Exploring the Hidden Connections Between Public Events and Health
Major events that capture public attention can create unexpected opportunities to study health. From highly anticipated music releases to national tragedies, these events can function as natural experiments, revealing how sudden changes in circumstances or people’s daily behavior can influence health outcomes.
In this Q&A, resident physician researchers Vishal Patel, MD, MPH, and Michael Liu, MD, MPhil, along with faculty Christopher Worsham, MD, MPH, and Anupam Jena, MD, PhD, discuss why they are drawn to studying events that seem unrelated to medicine, but can actually offer valuable insights into health and health care.
Their research together is based at Mass General Brigham and Harvard Medical School Department of Health Care Policy, with Jena also serving as a research associate at the National Bureau of Economic Research.
What drives you and your peers to explore “hidden” associations between seemingly unrelated events and health outcomes? How do you choose which events or topics to study?
Most health research studies what happens inside the health care system. But a great deal of what determines public health happens outside of it. That premise is a through-line of a book by two of us, Random Acts of Medicine: The Hidden Forces That Sway Doctors, Impact Patients, and Shape Our Health.
We look for moments when something changes abruptly for a large number of people at once, for reasons that have nothing to do with health, and are effectively random: A marathon closes city streets, a national cardiology meeting empties hospitals of specialists, a new music album drops on a Friday. These are accidental experiments, and you can find them happening everywhere if you look for them.
The event must be sharp in time, reach many people simultaneously and most importantly, its timing has to be unrelated to who happens to get sick or injured that day.
And then, of course, there must be a mechanism worth learning something about. If a question makes us say, “Huh, I wonder...” and there is data that can answer it, that is usually the one we go after.
What motivated you to examine the impact of two seemingly unrelated events (mass shootings and major music album releases) on traffic fatalities in your recent papers?
Both studies arose from the same problem: Distracted driving is very hard to study well.
You cannot randomize people to be distracted while on real roads, and self-report does not work — few drivers will volunteer what they were doing with their phone in the seconds before a crash.
So, we went looking for days when the entire country got distracted at once. For example, a mass shooting saturates the news and social media feeds, creating acute stress in people living thousands of miles away. Or a major album release sends tens of millions of people to their phones to listen.
The likely mechanisms are different—one emotional, one behavioral—but both are national in scope, abrupt in timing and unrelated to road or weather conditions. That combination is what makes them usable.
A key finding of the first study was that mass shootings may have ripple effects beyond the people and communities directly affected, influencing unexpected outcomes like traffic fatalities. Were you surprised by this finding?
We were surprised by the magnitude, less so by the direction. Acute stress is known to impair attention and judgment, and there is prior work showing that tragedies affect driving behaviors.
What we did not expect was a 21.9% increase—corresponding to roughly 21 additional deaths nationwide in a single day—or that it held after we removed the state where the shooting occurred and its immediate neighbors.
We also saw that for active shooter events without fatalities, which draw far less public attention, there was no increase in traffic fatalities. If attention and distress are the mechanism, this is what you would expect to see.
Taken together, these traffic deaths raise the total mortality attributable to major mass shootings by about 75% over the shooting deaths alone.
The second study suggests that surges in music streaming (a proxy for increased phone use) may contribute to more traffic fatalities. What do these findings add to the longstanding conversation about phones and driving?
That conversation has been about texting and calling for two decades, and “texting and calling” is where most state laws are focused. But phone use in cars has changed fast.
More than half of U.S. drivers now stream audio behind the wheel, and many use their car’s infotainment system to do so; that is rarely what people picture when they hear “distracted driving” and not what texting-while-driving laws had in mind.
On major album release days, streaming rose 43% and traffic fatalities rose about 15%—roughly 18 additional deaths per release day.
Two secondary findings of this study also stood out to us. The increase was larger in vehicles equipped with Apple CarPlay, which suggests that making streaming easier in the car may not necessarily make it safer overall. Also, there were less fatalities when a passenger was present, meaning someone else could work the phone.
The broader point is that distraction behind the wheel is not just obvious reckless behavior — it includes the routine, apparently harmless things people do without thinking twice.
How do natural, real-world experiments allow you to study questions that would be difficult to study otherwise? What are the strengths and limitations of this approach?
The strength is that they get you close to a randomized trial for questions where a real trial is impossible or unethical. Nobody is going to randomize drivers on real roads to experience a national tragedy or intentional distraction.
Since the timing of these public events is arbitrary with respect to who is doing what on the road, the surrounding days give you a credible comparison. And because we use real-world national crash records rather than a survey, we are not relying on anyone’s memory or willingness to be honest.
Like any research method, there are limitations. These are population-level associations, not individual mechanisms.
We can show that fatalities rose, not that any particular driver was on a phone. We can try to rule out alternative explanations that we are able to test, but not all of them. And the biggest, most impactful events are rare.
We focused on the 10 biggest mass shootings and music albums. That small number limits precision, and it also puts more specific questions (for example, does the music genre or the songs’ tempo matter?) out of reach.
How do you distinguish meaningful associations from mere coincidences in these kinds of natural experiments?
Mostly by making it a habit of trying hard to prove ourselves wrong and presenting those results along with our main findings.
First, we state in advance what should happen if the hypothesis is right. If public attention is the mechanism, the effect should track attention, and it did: Mass shootings with fatalities moved traffic deaths, while those without fatalities did not.
Second, we ask how often a result like ours would show up by chance alone. In the mass shooting study, only one in 10,000 simulations using randomly assigned, fake “placebo” dates of mass shootings produced a larger effect than the one we observed. Comparing each shooting date with the same calendar date in adjacent years ruled out holidays and heavy-travel periods.
Third, if we look within subgroups, any patterns we observe have to make sense. An effect that appears and disappears at random across drivers and conditions is a warning sign.
So, while any single research result might be a chance finding, a result that survives a dozen honest attempts to disprove it is usually not.
How do you hope that different audiences, such as drivers, clinicians, and policymakers, will use or act on these findings?
For drivers, the takeaway is small and practical. Queue the music before you pull out of the driveway, and either let it play, or let the passenger mix things up. If the news has you rattled, it might be a good idea to delay your drive or at least try to minimize any other distractions.
For clinicians, the point is that a major national event can reach our patients even when they are nowhere near it.
For policymakers, two things. First, distracted driving laws were written for the texting era and have not kept pace with how phones are now used in cars. And second, it’s important to recognize that the mortality cost of a mass shooting is meaningfully larger than the count at the scene, which matters for how prevention is valued.
Importantly, these studies do not lead us to conclude that people should stop listening to music or stop following the news. The point is that these are predictable moments of elevated risk that we should keep in mind.
Are there any other recent studies you’d encourage interested readers to explore next? For example, your research on cancer mortality among pilots and flight attendants?
The occupational cancer work, published in JAMA Internal Medicine, is an interesting study. Using national vital statistics covering hundreds of U.S. occupations, we found that pilots and flight attendants had among the highest proportions of radiation-related cancer deaths, while their non-radiation cancer mortality was roughly average.
In other words, we found a group whose ordinary working life delivers a highly unusual exposure and let that stand in for an experiment nobody could run.
For readers who want a deeper tour, Random Acts of Medicine tells stories about data in an approachable way, collecting a decade of this kind of work. It is the fullest version of the thinking behind these two papers and others.
Subscribe to BenchMarks!
Subscribe to our monthly research newsletter and be the first to know what’s shaking in science at Mass General Brigham. From groundbreaking discoveries to the latest “you-heard-it-here-first” breakthroughs, we’ve got the updates that’ll keep your curiosity grooving.