Patient Complications After Total Joint Arthroplasty: Does Surgeon Gender Matter?

Talia Ruth Chapman, MD; Benjamin Zmistowski, MD; Kaitlyn Votta, BS; Ayesha Abdeen, MD; James J. Purtill, MD; Antonia F. Chen, MD, MBA


J Am Acad Orthop Surg. 2020;28(22):937-944. 

In This Article


The source of patient complications data for this analysis was obtained from a risk-adjusted analysis of Medicare data performed by a nonprofit investigative reporting organization (ProPublica Surgeon Scorecard).[14] The initial analysis by this organization was performed for the purpose of publically reporting surgeon outcomes for widely performed surgeries to improve information available to patients during surgeon selection. The following is a brief description of their methods, which have been described in detail elsewhere.[15] This data set was generated from the Medicare 100% Standard Analytic Files from 2009 to 2013. All patients undergoing primary total hip arthroplasty (THA, n = 494,576) and primary total knee arthroplasty (TKA, n = 1,190,631) during the study period were identified by International Classification of Diseases-9 procedure codes (81.51 and 81.54) and linked to their treating surgeon. From this cohort, 30-day patient complication rates (based on death or readmission secondary to complications stemming from the surgery) were determined separately for THA and TKA for each surgeon. A readmission was deemed a complication if the primary diagnosis for readmission was considered to be related to the index procedure. Relatedness was determined by a panel of physicians and surgeons assessing each identified readmission diagnostic code.[16] Surgeon-specific complication rates were then adjusted for patient age, sex, and overall health score (Elixhauser). The database included surgeon-specific adjusted complication rates (ACRs) for all surgeons who performed a minimum of 20 hip or 20 knee arthroplasties, their name, national provider identifier, ACR confidence interval, procedure type, and procedure volume over the study period for the surgeon. This formed the basis of the data for an analysis of arthroplasty outcomes determined by surgeon gender.

The data set provided by ProPublica was merged (by national provider identifier) to a publically available Medicare dataset—Medicare's Physician Compare data set (Table 1).[17] Current and archived datasets were used. Physicians were included in these data set if they submitted a Medicare claim within the previous 12 months, enrolled in Medicare Provider Enrollment, Chain, and Ownership System (Medicare's enrollment system), and have them listed as a specialty within the Provider Enrollment, Chain, and Ownership System. The Physician Compare database contains demographic data (sex, practice location, name of medical school, and year of medical school graduation), self-reported specialty, group practice identifier, and number of practice members. Of the 9,284 physicians identified as arthroplasty surgeons in the ProPublica database, 8,742 (94.1%) were included in the Physician Compare database. To improve the classification of the surgeon's gender, the Medicare Provider Utilization and Payment Data[18] was used. This provides physician-specific data including sex, credentials, location, and billing characteristics. From this, the gender for an additional 223 surgeons was identified, for a combined cohort of 96.5% (8,965 of 9,284) of orthopaedic surgeons. In cases of discordance between the Physician Compare and Utilization databases, the surgeons were independently reviewed to identify the accurate sex. The Provider Utilization data set also provided credentials for surgeons when available. Medical schools for each surgeon were consolidated into a list. The location and medical school rank were established using the US News ranking in 2016.[19]

To adjust for geographic influences of outcome from arthroplasty, physicians were grouped by regions and divisions as defined by the United States Census Bureau based on their state of practice. In addition, the 2000 social deprivation index was obtained from the Health Innovation Program at the University of Wisconsin-Madison School of Medicine and Public Health.[20] Reported by nine digit zip codes, the Area Deprivation Index is a single local representation of the populations' education, employment, income, and personal assets. This was used as a reflection of the social resources in the communities being served.

The primary purpose of this analysis was to identify the differences (with the null hypothesis of no difference) in complication rates for common orthopaedic procedures (hip and knee arthroplasty) based on surgeon gender using this combined database. First, collected variables (procedure type, surgeon procedure volume, surgeon years since graduation, surgeon gender, and geographic region) were assessed for any association with the surgeon-specific ACR. To assess this, a t-test was performed to compare continuous variables between the groups, a Pearson correlation coefficient was calculated to compare two continuous variables, a Chi-squared analysis was performed to assess categorical variables, and an analysis of variance was performed to compare continuous variables across multiple groups. Skewness and kurtosis less than two and 12, respectively, were used to confirm data normality. For comparison of non-normal data between the two groups, a Wilcoxon rank sum test was used. To confirm the independence of potential predictors of outcomes after arthroplasty, a multivariate linear regression including variables with a P-value < 0.1 in univariate analysis was undertaken. In variables with non-normal data distribution, a log transformation was performed; skewness and kurtosis were used to confirm data normality in the new distribution. For each analysis, cases with missing data for that analysis were excluded. Analysis was performed with RStudio (Version 1.0.136).