“Patient–physician gender concordance and increased mortality among female heart attack patients” by Brad N. Greenwood, Seth Carnahan, and Laura Huang was just published in PNAS. It is getting lots of positive press. From The Atlantic:
Women More Likely to Survive Heart Attacks If Treated by Female Doctors
And male doctors do better when they have more female colleagues.
They showed that women are more likely to die when treated by male doctors, compared to either men treated by male doctors or women treated by female doctors.
“There are inequalities in a lot of different contexts, but when someone is suffering from a heart attack, you might expect that there would be no gender differences because every physician will go in trying to save their patient’s life,” says Huang, a professor of organizational psychology at Harvard Business School. “But even here, we see a glass ceiling on life.”
I have several concerns about this article:
- The authors make strong causal claims — both in the article itself and in press commentary — even though this is a 100% observational study. Physicians are not randomly assigned to heart attack victims. If male doctors are, for example, more likely than female doctors to seek out more difficult cases, then we would expect their patients to have higher mortality even if physician skill is equal.
Their description of the assignment mechanism is problematic at best:
In addition, when patients visit the emergency room (ER), they have little agency over their choice of attending physician, allowing for a quasirandom assignment of physician and patient.
- Why don’t they provide some information about how emergency rooms actually work? This is the single most important issue in determining how much credence to give their conclusions. If there is a fixed rotation and physician X is just assigned whichever patient comes in next, then “quasirandom” might not be a bad description of the process. But if nurses/physicians/administrators play any role in assigning patients to physicians, then any causal conclusions are deeply suspect.
- The authors have very little experience in writing about medical issues (I see a couple of articles by Greenwood) and no prior publications (that I can find) involving heart attacks or emergency rooms. Why not involve someone with relevant knowledge, especially since the entire article stands or falls on the basis of how physicians are assigned to patients?
- The authors, at least Huang, have a clear ideological preference for the result which, in fact, they did find. Female physicians are better than male physicians and, if we do have to have any male physicians, they perform best (although still not as well as female physicians) in departments with more women.
- There are a variety of technical issues (and general sloppiness) which also makes me suspicious. Using a linear model with a 0/1 dependent variable is just nonsense, even if it makes interpretation easier. The most important plot (Figure 2) is mislabeled. And so on.
Greenwood, at my request, kindly shared the code with me, but then declined my request to make it public (or allow me to share it here or with students). I never trust an empirical claim which does not provide this sort of transparency. Recall the exemplary behavior of the folks involved with the NEJM paper on Maria mortality.
Again, the major issue is making a strong causal claim when, at best, you have some evidence of correlations. Note Huang’s language: “Women are more likely to die from heart attacks when treated by male physicians.” That might be true, but this paper, at most, provides some hints in this direction. Why would Huang (and her co-authors) engage in such obvious exaggeration?