On HN, we need to do better than the internet trope of dismissing a study with a single drive-by, usually obvious objection. No doubt there are a lot of bad studies, but bad studies plus bad comments about them is worse.

We should inhibit the impulse to dish out snark for the brief thrill of feeling smarter than an author or researcher, not because they don't occasionally miss the obvious (who doesn't?), but because it lowers the standard of discussion here.

A better way to present such a comment might be as a question: "In my experience, X. Does the study Y?"

In presenting a possible explanation I didn't think a little humour would be taken so negatively

I thought that as more younger doctors are women, then perhaps younger doctors are better. I'm always a bit disappointed that science news doesn't dig that one level deeper. You'd hope they'd cover both the ideas we came up with in 5 seconds in the paper, but if they don't mention it in the story then you're left guessing.

> perhaps younger doctors are better.

That's an interesting observation. On the one hand younger doctors have been more recently exposed to "current" ideas, research, and stuff. On the other hand younger doctors lack the experience to catch things that are unusual and often discussed as theoretical.

I'd be curious to see if there were research on this topic now!

They did!

The paper reports that they included both physician age AND years in practice, along with a slew of other factors.

The paper says they "adjusted for patient and physician characteristics" which presumably includes the patient's sex.

Concretely:

> We accounted for patient characteristics, physician characteristics, and hospital fixed effects. Patient characteristics included patient age in 5-year increments (the oldest group was categorized as ≥95 years), sex, race/ethnicity (non-Hispanic white, non-Hispanic black, Hispanic, and other), primary diagnosis (Medicare Severity Diagnosis Related Group), 27 coexisting conditions (determined using the Elixhauser comorbidity index28), median annual household income estimated from residential zip codes (in deciles), an indicator variable for Medicaid coverage, and indicator variables for year. Physician characteristics included physician age in 5-year increments (the oldest group was categorized as ≥70 years), indicator variables for the medical schools from which the physicians graduated, and type of medical training (ie, allopathic vs osteopathic29 training).

They don't say _how_ they controlled for those characteristics though. Presumably, they divided each patient's calculated 30-day risk of death [1, table 10] by the relative ratios in risk of death of every category, calculated from the same data set. That should probably cut it for controlling for this particular difference, although I am not a statistician.

Paper link:

https://jamanetwork.com/journals/jamainternalmedicine/fullar...

[1] Supplemental material https://jamanetwork.com/data/Journals/INTEMED/0/IOI160102sup...

It's a regression model, so "accounted for" usually means that the factors were included in the models.

Ideally, we'd ask a group of male doctors and a group of female doctors to independently diagnose and treat the same set of patients. We can't actually do that--in addition to the cost, you obviously can't treat the same patient twice. However, we can try to estimate this effect statistically.

First, they build a model describing the probability of a patient dying within 30 days. They used a linear probability model, which essentially means that the probability of someone dying is the sum of the "weights" related to the patient, doctor, and hospital. These weights are estimated from the data using ordinary least squares (the same way you may have learned to fit a line to points at school).

Having built the model, they then ask what the marginal effect of the doctor's sex is. In other words, if you hold everything but that constant, how does the probability of dying change? They cite a nice Stata guide (#32 in the paper here: http://www.stata-journal.com/sjpdf.html?articlenum=st0260 ) which gives some background and examples.

The rest of the paper looks at different variations (only hospitalists, different diseases, etc), using a pretty similar approach.

They only looked at hospital visits, which they assume are pretty randomly distributed based on the physicians' work schedules. The patients don't get to pick their doctors in that case.

Is there a relationship between shifts, gender, and mortality? E.g. more male doctors take more night shifts and night shifts have higher mortality

Ha! That jibes with my experience too, my wife prefers a female doctor, and I do too for her. In fact, my mother and sisters have the same preference come to think of it.