Surgeon Volume and Operative Mortality in the United States.

Nearly 300 studies have been published over the last 25 years regarding the relationship between improved patient outcomes and greater hospital case volume. In these studies the relationship between hospital volume and outcomes has been robust and persistent for a wide variety of diseases and procedures. Despite this, many questions remain as to the underlying mechanisms responsible for improved outcomes at high volume centers. One potentially very important factor, the experience of individual surgeons, has received less attention. A recent article by Birkmeyer and colleagues (New Engl J Med 349:2117–2127, 2003) addresses this issue by focusing attention on the experience of the individual surgeon as a factor in determining patient outcomes following surgery. The national Medicare claims database for 1998-1999 was examined to determine mortality among nearly 475,000 patients following eight surgical procedures. Nested regression models were used to determine the relationship between operative mortality and surgeon volume as well as between mortality and hospital volume. Surgeon volume was inversely related to operative mortality for all eight procedures with adjusted odds ratios for operative death between low volume and high volume surgeons ranging between 1.24 and 3.61 depending on the operative procedure. For most procedures, mortality rates of low volume surgeons at high volume centers were greater than mortality rates of high volume surgeons at low volume centers. In other words, it was the surgeon and not the institution that had the greatest effect on outcome.FigureFor the one neurosurgical procedure studied, carotid endarterectomy, the findings were as follows: The adjusted odds ratio for operative mortality between low volume and high volume surgeons was 1.64, with a 95% confidence interval of 1.47–1.87. If the data were adjusted for hospital volume, the odds ratio was 1.70 with a 95% confidence interval of 1.51–1.91. The proportion of the effect of surgeon volume that could be attributed to hospital volume was 0% In other words, low volume surgeons had a mortality rate that was more than 1.5 times higher than the mortality rate of high volume surgeons, regardless of the total number of endarterectomies performed at the hospital. The case volume of the institution in which the endarterectomies were performed had no effect on outcome. This is a very important paper. It indicates that there is a positive association between the experience of individual surgeons and patient outcomes that is independent of the case volume of the hospitals in which surgeons practice. Although this has been more difficult to document than the relationship between hospital case volume and outcomes, it is no surprise. Most neurosurgeons recognize that surgical talent varies from individual to individual and that affiliation with a high volume hospital does not miraculously confer surgical skills. This paper raises some fascinating questions. Do surgeons improve their skill by performing more operations or do more skillful surgeons get more referrals? If it is the former, regionalizing care by imposing referral guidelines should lower surgical morbidity and mortality by increasing the experience of a small number of surgeons. However, if surgical skill is the independent variable, mandated regionalization of surgical care could direct patients away from high quality surgeons at low volume institutions to low quality surgeons at high volume institutions with disastrous results. This should be food for thought for those who have blithely advocated regionalization of surgical care as a means to improve patient outcomes. These results cannot be necessarily generalized to neurosurgical procedures other than carotid endarterectomy, which were not studied herein, although similar results have recently been widely published for intracranial aneurysms. And there may be exceptions related to certain low volume procedures, where overall surgeon experience with other operations may confer surrogate competence in others. The patterns will need to be studied specifically in the settings of subspecialty teams, housestaff involvement, and community versus academic practices, where the effects may not necessarily be the same, and will need further analysis. The findings of this paper also call into question the validity of results from randomized clinical trials in which the number of procedures per practitioner in one arm of the trial is significantly different than the number of procedures per practitioner in the other arm. Such trials are likely to be evaluating the relative skills of the practitioners and not the safety of the procedures per se (1). ROBERT E. HARBAUGH, M.D. OUTCOMES AND POPULATION SCIENCE