New and Experienced Surgeons Have Similar Patient Mortality Rates

Samuel D. Pimentel Doctoral student Statistics Department Wharton School of the University of Interview with:
Samuel D. Pimentel
Doctoral student Statistics Department
Wharton School of the University of Pennsylvania


MedicalResearch: What is the background for this study? What are the main findings?

Response: Surgical training has undergone major changes in recent years – including a reduction of six to twelve months of training time – and there is controversy about whether these changes have been good or bad for patient outcomes.   Our work partially addresses the issue by asking whether newly-trained surgeons perform better or worse than experienced surgeons.  We compared surgical patients treated by new surgeons to a similar group of patients treated by experienced surgeons using a new statistical technique called large, sparse optimal matching.  Our analysis found no significant differences in mortality rates between the two groups.

MedicalResearch: What should clinicians and patients take away from your report?

Response: It is reassuring that new surgeons achieved similar mortality rates to their experienced counterparts when operating on similar patients.  However, mortality is a rare event and mortality rates may not expose the benefits of surgical experience.  Furthermore, the changes in surgical training happened very recently and it is not clear how fully these results will generalize to surgeons trained under this new regime.  So further follow-up is needed to confirm and clarify our findings.

The findings also demonstrate the importance of accounting for pre-existing differences in patient populations in clinical research.  The patients treated by new surgeons in our data had higher risk factors and more severe illness than the patients of experienced surgeons, and they were more also likely to have been admitted to the hospital via the emergency room.  Only after conducting large, sparse optimal matching, a statistical methodology designed to remove these discrepancies, were we able to make a fair comparison and recognize the absence of a significant difference in mortality.  Clinical researchers should be aware of the importance of making such statistical adjustments, and of the availability of effective new matching technologies that can perform them.

MedicalResearch: What recommendations do you have for future research as a result of this study?

Response: Further comparative analyses of new and experienced surgeons are needed in order to take into account additional factors of interest.  For example, it would be valuable to consider additional outcomes (such as post-operative complications and health care costs) and to conduct targeted analyses of clinically relevant subsets of patients.  In addition, the original analysis was conducted on a pilot database from Medicare covering only a few states.  A larger and more representative dataset is needed to obtain more precise statements about equivalence between patient outcomes and to ensure that our results generalize.  We have applied for funding to run a broader study along these lines.

In addition, the statistical methodology we demonstrate should be widely applicable in clinical research.  In many settings where a treatment or procedure has been administered to patients in a non-randomized manner, large, sparse optimal matching can help researchers make fair comparisons between the outcomes of those who received the treatment and those who did not using existing medical databases.


Samuel D. Pimentel, Rachel R. Kelz, Jeffrey H. Silber, Paul R. Rosenbaum. Large, Sparse Optimal Matching with Refined Covariate Balance in an Observational Study of the Health Outcomes Produced by New Surgeons. Journal of the American Statistical Association, 2015; 0 DOI: 10.1080/01621459.2014.997879 Interview with: Samuel D. Pimentel (2015). New and Experienced Surgeons Have Similar Patient Mortality Rates

[wysija_form id=”1″]

Last Updated on February 25, 2015 by Marie Benz MD FAAD