Patient mortality after surgery on the surgeon’s birthday: observational study

Dear Editor,

We appreciate comments on our article that showed that patients who underwent a surgical procedure on surgeons’ birthday exhibited a higher 30-day post-operative mortality compared with patients who received a procedure on other days.[1] We agree with Osler that the plausibility of the hypothesis we tested is important to address. The plausibility of our hypothesis was supported by three practicing surgeons we consulted (whom we acknowledged in our article), several clinicians (both surgeons and non-surgeons) who provided insightful feedback during research presentations we made, as well as surgeon reviewers. In particular, the potential mechanisms we described in our article were developed based on inputs from practicing surgeons and reviewers. We also agree that statistical significance is different from clinical significance. However, we believe that a 23% increase in patient mortality, if causal, is a meaningful difference from patients’ point-of-view. We also view this study as contributing to a broader economics literature on how performance on complex tasks, in this case surgery, can be affected by external factors, including distractions. Further research is warranted to understand the work environment and schedule that allows surgeons to perform well without distractions.

We addressed a concern raised by Sundar that surgeons’ birthdays may be concentrated on weekends by adjusting for the day of the week of surgery in all analyses we conducted in this article. We have also shown surgeons’ birthdays were distributed evenly throughout the year (see Supplementary eFigure 1), and excluding outlier birthdays (i.e., a larger than expected number of surgeons reported birthdays) did not affect our findings (eTable 13). Furthermore, additional adjustment for the day of the year variable in our sensitivity analysis showed consistent results (eTable 12). Given that we used Medicare claims data, we agree that differential coding may confound our findings. However, for differential coding to explain the observed findings, surgeons would need to code more comorbidities on their birthdays compared with other days of the year. We believe this hypothesis is implausible given that we found no increase in patient mortality for elective surgeries (it is unlikely that surgeons code more only for emergency surgeries on their birthday), and if anything, surgeons may code less on their birthday (which bias our estimates toward the null) if they have competing schedules.

The selection of emergency procedures in our article, pointed out by Smith, was based on inputs from surgeon collaborators we acknowledged in our article. With respect to four cardiovascular surgeries, we used identical ICD-9 codes (eTable 1) as our previous studies, although we used slightly different wording. The exclusion of cancer patients or the use of a slightly different definition of patient mortality (30-day post-operative mortality vs. death before discharge or within 30 days of operation) did not qualitatively affect our findings. Indeed, we found our findings remained largely unchanged by using in-hospital mortality rates instead of 30-day post-operative mortality rate, supporting the robustness of our findings (eTable 5).

Reference
1. Kato H, Jena AB, Tsugawa Y. Patient mortality after surgery on the surgeon’s birthday: observational study. BMJ 2020;371:m4381. doi:10.1136/bmj.m4381

Dear Editor,
R.E. Patient mortality after surgery on the surgeon’s birthday: observational study
Hirotaka Kato,1,2 Anupam B Jena,3,4,5,6 Yusuke Tsugawa1,7
BMJ 2020;371:m4381 | doi: 10.1136/bmj.m4381

As a retired general surgeon, I read with interest Kato et al.’s examination of the effect of a surgeon’s birthday on their patients’ subsequent survival with growing dismay. The authors find that patients suffered higher mortality when surgeons operated on their personal birthday, and the authors go on to suggest that surgeons might be “distracted” by “life events not directly related to work”, such as “… receiving birthday messages on their phones in the operating room”.

The claim that distracted birthday surgeons are responsible for patient deaths, if true, is a damning allegation. But is it true? And is it true of all surgeons?

The suggested scenarios sound implausible to this surgeon, but may seem plausible to the authors of this paper only one of whom has been to medical school, and none of whom have any surgical experience. Absent familiarity with how surgeons conduct their medical practices and their lives, it’s perhaps unsurprising that the authors chose this hypothesis.

But this paper is driven by an hypothesis so odd that one wonders if it was prespecified before seeing the dataset, or the result of a fishing expedition in search of statistical significance. It seems unlikely that the authors undertook the substantial task of cleaning this dataset solely to look at the effect of surgeon’s birth date on mortality, a suspicion confirmed by noting that these authors have used this dataset in a previous publication in this very journal (BMJ 2018; 361). How many other hypotheses did the authors examine, and reject, because they did not lead to a publishable p-value? We have no way of knowing for sure, but all the hallmarks of p-hacking are on display in this paper.

P-hacking, the practice of examining multiple hypotheses in a single dataset until, by chance alone, a p-value is obtained that is “significant” is, of course, statistical malpractice, but seems likely to have been the approach taken by these authors. Marginally significant p-values (0.03) such as the authors report after very significant statistical exertions (29 pages of Online Supplement) are typical of p-hacking; real findings are usually apparent with standard statistical approaches. Moreover, not only are the p-values reported by the authors marginal, but the effect size is minuscule: in the 2,064 operations performed on surgeons’ birthdays, the observed mortality was 7.0%, compared to an observed mortality of 5.6% on surgeon non-birthdays, “… a 23% increase in mortality” as Kato observes, but in reality only a total of 29 deaths (1.4% of 2,064 birthday operations) over 4 years and almost one million operations. Further, these deaths are spread over a total of 47,489 surgeons. So, even if 29 deaths are due to “surgeon distraction”, only 29 of 47,489 surgeons, or 0.06% of surgeons are guilty of such fatal distraction.

In an effort to bolster their conclusion Kato et al. examine the effect of surgeon birthday on complication rates. But, when inconveniently enough they are unable to find such a relationship, they simply wave away this finding, cryptically noting that it is “… perhaps related to the narrowness of he AHRQ’s Patient Safety Indicators for surgery”. However, in my experience, patients rarely die unless something goes wrong, that is, a complication occurs. Deaths without cause seem another indication of an analysis gone wrong.

The authors conclusion, even if true, is of little value: very few lives seem to have been at stake, the authors don’t propose enjoining surgeons from operating on their birthdays, and the authors propose no mechanism for finding the 29 bad-on-their- birthday surgeons. So, what is their point? Disappointingly, the point seems to be adding a publication to the authors’ CVs.

The problems with this manuscript are so profound that Kato’s willingness to impugn the reputation of every surgeon in America is not just wrong; it’s irresponsible. Kato’s manuscript’s conclusions have already been picked up by the lay press, and are rapidly spreading over the internet as “click bait”, no doubt boosting the authors’ careers, but in the process undermining, patients’ trust in their surgical health care team. Patient confidence is a vital asset that must not be squandered on conjectures as silly and pointless as the one put forth in this paper.

Tragically, the damage has been done. Even if the authors withdraw their paper, the internet is already abuzz, undermining the credibility of all surgeons in America, and further estranging patients from our health care system.

Dear Editor,

It is a damning indictment if patients are indeed dying because surgeons are distracted by birthday plans and well wishes from their assistants [1]. However, several red flags suggest the results may have been bolstered by cherry picking.

The study used data for 4 common cardiovascular surgeries and the 13 most common non-cardiovascular surgeries in the U.S. Medicare population. The authors justify their heart surgery selections by references to four similar studies that investigated the relationship between surgical mortality and surgeon age [2], surgeon experience [3], hospital volume [4], and surgeon age and sex [5]. Two of the co-authors of the fourth study were also co-authors of the birthday study.

A comparison of the birthday study with these four papers is helpful for identifying choices that may have been made to bolster the case the authors wanted to make. For example, one paper considers 6 cardiovascular and 8 cancer operations; two papers examined 4 cardiovascular and 4 cancer operations; and the fourth paper considered 4 cardiovascular surgeries and the 16 most common non-cardiovascular surgeries in the Medicare population. The birthday paper’s choice of 17 surgeries is suspiciously peculiar.

There may also be a slight hiccup in the choice of cardiovascular surgeries in that some studies refer to aortic valve replacement (AVR) and others to “heart valve procedures,” which may be a more general term.

None of the other four studies exclude patients with cancer, but the birthday study does, with this unconvincing explanation: “To avoid patients’ care preferences (including end-of-life care) affecting postoperative mortality” [1, p. 2].

The birthday study restricts its population to emergency surgical procedures; only one of the four cited papers does this—the paper with overlapping co-authors.

The birthday study defines operative mortality as death within 30 days after surgery. All four of the other papers (including the paper with overlapping co-authors) define operative mortality as death before hospital discharge or within 30 days after the operation. One paper explains that, “Because, for some procedures, a large proportion of operative deaths before discharge occurred more than 30 days after surgery, 30-day mortality alone would not adequately reflect the true operative mortality.” [3, p. 2019].

Bibliographic references.
1. Kato H, Jena AB, Yusuke Tsugawa Y. Patient mortality after surgery on the surgeon’s birthday: observational study. BMJ 2020;371:m4381 http://dx.doi.org/10.1136/bmj.m4381
2. Waljee JF, Greenfield LJ, Dimick JB, Birkmeyer JD. Surgeon age and operative mortality in the United States. Ann Surg 2006;244:353- 62. doi:10.1097/01.sla.0000234803.11991.6d
3. Birkmeyer JD, Stukel TA, Siewers AE, Goodney PP, Wennberg DE, Lucas FL. Surgeon volume and operative mortality in the United States. N Engl J Med 2003;349:2117-27. doi:10.1056/NEJMsa035205
4. Finks JF, Osborne NH, Birkmeyer JD. Trends in hospital volume and operative mortality for high-risk surgery. N Engl J Med 2011;364:2128-37. doi:10.1056/NEJMsa1010705
5. Tsugawa Y, Jena AB, Orav EJ, et al. Age and sex of surgeons and mortality of older surgical patients: observational study. BMJ 2018;361:k1343. doi:10.1136/bmj.k1343