Association Between Gender and Salary Among Pediatric Hospital Medicine Physicians Free

We conducted a survey-based, cross-sectional study of pediatric hospitalists in December of 2021.

We adapted questions from an existing instrument used to examine pay disparities in neonatology and included new questions pertinent to hospitalists.¹⁴  We asked about compensation models, incentives, administrative supplements, and salary data source. The final survey was reviewed by an expert in measure development to assess question clarity and piloted by 7 physicians. We revised the instrument in response to feedback (Supplemental Information).

We distributed the instrument electronically to a broad group of individuals working as pediatric hospitalists in academic and community settings who responded to a previous survey on the PHM workforce in 2021.⁸  In this previous survey of individuals practicing in PHM, which did not include trainees, all respondents completed 2 sections, and those who indicated that they were group/division leaders completed a third section. The survey distribution list for this current study included members of the American Academy of Pediatrics’ Section on Hospital Medicine Listserv, the Academic Pediatric Association PHM Special Interest Group, and pediatric members of the Society of Hospital Medicine, and was inclusive of individuals who did and did not complete the division leader survey. This allowed us to send the instrument to a discrete group of people known to have working e-mail addresses and to ensure an accurate denominator was available. Although the survey was distributed to individual e-mail addresses, no identifiable information (eg, e-mail addresses, Internet protocol address, etc) was collected, Each e-mail contained a unique link to the instrument tied to the respondent’s e-mail address, which prevented either forwarding of the link or multiple responses for each respondent. We were able to ensure that trainees were not included in this current survey because the initial study from where our distribution list was derived was not sent to trainees. We excluded respondents who did not report either gender or salary. We allowed respondents the opportunity to either select their gender from a list of options or self-describe. We decided a priori to analyze the data from respondents who listed a nonbinary option as their gender identity as a separate group. Similarly, we allowed respondents to either select their race and ethnicity from a list of options or to self-describe.

The e-mail associated with the survey link contained the information necessary for informed consent by the approving institutional review board. Informed consent was presumed by the respondents’ decision to complete the survey. This study was approved by the institutional review board at The American Academy of Pediatrics.

Primary outcomes were base salary (salary without bonuses, supplements, or incentives) and total salary (base salary plus bonuses, supplements, and incentives).

We used a 2-sample Wilcoxon test and χ² test to compare continuous and categorical variables, respectively, between genders. Propensity scores were derived using the covariate balancing propensity score method applied to age, years as a pediatric hospitalist, years with current employer, self-reported race, self-reported ethnicity (for primary analysis), geographic region, urbanicity, employer type, binary presence of a leadership position, marital status, primary wage-earner status, average weekly work hours, and clinical full-time equivalent. We specifically asked about average hours worked per week for 2 main reasons. First, there is significant heterogeneity in how different organizations define full time,¹⁵  and thus what is considered a 1.0 full-time equivalent at 1 institution may vary significantly from another. Second, many nonuniversity-affiliated sites do not use the concept of full-time equivalent. Because a significant portion of the physicians within PHM practice in nonuniversity-affiliated programs, it was important to consider a method to capture the amount of time worked that could be consistent among all respondents.

We compared absolute mean differences before and after propensity score adjustment, which is shown in Supplemental Figs 1–3. After weighting, the absolute mean differences were <0.1, indicating that balance was achieved. After ensuring propensity score balance, we used a weighted t test using inverse probability weighting weights to compare the outcomes between genders.

We performed several subanalyses in this work, including comparing salary between people with and without self-described leadership positions, as well as between respondents who self-described as white and nonwhite. Given the low number of nonwhite respondents, we combined all nonwhite respondents and compared with white respondents. Propensity scores were derived and balanced (as described above) for each subgroup.

We performed a sensitivity analysis using respondents who used a definitive source for salary (eg, W-2, contract).

A total of 567 people responded to the personalized invitation (response rate 51.7%), of which 559 were eligible (51.0%). Our population consisted predominantly of women (71.7%), respondents that self-identified as white (69.8%), and those that were employed by a hospital/health care system (49.1%). Less than 4% of respondents identified as Black, Middle Eastern or North African; Native American or Alaskan Native; or Pacific Island Native. Similarly, our respondents identified mostly as non-Hispanic (93.3%). Respondents who were men were older than women (P < .01), had worked in PHM longer than women (P < .01), and reported working more hours per week than women (P < .01). There was also a higher proportion of men who reported being the primary wage-earner within their families (P < .01). There was no difference in proportion of men and women who had a self-described leadership position (P = .18) (Table 1). The type of self-described leadership positions differed by gender, with a higher proportion of women who reported having a leadership role in medical education (ie, clerkship director, program director, etc) compared with men who reported having a leadership role (43.9% of women versus 28.4% of men, P = .02). However, there was no difference between women and men who reported a leadership role designated as either associate or assistant (eg, associate program director or assistant clerkship director) (13.8% of women versus 13.4% of men, P = .73).

After propensity score weighting, women’s mean base salary was 87.7% of men’s base salary (95% confidence interval [CI] 79.8%–96.4%, P < .01), and women’s total salary was 85.6% of men’s total salary (95% CI 73.2%–100.0%, P = .05). In our subgroup analysis of respondents with a self-described leadership position, women’s base salary was 87.2% of men’s total salary (95% CI 78.6%–96.6%, P < .05) and women’s total salary was 80.6% of men’s total salary (95% CI 68.7%–94.4%, P < .01). There was no difference in either base or total salary between men and women who did not report a leadership position. (Table 2). Although women who identified as white had base salaries that were 86.6% of white men’s base salary (95% CI 78.5%–95.5%, P < .01), there was no gender-based difference noted between respondents that identified as nonwhite (women made 88.4% [69.9%–111.7%] of men’s base salary and 80.3% [57.2%–112.7%] of men’s total salary) (Table 2).

Our sensitivity analysis of 322 respondents who used a validated method to report salary (ie, offer letter, W-2, current contract, etc) found lower magnitudes of difference, but similar trends as the primary analysis: In the sensitivity analysis, the percentage of men’s salaries earned by women for base and total salaries was 93.0% (86.6%–99.8%, P < .05) and 91.5% (84.2%–99.5%, P < .05), respectively. Comparatively, in the full cohort, the percentage of men’s salaries earned by women for base and total salaries was 87.7% (79.8%–96.4%, P < .05) and 85.6% (73.2%–100.0%, P = .05), respectively. One notable difference in the sensitivity analysis was seen in the subgroup of nonwhite respondents who used a validated source of their salary: In this subgroup, there was a significant difference between men and women for base and total salary (Table 3).

Similar to previous studies demonstrating gender disparities in PHM,^16,17  we report gender-based salary differences among pediatric hospitalists. We also report that the magnitude of differences in total salary are greater in respondents with a self-described leadership position. Finally, we did not report a difference in salary between genders among nonwhite respondents.

The field of PHM is estimated to be 70% women,¹⁶  which is consistent with the gender-based distribution of our survey respondents. However, despite the predominance of women in PHM, there have been gender-based differences in several areas, including leadership,¹⁷  senior authorship on articles, opportunities as invited speakers at national conferences,¹⁶  and service on journal editorial boards. Although there are no previous PHM-specific studies examining salary, there are multiple studies in the literature that demonstrate differences in financial compensation between genders.¹⁸  In one study that examined physician salary in public medical schools, there were differences in salary between genders in all specialties, with the exception of obstetrics/gynecology and radiology. This same study, which did not account for leadership positions, found that discrepancies between men and women worsened at higher ranks, with the largest difference between women and men who were full professors. More recently, an unadjusted analysis using data from the Association of American Medical Colleges reported a gender-based difference in starting salaries in PHM.⁷  This study only included full-time faculty employed by university-affiliated institutions, so although our results are not directly comparable, our findings add to our understanding of the extent of pay disparities between men and women in PHM by including physicians who work in a nonuniversity-affiliated setting, as well as those that work less than full time.

Some medical subspecialities have demonstrated apparent decreased differences in salary between genders over time,²  potentially reflecting efforts to mitigate these gaps. Therefore, because PHM has significantly grown in the past few years, it is possible that new physicians in the field were hired in an environment where measures are in place to avoid disparities in salary. Other explanations for the pay gap have included gender variations in personal work-related decisions, like the transition to part-time work, because women are more likely than men to work part time.⁴  Our findings suggest an alternative explanation is more likely, because pay discrepancies persist even when compensation was adjusted per hours worked per week. One of our notable findings is that increased salary disparities are seen in those respondents with leadership roles, and that there is a lack of salary disparities among those without leadership roles, suggesting that compensation for leadership roles may be a significant driver of salary disparities in this group. Thus, on the basis of our findings, examining and addressing compensation for various leadership roles may be a place where institutions can begin to make substantive changes in addressing salary disparities.

It is important to note that our instrument did not include a definition of leadership position, but rather simply asked if the respondent had a leadership position. Therefore, there may be differences in what men and women considered to be leadership roles, or the possibility that women more often hold uncompensated leadership roles. This would align with the evidence related to “invisible work,” a construct referring to work that, although critical, is traditionally not formally accounted for and is therefore undervalued with regard to promotion, tenure, or supplementary compensation.¹⁹  Additionally, we did note that women had a higher proportion of leadership roles in education. This may partially explain the discrepancy because these roles are often compensated at a lower rate than other administrative positions.^20,21 

One suggested mitigation strategy is salary transparency. However, anecdotally, individuals are reluctant to share salary data, and finances are often treated as a confidential matter. The robust response rate in our electronically distributed study suggests that, nationally, individuals may be willing to share salary data if confidentiality is assured. Sharing salary data, as was done in this study, on a national scale is an intriguing, if fraught, possibility as a strategy to address the pay gap and provide data to facilitate negotiation. Institutions can also improve hiring and promotion processes to include transparent information about compensation related to academic rank and leadership position.

In our study, we chose to normalize salary by hours worked per week, because what is considered to be a full-time equivalent, or whether that concept is even used, varies between institutions.¹⁵  Normalizing salary by hours worked per work is imperfect because it requires respondents to estimate hours worked per week. This approach effectively assigned equivalent value to time spent in a range of activities, from clinical to administrative, without specific weighting given to tasks conducted during less preferable times (eg, weekends, evenings) as is common practice at the authors’ institutions. As such, our results may not have accurately captured respondents who were part time as compared with respondents who underestimated their hours worked per week. In spite of these limitations, this approach enabled inclusion of members of the PHM community that work within a nonuniversity-affiliated program, and thus our findings are more representative of the field of PHM as a whole and could allow for tracking of compensation across different compensation structures.

In this work, we examined comparisons between respondents who self-identified as white and nonwhite. However, we acknowledge that many personal characteristics may result in individuals experiencing discrimination and resultant pay disparities including national origin, sexual orientation, disability, presence of an accent when speaking English, and colorism (eg, preference for lighter skin tones), which our survey did not address. Our results add urgency and support the need for more detailed studies including quantitative and qualitative assessments focusing on and partnering with individuals in those communities to identify how they are negatively impacted by pay and other types of discrimination. Our results, which include a low number of nonwhite respondents, also highlight how critical it is to address the low numbers of individuals from diverse backgrounds entering medicine, as well as the potential for lower engagement from those individuals in national professional groups which may drive policy and practices nationally.

Our study also has limitations related to the complexities of comparing compensation between different models. Our instrument (Supplemental Information) explicitly described the concepts of base pay and supplementary income; however, it is possible that respondents may have had difficulty interpreting their own financial information. We would anticipate that those difficulties would be randomly distributed among respondents, and so the likelihood that this systematically influenced results in one direction or another seems less likely. Additionally, we performed a sensitivity analysis of respondents who used a validated source for their financial information, which showed similar trends as our primary analysis. However, the variation in interpretation of this response may affect our results. Our sampling frame is another limitation of this work. Although it allowed us to survey a cohort of people known to have a working e-mail address and enabled the inclusion of a denominator (and thus calculation of response rate), this sampling frame may not have been representative of the entire PHM community. Indeed, our study population included members of the PHM community who were members of either the American Academy of Pediatrics, the Academic Pediatric Association, or the Society of Hospital Medicine. Thus, this survey may not reflect those pediatric hospitalists who are not in 1 of these 3 societies, and may represent a particularly engaged and academically active group. As a consequence, we may have oversampled individuals with research or operational roles, for example, and underrepresented individuals working primarily clinical roles. Our respondents do, however, represent the estimated gender breakdown that exists within the field at large, and respondents reported working in a variety of settings.

There are also some limitations related to our subgroups. First, we did not detect a difference in salary between men and women who did not report a leadership position, which may be because of a lower number of respondents in this category. In addition, we have low numbers of respondents who identify as nonwhite. Although the percentage of respondents who identified as white was quite high in our sample (70%), it is similar to the racial breakdown of all medical school faculty, in which 63.9% of faculty members identify as white,²²  suggesting that our sample is representative of the larger field of medicine. We also acknowledge that grouping all nonwhite respondents does not reflect the experiences of all included groups, and that people of different races likely experience different degrees of racism that is not reflected in our results. This low number of nonwhite respondents is likely also responsible for the wide CI seen when we examined salary differences for nonwhite respondents. This wide CI, and nonspecific grouping of nonwhite respondents, may be responsible for not reporting a significant difference when one, in fact, does exist because of inadequate sample size.²³  Indeed, other studies in the literature have shown greater differences in compensation for women who are also underrepresented in medicine. However, given the limited utility of post-hoc power analysis in determining optimal sample size,²⁴  we have chosen not to perform one.

In conclusion, in a national sample of pediatric hospitalists, men had higher base salaries than women. Although these differences were most pronounced among individuals who held leadership positions, there was no gender-based disparity between women and men without a leadership role. Our data suggest that, although pay equity may have been achieved within at least a subset of the field, significant differences continue to exist and require institutional-level changes to continue to work toward salary parity. Future directions include confirmation of these results in a larger cohort of pediatric hospitalists, which will allow for further examination of subgroups, specifically with regard to women with intersectional identities that we are likely underpowered to detect.