Read on Twitter
Do surgeons have worse outcomes when they operate on their birthday?
Our @bmj_latest X-mas study, led by Hiro Kato & @ytsugawa1
We analyzed 1M surgeries, linked to surgeon & surgeon's birthday
We found higher mortality for surgeries performed on a surgeons bday
Thread
Our @bmj_latest X-mas study, led by Hiro Kato & @ytsugawa1
We analyzed 1M surgeries, linked to surgeon & surgeon's birthday
We found higher mortality for surgeries performed on a surgeons bday
Thread
2/
Study here: https://www.bmj.com/content/371/bmj.m4381.full
It was a fun question but broader idea was to identify situations where surgeons may be distracted and to study the effect of those distractions on their patients' outcomes.
Why distracted? Texts, phone calls, pages, evening plans, etc.
Study here: https://www.bmj.com/content/371/bmj.m4381.full
It was a fun question but broader idea was to identify situations where surgeons may be distracted and to study the effect of those distractions on their patients' outcomes.
Why distracted? Texts, phone calls, pages, evening plans, etc.
3/ There are, of course, other things that can distract surgeons but birthdays offer an "exposure" that is individualized and allows us to study variation in outcomes within surgeons on cases performed on their birthdays vs on other days, a within-surgeon analysis.
4/ The basic finding is that 30-day mortality for patients undergoing surgery on a surgeons' birthday is appreciably higher than on other days (~7 vs 5.5 percent 30-day mortality). This is a big effect and we recognize that it may be "too big." The event study plot is below.
5/ Here is the regression table.
6/ Of course, the key concern here is confounding. Are patients who undergo surgery on a surgeon's birthday at higher risk? Could be if surgeon's alter their schedule and prioritize sicker patients.
We took this issue seriously. We showed:
a. Balance of patient characteristics
We took this issue seriously. We showed:
a. Balance of patient characteristics
7/
b. Similarity of surgery types performed, i.e., not higher risk surgery types being performed.
c. Predicted mortality was similar for surgeries performed on surgeons' birthdays vs other days. The same is true if you plot predicted mortality as an event study plot, shown below
b. Similarity of surgery types performed, i.e., not higher risk surgery types being performed.
c. Predicted mortality was similar for surgeries performed on surgeons' birthdays vs other days. The same is true if you plot predicted mortality as an event study plot, shown below
8/
d. There is no reduction in number of procedures performed on surgeon's birthdays vs other days, which goes against the idea of surgeons selecting fewer, higher risk/need cases. Plot here.
d. There is no reduction in number of procedures performed on surgeon's birthdays vs other days, which goes against the idea of surgeons selecting fewer, higher risk/need cases. Plot here.
9/
e. On extensive margin, we find that surgeons are less likely to operate at all (ie, perform a single surgery) on bdays. This doesn't affect analysis, which looks at surgeons on their bdays vs other days. But shows bdays are meaningful enough to affect practice behavior
e. On extensive margin, we find that surgeons are less likely to operate at all (ie, perform a single surgery) on bdays. This doesn't affect analysis, which looks at surgeons on their bdays vs other days. But shows bdays are meaningful enough to affect practice behavior
9/
We also looked at whether surgeon birthday effect was greater when birthday was for a milestone (e.g., age 40, 50, 60) or if the birthday occurred on a Friday, in which case evening plans or weekend getaways may be more likely and more distracting. No evidence of either
Fin
We also looked at whether surgeon birthday effect was greater when birthday was for a milestone (e.g., age 40, 50, 60) or if the birthday occurred on a Friday, in which case evening plans or weekend getaways may be more likely and more distracting. No evidence of either
Fin
