Risk aversion in surgeons is widely believed to be true but the evidence suggests it isn't universal
The potential for surgeons to behave in risk averse ways occurs when surgeon level outcome data is published. This clearly creates an incentive to select patients whose risk of complications is low so that the published statistics for the surgeon look better.
But–especially when the evidence is handled carefully including proper risk-adjustment and careful presentation–it is less clear that the incentive to be risk-averse is clear and simple. If risk-adjustment can be done well and presented well so even the public can understand the results, there may be little real incentive to avoid risky patients.
The belief that open publication of outcome data is a problem is clearly widespread. This 2014 article in the BMJ is one of a series arguing against surgeon-level outcome data for related reasons (note the date for context later).
The UK experience of using surgeon-level data is a useful story. In the 1990s a whistleblower exposed a serious problem with heart surgery in children. The Bristol Heart scandal as it is known exposed serious deficiencies in surgical competence in one department that had been hidden for some time because there was no open publication of surgeon-level outcome data. One outcome of the eventual enquiry prompted the NHS to promote open publication of risk-adjusted surgical outcomes for all cardiac surgeons. The profession resisted this at the time (largely because of the supposed incentives for surgeons to "game" their numbers by risk-averse selection of patients but they were browbeaten into accepting the proposal by the NHS Medical Director, Bruce Keogh and public pressure from the Bristol scandal.
But the early results of the use of surgeon-level data did not create the expected effects. One BMJ report on the results in 2009 "Survival after heart surgery continues to improve after publication of mortality data" reports:
Outcomes for adult cardiac patients have improved substantially in the past five years, even though more elderly and high risk patients are now being treated. Mortality rates after coronary artery surgery have fallen by 21% and for isolated valves by a third, a study by the Society for Cardiothoracic Surgery of Great Britain and Ireland has found...
Ben Bridgewater, consultant cardiac surgeon at the University Hospital of South Manchester and author of the report, said, “One of the benefits we are now seeing from public reporting of outcomes is not just about bringing poor performers ‘into the pack’ but improving the performance of the pack as a whole. The very act of auditing services brings about improvements as centres learn from one another.”
And these results were so convincing that the Royal College of Surgeons though other specialties should also openly publish their data (my emphasis):
The results have prompted the Royal College of Surgeons of England to urge all surgical specialties to follow the lead as soon as possible and publish mortality data.
The college said that although critics expected that publishing mortality data would lead to risk averse behaviour from surgeons, with the most sick and elderly patients denied surgery for fear of worsening the statistics, in practice, the opposite has turned out to be true.
Note that this report was published 5 years before the 2014 article arguing the opposite. As one critic responding to the 2014 argument summarised well (my highlighting):
The article by Steve Westaby dealing with publication of surgical mortality data is at best a superficial examination of the issue and at worst misleading. The evidence has been irrefutable for many years that the collection and feedback of risk-adjusted mortality data will reduce mortality rates by between 40% and 24% in cardiac surgery and other specialties. The issue of ‘gaming’ cardiac surgery outcomes by avoiding high risk patients was proposed and investigated at the time of the initial publication of the results and no evidence was found to support the assertion. More recently it has been suggested that without the knowledge of their own mortality or complication rates and that of their colleagues it is not possible for surgeons to obtain full informed consent from their patients. Thus complication and mortality rates become a tool for continuous quality improvement, patient information and informed consent rather than a stick with which to beat the surgeon.
The UK data and discussion illustrates two key things. One is that many in the profession still believe that open data promotes risk-aversion. But the second is that this is refuted by the actual experience of publishing the data.
Whether this is universal in different health systems is unclear. The NHS has few if any direct financial incentives for its surgeons but the US system has strong financial incentives which might promote the pursuit of volume rather than quality. It is clear, though, from the UK data that the expected risk-aversion does not always occur when data is published and that publication can lead to systematic improvement over time.