Analysis of Physician Compensation Studies by Gender, Race, and Ethnicity

Purpose: This report investigated physician compensation studies by gender, race, and ethnicity. Methods: Published U.S. physician compensation studies were assessed. Results: Of the 47 data sets within 46 studies, 36 analyzed compensation by gender and 32 (88.9%) found disparities. Thirteen and eight analyzed for race and ethnicity, with disparities found in four (30.8%) and none, respectively. The sample sizes of the four data sets with differences by race were among the largest in the subset. Conclusion: Most studies demonstrate pay disparities for women, but not for people who identify with underrepresented race/ethnic groups; however, small sample sizes may affect results.


Introduction
Among the most important issues confronting the increasingly diverse physician workforce is fair pay. In the United States, both federal law and many states' laws support compensation for people based on the work that is done rather than who is doing the work. 1 Despite this, there is a large body of evidence that demonstrates pay gaps for women in general. A recent systematic review found that across countries and medical specialties, women physicians earned significantly less than men despite similar demographic and workrelated profiles. 2 In this review, the pay disparities were often tens of thousands of dollars less annually, which can translate into millions of dollars in lost income and investments throughout one's career. 3 Less is known about compensation disparities for physicians who identify with racial or ethnic minority groups; however, large surveys such as Medscape 4 suggest that disparities exist for people who identify with these groups. In this report, we analyzed physician compensation studies published in medical journals to determine what is known about pay disparities as they relate to gender, race, and ethnicity.

Methods
We searched PubMed on July 1, 2020, for studies on physician compensation published between January 1, 2013, and June 30, 2020. We included studies if they used terms in the title or abstract: ''salary'' or ''compensation'' or ''wage'' or ''payment'' or ''research support'' and the term(s) ''physician'' or ''faculty.'' We excluded studies that did not include U.S. physician compensation, were not in English, were secondary sources (e.g., reviews, perspectives) that did not present novel data, and studies that focused on Medicare payments only or supplementary income (e.g., industry payments, grant awards). IRB approval was not required as all data collected were publicly available.
In a second round of review, we excluded studies that presented data reported as a percentage (percent funding or percent effort) or a partial component of compensation (not total compensation/salary) or billing metrics (e.g., relative value units).
Two authors (A.R.L. and M.J.E.) independently reviewed the 4,563 articles for inclusion and came to consensus on 62 studies that met the initial criteria. Next, two authors (A.R.L. and Q.R.Y.) independently verified the initial inclusion criteria as well as checked for numeric data on total salary/compensation and came to consensus on 46 studies that met the full inclusion criteria, which included an analysis of 47 data sets in total (one study analyzed two data sets separately). We further evaluated each of the 47 data sets to determine specifics of the analyses and findings of disparities by gender, race, and ethnicity.

Results
Twelve data sets conducted a multivariable analysis considering at least gender and race and, in some instances, ethnicity. Three of these studies included race/ethnicity in their multivariable model, but did not consider the impact of these terms on compensation separately. These 12 are reported in the first section of Table 1. One study considered the impact of gender and race on compensation individually (nonmultivariable model). Three studies analyzed ethnicity and race, but not gender, and, of these, two used multivariable models adjusting for covariates. Twentythree studies collected and analyzed data on gender, but not race or ethnicity. Eight studies did not analyze physician compensation data by gender or race or ethnicity (Table 1 and Fig. 1). Some studies reported gender, race, or ethnicity within general demographic information on participants or used these data as a confounding variable for adjustment in the analysis. Only those studies that reported the impact of each variable on compensation are listed in Table 1 as having analyzed/studied that variable. Table 2 lists the gender, racial, and ethnic breakdown for each study. For studies that reported these categories as percentages, the numbers are noted to be approximate.
Of the 36 data sets for which compensation was analyzed by gender, 32 found gender-based compensation disparities (Table 1 and Fig. 2). In contrast, 13 data sets were analyzed by race and 4 found racebased compensation disparities. For ethnicity, zero out of eight data sets showed differences between ethnic groups (Fig. 2). The median sample size for the data sets that were analyzed by race and ethnicity was 1,012 and 987.5. The four data sets that revealed dif-ferences in compensation by race had four of the five largest sample sizes in the group of 13 at 61,327, 41,396, 17,583, and 2,075.

Discussion
In this report, we found that the majority of data sets on physician compensation focus on women and most of these (88.9%) had documented disparities. A smaller number of data sets considered race and/or ethnicity, and of these, four (30.8%) had documented disparities by race.
Our findings are consistent with other reports on gender-related disparities in compensation for physicians. 2,5 Pay gaps begin early in a physician's career 6 and persist into the highest echelons of academia. 7 Documented disparities exist even after accounting for confounding variables, such as years of experience, academic rank, and specialty, among others (Table 1). [8][9][10][11][12] A recent study showed that research on pay disparities is primarily conducted by women and the majority of this work is unfunded. 13 Some of the institutions that did not show pay disparities in our study (Table 1) were based on a regimented/formulaic model of compensation, 14,15 which is one possible approach to address this problem.
Larger studies have reported compensation disparities based on race and/or ethnicity. A 2019 study by   Medscape on 19,328 U.S. physicians found that Caucasian physicians receive the highest compensation and African American physicians the lowest. 4 This racial disparity persisted after adjusting for specialty. 4 Ly et al. compared the income of White and Black physicians within two large data sets and found that White men made significantly higher compensation than Black men and that, while women physicians made significantly lower income than men physicians, there was no statistically significant difference in compensation for White compared with Black women. 12 This finding was the same for both data sets they considered. 12 Marcelin et al. analyzed unadjusted data from a national society report on compensation and found lower compensation for African American physicians within the society. 16 The authors mentioned that the relatively small number of physicians from underrepresented racial or ethnic groups limited the analysis 16 -a common problem for many compensation studies and databases.
Disparities in rates of promotion can compound compensation disparities. Promotional gaps exist for women physicians. 17 These are present after adjusting for age, experience, research productivity, and other factors. 18 Studies have also shown differences in academic rank based on race. 19 Multivariable compensation studies often adjust for academic rank since compensation is expected to be higher with ascending rank. It is therefore important to consider the additive effect promotional disparities can have on differences in compensation.

Limitations
This study is limited to articles published in the English language and reported in PubMed, as well as by the search terms used to discover these articles.

Conclusion
In conclusion, the majority of reports on physician compensation analyzed for and discovered disparities based on gender. A minority of compensation articles considered disparities based on race and/or ethnicity and this analysis was often limited by a small sample size. Disparities in compensation for racial/ethnic minority groups are understudied and further research is needed.