on today. He simply collected data and analyzed it with an eye toward causality. We know now that Snow was right: Vibrio cholerae is the bacteria that causes cholera, and it’s spread via ingestion of food or water contaminated by infected sewage.
Our health problems today are different in countless ways from those of Snow’s time (though, clearly, infectious disease remains a threat). But striving to improve and prolong life, free from disease, remains a priority of modern medicine, and the natural experiment remains as powerful a tool now as it was in 1854.
* * *
The challenge in identifying and studying natural experiments in healthcare, or in the world more broadly, is that it’s not always obvious when they’re happening! (Part of why we remain gainfully employed.) It takes practice to find them hiding in the data — we’re not all born with the instincts of John Snow. Indeed, the ways in which we’ve stumbled onto our studies have often been accidental, as much a product of chance as anything we discover in our research — a chance conversation with a spouse, a chance encounter while waiting in line for coffee, a chance experience with a patient or colleague.
Our hope is to show you that there’s much for us to learn from these natural experiments, as doctors and patients alike. We can expose the hidden forces in our health care that send two otherwise-similar people down very different paths of care, by chance alone. By examining those forces, we can identify their implications in the broader context of medicine and our daily lives.
Published by arrangement with Doubleday, an imprint of The Knopf Doubleday Publishing Group, a division of Penguin Random House LLC. Copyright © 2023 by Anupam Jena and Christopher Worsham.