If doctors can’t help their parents, what good are they?

I’ve been teasing doctor-friends this week by sending them this abstract of a new working paper from the London-based Centre for Economic Policy Research. It’s called “Do doctors improve the health care of their parents? Evidence from admission lotteries” by three economists at the University of Amsterdam—Elisabeth Artmann, Hessel Oosterbeek and Bas van der Klaauw. You can read their entire paper here.

The answer, as you might have guessed from the set-up, is “No.” To quote from the abstract: “Results reject that health outcomes of [Dutch] doctors’ parents differ from those of non-doctors’ parents. This suggests that easy, informal access to medical expertise is not an important driver of differences in health-care use and mortality.”

Mind you, when you look at Dutch doctors’ parents compared to all Dutch parents of similarly-aged children the doctors’ parents do better, at least in terms of living longer. But that comparison is a trap. Doctors’ families could differ from regular families in lots of ways—by being better educated and having higher incomes, to name two—and it might be these other differences that explain the observed difference in health outcomes.

How do the economists deal with this possibly spurious correlation? They take advantage of a kind of natural experiment that the Dutch way of choosing medical school students offered up until it was done away with in 1999. Until then the Dutch chose from all high-school graduates who applied to medical school by using a lottery, though with heavier weighting for students with better grades (which was reassuring to Dutch patients, I assume).

To figure out the effects of having a doctor in the family the three economists compared health outcomes in families where children won the lottery and went on to become doctors with families whose kids qualified for entry into the lottery but didn’t win. In total, they looked at 22,000 families of lottery participants. If the lottery was truly random, and it evidently was, then the families that ended up being compared should have been alike in virtually all respects except for having their child’s number drawn for medical school. By making this comparison you have effectively removed family effects when comparing the health outcomes of parents. And when you do that you get the study’s headline result—there’s no statistically significant difference in outcomes. Having a doctor in the family doesn’t provide a longevity advantage, at least not in the Netherlands.

A couple of conclusions are possible here. One is that Holland has a good health-care system —which it does seem to have despite the fact (Canadians take note) that it’s founded on compulsory purchase of private insurance, complete with deductibles. As a result of this good system, people who don’t have doctors in the family get as good coverage as those who do. A second possible conclusion is more nihilistic—since having a doctor in the family doesn’t seem to bring a health payoff, maybe doctors don’t make that much difference to people’s health. The doctors I’ve sent this result to have responded by musing in turn about what possible benefit there might be from having an economist in the family.

In fact, what I get out of this study has to do more with economics than medicine. A first lesson is that though anyone can make data talk you have to be very clever to ensure your data aren’t talking to you in a biased way. Random double-blind testing is the gold standard in science but because—perfectly reasonably—most people don’t want to be experimented on it’s hard to do that sort of thing in the social sciences. So you have to look for “natural experiments.” The three economists who wrote this study found a good one and made clever analytical use of it.

A second point about economics demonstrated here is that economists study lots of different social phenomena, not just ones revolving around money and finance. We’re always having to explain to non-economists that there’s a lot more to the discipline than that—and a lot more to attract people to it who aren’t necessarily so interested in money and finance.

A third and final point (for today at least) is that Big Data has made it possible to study all sorts of questions with a precision that old-fashioned sampling simply wasn’t capable of. The economists who wrote this study had complete health care and socio-economic data sets for the entire country and did their detailed work with 22,000 families of medical school lottery participants. They don’t actually mention how much of this data was “anonymized.” I assume—and hope—all of it was. Privacy is a big concern with these new technologies. Big Brother himself loved and thrived on Big Data, though Orwell didn’t call it that. But even as we worry about the implications of making all this data available to researchers and, inevitably, in one way or another, to genius hackers, we can marvel at how things have changed from not so long ago when researchers had to try to make inferences from just a few hundred data points.

I hope we use all these new Niagaras of knowledge wisely.