[onomonomono]

There's a typo that appears twice in the appendix where both 'female physician/female patient' and 'female physician/male patient' were labeled female/female in both Table S2 and Table S3.

Also interesting are the raw numbers from Table S2 of patient survival:

                        Female Dr.        Male Dr.
    Female Patient        0.866            0.854
    Male Patient          0.888            0.883

These numbers make it look like both male and female doctors have a harder time saving female patients, the difference being a 2.5% vs 3.3% harder time. It's hard to reconcile this with the paper's abstract where they say "Male patients and female patients experience similar outcomes when treated by female physicians, suggesting that unique challenges arise when male physicians treat female patients."

These numbers are obviously too broad to tell the whole story but if you trained the male doctors to have the same male-to-female patient mortality drop as female doctors, that still leaves 3/4 of the male doctor male-to-female patient mortality difference untouched. That is unless you throw out the .888->.866 drop, which they apparently did, and I wish I had the whole paper in front of me right now to figure out why.

It looks to me like there might be more productive questions to answer here: Why are female doctors outperforming male doctors in this area? Why are female patients (ballpark) twice as sensitive to the gender of their doctor?

source: http://www.pnas.org/content/pnas/suppl/2018/07/31/1800097115...

[onomonomono]

Somebody with more expertise than me needs to chime in here.

If you take the data from Table S2 and perform a two-sample binomial test (as seen here[0] and at 3 minutes here[1]) you can directly contradict a statement in the paper's abstract ("Male patients and female patients experience similar outcomes when treated by female physicians") Using the female doctor numbers I get a z-value of 8.36 which is well over the standard two-tailed significance value of 1.96 for p<0.05. With N being so large it looks like the gender of the patient is absolutely statistically significant with female doctors as well as male.

I feel like I must be missing something here but I'm such a beginner with statistics that I don't know what it is. Is it not a binomial distribution, is it not normally distributed? Am I plugging my numbers in wrong?

[0] - https://stats.stackexchange.com/questions/113602/test-if-two... [1] - https://www.coursera.org/lecture/biostatistics-2/two-sample-...