Abstract
Purpose:
To determine factors that contribute to current salary differences among ophthalmologists employed by the Veterans Health Administration (VHA).
Methods:
We performed a cross-sectional analysis of full-time equivalent salaries of Veterans Health Administration ophthalmologists who received a federal salary in 2021. Differences in average salary with respect to various demographic variables were analyzed.
Results:
Data for 632 salaried VHA ophthalmologists (262 females [41.5 %]) were included. Although biological sex alone did not have a crude, statistically significant effect on average VHA ophthalmologist salary (p = 0.116), a substantial discrepancy was noted in adjusting for other significant demographic variables, including years spent at the VHA. There was a significant interaction effect between sex and years at the VHA: female ophthalmologists initially earned substantially and significantly less than their male counterparts (−$9726 [p = 0.018]). Regarding other variables, chief status and paygrade had the greatest effect on median overall salary, with an average increase of $33,673 for chiefs (p = 5.52e-11) and $28,472 for full-time VHA physicians (p = 1.96e-23). Years since residency (p = 3e-4) and h-index (p = 0.03) were each positively and significantly correlated with overall salary. All Veterans Integrated Services Network (VISN) regions, especially the Northeast region, were associated with significantly lower salaries compared with the Southern region.
Conclusion:
Female sex is associated with a significantly lower initial compensation for ophthalmologists employed by the VHA when adjusting for chief status, paygrade, VISN region, complexity level, academic rank, and years at the VHA. Although several factors could be involved in the apparent salary disparities found within this study, policy changes may be important to ensure compensation is equitable across VHA centers.
Keywords: Compensation, Pay gap, Veterans Health Administration, Salary, Sex
Introduction
Data published by the United States Bureau of Labor Statistics in 2022 showed that women with an advanced degree earned $23,504 less in median annual earnings than their male counterparts with equal degrees.1 Multiple groups have shown that this gender pay gap also applies to physicians working in academic or private practice settings and in both medical and surgical sub-specialities, even after adjusting for factors such as faculty rank, age, years since residency, hours worked, and research productivity.2–5 A recent study by Emami-Naeini et al.6 suggests that the gap between female and male academic ophthalmologists is even larger than that reported by the U. S. Bureau of Labor Statistics, with women being paid an average of $50,300 less than their male counterparts.
In contrast, several groups have reported on the lack of pay gap among male and female physicians working in the U.S. Veterans Health Administration (VHA), within the Department of Veteran Affairs, and credit the objective criteria used to establish physician salaries as well as the transparency of VHA salary data as the cause.7–9 Maxwell et al.9 demonstrated that a gender pay gap was nonexistent in the year 2016 for a number of surgical sub-specialities in the VHA system, ophthalmology included. However, the authors did not compare salaries among women and men in varying ophthalmic subspecialties; nor did they analyze how VHA complexity level, paygrade, and chief status affect compensation.
The purpose of the current study was to identify how biological sex (“sex”) and other variables impact salary among VHA ophthalmologists. We created the largest salary database for Department of Veterans Affairs—employed ophthalmologists―more than twice the size of previous VHA ophthalmologist salary studies―across a variety of subspecialties that included several previously unexplored demographic and scholarly parameters.9 We hypothesized that sex and geographic location would not have a significant effect on salary but that ophthalmic subspecialty, academic rank, years since residency, h-index, chief status, and paygrade would. We analyzed each of these variables in the context of accounting for and disregarding sex as a factor. This data may aid in career decision making for prospective VHA ophthalmologists and VHA administrators, human resource teams, and leadership. Additionally, because of the importance of the VHA on resident education, it is important to be able to recruit and maintain a diverse population of ophthalmologists across the United States based on the assurance of equitable pay.
Methods
Data acquisition
The methodology is based on that employed by previously published VHA equity studies.7–9 Ophthalmologists employed by the VHA were identified using the VHA’s publicly available “Our Providers” online directory (https://www.accesstocare.va.gov/FindProviders), which was used to compile ophthalmologist name (later deidentified), sex, VHA parent facility, VHA paygrade, and VHA service line.
The “Federal Employee Lookup” tool on FederalPay.org was used to collect full-time equivalent (FTE) salary data in U.S. dollars (USD) for the year 2021 as well as paygrade and first year employed by the VHA. Physicians whose salaries were not included in this database were excluded from our analysis. Each individual’s Hirsch index (h-index) was collected via Scopus author profile search. In the case of multiple entries for the same physician, the highest h-index was used.
Information regarding ophthalmic subspecialty, completion of a fellowship, academic rank at an affiliated university, years since residency, and VHA chief status was compiled using Doximity, an online networking service for physicians, as well as university websites, and LinkedIn. For additional information regarding descriptions of complexity level, VISNs, paygrade, service lines, and h-index, see Supplemental Methods (https://www.aupojournal.org/cgi/viewcontent.cgi?filename=0&article=1310&context=jao&type=additional&preview_mode=1).10–12
Our final dataset comprised 14 variables: one outcome (salary in USD) and thirteen covariates. For definitions of VHA terminology, refer to Supplemental Methods (https://www.aupojournal.org/cgi/viewcontent.cgi?filename=0&article=1310&context=jao&type=additional&preview_mode=1). Three of the covariates were continuous (years since residency, years at the VHA, h-index) and 10 were categorical. Five of the categorical variables were treated as binary: sex (male/female), paygrade (part-time/full-time), service line (eye/surgery), fellowship status (yes/no), and chief status (yes/no). Five of the categorical variables were treated as nominal: state (the 48 contiguous states, as well as District of Columbia and Puerto Rico), subspecialty (comprehensive, cornea, glaucoma, medical retina, neuro-ophthalmology, ocular oncology/pathology, oculoplastics, pediatrics/adult strabismus, surgical retina, and uveitis), VISN region (Central, Northeast, Southern, Western), complexity level (1a, 1b, 1c, 2, 3), and academic rank (none, instructor, assistant professor, associate professor, adjunct professor, professor, chair).
Statistical analyses
All data analysis was performed using R version 4.4.0 in R Studio. Outliers were removed if the reported salary was greater or less than 1.5 times the interquartile range (IQR). Missing data was handled using complete case analysis, assuming the missing data for each variable appeared to be missing completely at random, and imputation would not offer considerable benefit.
Since many states had very few ophthalmologists, those with fewer than 10 reported ophthalmologists were aggregated into “other.” Each of the continuous variables were also divided into quartiles of roughly equal size to enhance data exploration and reporting. Any regression models used the continuous version of these variables.
Differences in distribution of variables by sex were analyzed using Pearson’s χ2 test (categorical variables when all expected values ≥5), the Fisher exact test (categorical variables when any of the expected values <5), or a Wilcoxon rank-sum test (continuous variables); differences in salary for all variables by sex were analyzed using two-sided t tests. Individual relationships between variables and physician salary were analyzed using simple linear regression. Interaction effects of each variable with sex were also evaluated using multiple linear regression. To compare salary between each category for multicategorical variables, pairwise two-sided t tests were performed. Significance for all tests were determined using an α value of 0.05, and a Bonferroni adjustment was applied when multiple tests were performed. A final multivariate regression model that considered all significant variables from the univariate and multivariate interaction effect models was created using bidirectional stepwise variable selection.
Results
Of the 940 ophthalmologists listed on the VHA’s “Our Provider’s” website, 297 (32 %) did not have salary information listed on FederalPay.org, and 8 with salary data were excluded because they were outliers. An additional 3 were excluded because they had an unclear paygrade label of “SR-PH.” The remaining 632 VHA ophthalmologists (370 males [59 %]) were included in our analysis. There were three variables with missing values, meaning the data could not be found or obtained: 6.8 % of the years since residency variable, 3.3 % of the service line variable, and 1.9 % of the subspecialty variable. The distribution of missing values for each variable is shown in Figure S1 and Table S1 (https://www.aupojournal.org/cgi/viewcontent.cgi?filename=0&article=1310&context=jao&type=additional&preview_mode=1).
Variable distributions
Summary statistics for all thirteen variables are included in Table 1, with and without sex stratification. No statistically significant differences were noticed in the distributions of any categorical variables by sex. For three variables―academic rank, subspecialty, and state―statistical tests were not appropriate because of the large number of categories and limited sample size per category. The summary statistics for these variables are reported for informational purposes.
Table 1.
Demographicsa.
| Variable | Overall N = 632 | Female n = 262 | Male n = 370 | P valueb |
|---|---|---|---|---|
| Chief status | 48 (7.6 %) | 25 (9.5 %) | 23 (6.2 %) | 0.12 |
| Pay grade | 0.9 | |||
| Part-time | 430 (68 %) | 179 (68 %) | 251 (68 %) | |
| Full-time | 202 (32 %) | 83 (32 %) | 119 (32 %) | |
| Service line | 0.7 | |||
| Eye | 295 (48 %) | 119 (47 %) | 176 (49 %) | |
| Surgery | 316 (52 %) | 132 (53 %) | 184 (51 %) | |
| Unknown | 21 | 11 | 10 | |
| Fellowship | 334 (53 %) | 137 (52 %) | 197 (53 %) | 0.8 |
| VISN region | 0.3 | |||
| Central | 144 (23 %) | 59 (23 %) | 85 (23 %) | |
| Northeast | 107 (17 %) | 37 (14 %) | 70 (19 %) | |
| Southern | 249 (39 %) | 104 (40 %) | 145 (39 %) | |
| Western | 132 (21 %) | 62 (24 %) | 70 (19 %) | |
| Complexity level | 0.2 | |||
| 1a | 360 (57 %) | 156 (60 %) | 204 (55 %) | |
| 1b | 122 (19 %) | 49 (19 %) | 73 (20 %) | |
| 1c | 100 (16 %) | 44 (17 %) | 56 (15 %) | |
| 2 | 40 (6.3 %) | 10 (3.8 %) | 30 (8.1 %) | |
| 3 | 10 (1.6 %) | 3 (1.1 %) | 7 (1.9 %) | |
| Years since residency | 16 (9, 26) | 13 (8, 20) | 20 (10, 30) | <0.001 |
| Years since residency (quartiles) | <0.001 | |||
| 0–9 years | 150 (25 %) | 73 (30 %) | 77 (22 %) | |
| 10–16 years | 157 (27 %) | 86 (35 %) | 71 (21 %) | |
| 17–26 years | 140 (24 %) | 57 (23 %) | 83 (24 %) | |
| 27+ years | 142 (24 %) | 29 (12 %) | 113 (33 %) | |
| Unknown | 43 | 17 | 26 | |
| Years at VHA | 7.0 (2.0, 14.0) | 7.0 (3.0, 14.0) | 7.0 (2.0, 14.0) | 0.6 |
| Years at VHA (quartiles) | 0.3 | |||
| 0–2 years | 161 (25 %) | 62 (24 %) | 99 (27 %) | |
| 3–7 years | 172 (27 %) | 80 (31 %) | 92 (25 %) | |
| 8–14 years | 153 (24 %) | 66 (25 %) | 87 (24 %) | |
| 15+ years | 146 (23 %) | 54 (21 %) | 92 (25 %) | |
| H-index | 3 (1, 8) | 2 (0, 6) | 3 (1, 10) | 0.030 |
| H-index (quartiles) | 0.025 | |||
| 0 | 158 (25 %) | 67 (26 %) | 91 (25 %) | |
| 1–2 | 155 (25 %) | 73 (28 %) | 82 (22 %) | |
| 3–7 | 150 (24 %) | 68 (26 %) | 82 (22 %) | |
| 8+ | 169 (27 %) | 54 (21 %) | 115 (31 %) | |
| Subspecialty | ||||
| Comprehensive | 230 (37 %) | 96 (37 %) | 134 (37 %) | |
| Cornea | 78 (13 %) | 30 (12 %) | 48 (13 %) | |
| Glaucoma | 97 (16 %) | 52 (20 %) | 45 (12 %) | |
| Medical retina | 26 (4.2 %) | 15 (5.8 %) | 11 (3.0 %) | |
| Neuro-ophth | 31 (5.0 %) | 13 (5.0 %) | 18 (5.0 %) | |
| Ocular oncology/pathology | 5 (0.8 %) | 3 (1.2 %) | 2 (0.6 %) | |
| Oculoplastics | 52 (8.4 %) | 20 (7.7 %) | 32 (8.9 %) | |
| Peds/adult strab | 14 (2.3 %) | 8 (3.1 %) | 6 (1.7 %) | |
| Surgical retina | 72 (12 %) | 14 (5.4 %) | 58 (16 %) | |
| Uveitis | 15 (2.4 %) | 8 (3.1 %) | 7 (1.9 %) | |
| Unknown | 12 | 3 | 9 | |
| Academic rank | ||||
| None | 221 (35 %) | 94 (36 %) | 127 (34 %) | |
| Instructor | 16 (2.5 %) | 6 (2.3 %) | 10 (2.7 %) | |
| Assistant professor | 174 (28 %) | 78 (30 %) | 96 (26 %) | |
| Associate professor | 111 (18 %) | 54 (21 %) | 57 (15 %) | |
| Adjunct professor | 25 (4.0 %) | 7 (2.7 %) | 18 (4.9 %) | |
| Professor | 81 (13 %) | 22 (8.4 %) | 59 (16 %) | |
| Chair | 4 (0.6 %) | 1 (0.4 %) | 3 (0.8 %) | |
| State | ||||
| CA | 68 (11 %) | 32 (12 %) | 36 (9.7 %) | |
| DC | 10 (1.6 %) | 5 (1.9 %) | 5 (1.4 %) | |
| FL | 42 (6.6 %) | 24 (9.2 %) | 18 (4.9 %) | |
| GA | 17 (2.7 %) | 6 (2.3 %) | 11 (3.0 %) | |
| IA | 11 (1.7 %) | 2 (0.8 %) | 9 (2.4 %) | |
| IL | 45 (7.1 %) | 19 (7.3 %) | 26 (7.0 %) | |
| LA | 12 (1.9 %) | 5 (1.9 %) | 7 (1.9 %) | |
| MI | 13 (2.1 %) | 9 (3.4 %) | 4 (1.1 %) | |
| MO | 12 (1.9 %) | 2 (0.8 %) | 10 (2.7 %) | |
| NC | 25 (4.0 %) | 15 (5.7 %) | 10 (2.7 %) | |
| NE | 10 (1.6 %) | 6 (2.3 %) | 4 (1.1 %) | |
| NY | 49 (7.8 %) | 21 (8.0 %) | 28 (7.6 %) | |
| OH | 14 (2.2 %) | 4 (1.5 %) | 10 (2.7 %) | |
| OR | 11 (1.7 %) | 7 (2.7 %) | 4 (1.1 %) | |
| PA | 24 (3.8 %) | 8 (3.1 %) | 16 (4.3 %) | |
| RI | 11 (1.7 %) | 2 (0.8 %) | 9 (2.4 %) | |
| SC | 17 (2.7 %) | 4 (1.5 %) | 13 (3.5 %) | |
| TN | 21 (3.3 %) | 11 (4.2 %) | 10 (2.7 %) | |
| TX | 48 (7.6 %) | 21 (8.0 %) | 27 (7.3 %) | |
| UT | 22 (3.5 %) | 9 (3.4 %) | 13 (3.5 %) | |
| VA | 10 (1.6 %) | 3 (1.1 %) | 7 (1.9 %) | |
| WI | 20 (3.2 %) | 6 (2.3 %) | 14 (3.8 %) | |
| Other | 120 (19 %) | 41 (16 %) | 79 (21 %) |
Summary statistics for ophthalmologists employed by the Veterans Health Administration, overall and stratified by sex. The four variables shown are continuous, displaying the median (interquartile range), and the rest are categorical, showing number (%). The four continuous variables are divided into quartiles of roughly equal size for exploratory purposes.
Pearson’s χ2 test, Fisher exact test, and Wilcoxon rank-sum test.
Regarding the continuous variables, although there was no significant difference for the number of years at the VHA by sex (p = 0.6), we observed a significant difference in number of years since residency, with a median of 13 years for females and 20 years for males (p < 0.001). The “27+ years” quartile was the largest contributor, with 113 of the male physicians (33 %) compared with 29 of the female physicians (12 %). There was a small but significant difference in h-index, with a median of 2 for females and 3 for males (p = 0.03). Similarly, the highest “8+” quartile is the greatest contributor, with 115 of the males (31 %) compared with 54 of the females (21 %).
Salary information
Table 2 provides the average salaries for each subgroup of the fourteen variables with and without sex stratification (see Figure S2) (https://www.aupojournal.org/cgi/viewcontent.cgi?filename=0&article=1310&context=jao&type=additional&preview_mode=1). The average salary for all 632 ophthalmologists was $291,866 ± $34,750. Stratified by sex, the average was $293,694 ± $35,191 for males and $289,284 ± $34,018 for females (p = 0.12). While this crude difference of males making $4410 more is not statistically significant, it aligns with a slight trend toward higher male salaries observed in most variable subgroups. Figure S3 (https://www.aupojournal.org/cgi/viewcontent.cgi?filename=0&article=1310&context=jao&type=additional&preview_mode=1) shows regional differences in salary.
Table 2.
Mean and standard deviation of average salary for total, female, and male ophthalmologists employed by the VHAa.
| Variable | Salary | Diff | P value | Adj p valueb | ||
|---|---|---|---|---|---|---|
| Overall | Female | Male | ||||
| Overall (N = 632) | 291,866 (34,750) | 289,284 (34,018) | 293,694 (35,191) | 4410 | 0.116 | 0.116 |
| Chief status | ||||||
| No (n = 584) | 289,309 (33,958) | 286,081 (32,823) | 291,513 (34,586) | 5432 | 0.058 | 0.115 |
| Yes (n = 48) | 322,982 (28,953) | 319,652 (30,474) | 326,601 (27,414) | 6949 | 0.412 | 0.824 |
| Pay grade | ||||||
| Part-time (n = 430) | 282,766 (33,573) | 280,777 (33,856) | 284,184 (33,366) | 3407 | 0.300 | 0.600 |
| Full-time (n = 202) | 311,238 (28,826) | 307,631 (26,392) | 313,753 (30,263) | 6122 | 0.138 | 0.276 |
| Service line | ||||||
| Eye (n = 295) | 289,869 (34,407) | 288,520 (30,545) | 290,781 (36,846) | 2261 | 0.581 | 1.000 |
| Surgery (n = 316) | 293,561 (34,316) | 289,987 (36,100) | 296,125 (32,837) | 6138 | 0.117 | 0.234 |
| Fellowship | ||||||
| No (n = 298) | 292,855 (35,717) | 289,502 (36,218) | 295,278 (35,257) | 5777 | 0.169 | 0.337 |
| Yes (n = 334) | 290,983 (33,894) | 289,086 (32,012) | 292,303 (35,164) | 3217 | 0.394 | 0.789 |
| VISN region | ||||||
| Central (n = 144) | 285,423 (34,612) | 277,851 (36,821) | 290,679 (32,173) | 12,828 | 0.028 | 0.113 |
| Northeast (n = 107) | 271,812 (39,710) | 260,218 (30,366) | 277,941 (42,792) | 17,723 | 0.027 | 0.110 |
| Southern (n = 249) | 303,571 (31,318) | 302,562 (29,853) | 304,294 (32,412) | 1732 | 0.668 | 1.000 |
| Western (n = 132) | 293,071 (27,207) | 295,238 (25,255) | 291,152 (28,871) | −4085 | 0.391 | 1.000 |
| Complexity level | ||||||
| 1a (n = 360) | 291,954 (33,352) | 290,431 (34,032) | 293,118 (32,858) | 2688 | 0.449 | 1.000 |
| 1b (n = 122) | 292,638 (34,805) | 289,219 (33,147) | 294,933 (35,916) | 5714 | 0.376 | 1.000 |
| 1c (n = 100) | 281,813 (38,595) | 279,334 (35,179) | 283,761 (41,291) | 4427 | 0.572 | 1.000 |
| 2 (n = 40) | 308,675 (30,150) | 310,054 (26,707) | 308,215 (31,627) | −1840 | 0.870 | 1.000 |
| 3 (n = 10) | 312,589 (28,653) | 307,447 (11,837) | 314,793 (34,146) | 7345 | 0.733 | 1.000 |
| Years since residency | ||||||
| 0–9 years (n = 150) | 281,591 (30,627) | 281,891 (31,492) | 281,307 (29,989) | −584 | 0.908 | 1.000 |
| 10–16 years (n = 157) | 290,844 (34,300) | 288,416 (34,254) | 293,784 (34,368) | 5367 | 0.331 | 1.000 |
| 17–26 years (n = 140) | 299,244 (35,939) | 297,589 (35,757) | 300,381 (36,237) | 2791 | 0.653 | 1.000 |
| 27+ years (n = 142) | 296,328 (35,069) | 293,933 (30,960) | 296,942 (36,149) | 3009 | 0.682 | 1.000 |
| Years at VHA | ||||||
| 0–2 years (n = 161) | 291,683 (33,871) | 283,522 (33,926) | 296,793 (32,987) | 13,271 | 0.015 | 0.060 |
| 3–7 years (n = 172) | 289,468 (31,445) | 284,444 (30,716) | 293,837 (31,584) | 9393 | 0.050 | 0.202 |
| 8–14 years (n = 153) | 293,523 (36,619) | 293,622 (36,557) | 293,447 (36,877) | −175 | 0.977 | 1.000 |
| 15+ years (n = 146) | 293,157 (37,497) | 297,768 (34,027) | 290,451 (39,318) | −7318 | 0.256 | 1.000 |
| H-index | ||||||
| 0 (n = 158) | 293,018 (35,744) | 288,751 (34,393) | 296,160 (36,576) | 7410 | 0.199 | 0.795 |
| 1–2 (n = 155) | 288,811 (33,363) | 289,578 (32,029) | 288,128 (34,690) | −1451 | 0.788 | 1.000 |
| 3–7 (n = 150) | 286,825 (31,614) | 282,581 (32,194) | 290,344 (30,878) | 7763 | 0.135 | 0.539 |
| 8+ (n = 169) | 298,066 (36,928) | 297,991 (37,237) | 298,101 (36,946) | 110 | 0.986 | 1.000 |
| Subspecialty | ||||||
| Comprehensive (n = 230) | 293,301 (34,792) | 288,776 (33,869) | 296,543 (35,206) | 7767 | 0.095 | 0.951 |
| Cornea (n = 78) | 295,508 (32,858) | 294,127 (34,690) | 296,371 (32,004) | 2244 | 0.771 | 1.000 |
| Glaucoma (n = 97) | 288,294 (31,055) | 287,806 (29,318) | 288,858 (33,276) | 1052 | 0.869 | 1.000 |
| Medical retina (n = 26) | 299,789 (39,231) | 295,529 (32,860) | 305,599 (47,661) | 10,070 | 0.529 | 1.000 |
| Neuro-ophth (n = 31) | 274,982 (35,219) | 262,372 (32,891) | 284,089 (34,870) | 21,716 | 0.090 | 0.904 |
| Ocular oncology/pathology (n = 5) | 311,460 (24,520) | 319,727 (6735) | 299,061 (42,446) | −20,666 | 0.434 | 1.000 |
| Oculoplastics (n = 52) | 281,965 (37,368) | 282,266 (36,182) | 281,776 (38,663) | −490 | 0.964 | 1.000 |
| Peds/adult strab (n = 14) | 281,290 (38,450) | 285,841 (47,307) | 275,222 (25,165) | −10,618 | 0.629 | 1.000 |
| Surgical retina (n = 72) | 299,719 (34,379) | 311,176 (37,371) | 296,954 (33,371) | −14,222 | 0.166 | 1.000 |
| uveitis (n = 15) | 295,800 (29,915) | 290,936 (28,235) | 301,359 (33,021) | 10,423 | 0.521 | 1.000 |
| Academic rank | ||||||
| None (n = 221) | 292,043 (32,053) | 286,623 (31,036) | 296,055 (32,324) | 9432 | 0.030 | 0.212 |
| Instructor (n = 16) | 280,450 (26,956) | 280,372 (25,275) | 280,497 (29,260) | 125 | 0.993 | 1.000 |
| Assistant prof (n = 174) | 284,912 (37,398) | 280,900 (36,109) | 288,172 (38,291) | 7272 | 0.203 | 1.000 |
| Associate prof (n = 111) | 293,865 (32,883) | 297,096 (32,119) | 290,803 (33,586) | −6293 | 0.316 | 1.000 |
| Adjunct prof (n = 25) | 300,104 (38,349) | 304,829 (31,086) | 298,266 (41,501) | −6564 | 0.709 | 1.000 |
| Professor (n = 81) | 302,820 (35,235) | 305,615 (35,412) | 301,777 (35,416) | −3838 | 0.666 | 1.000 |
| Chair (n = 4) | 301,458 (47,616) | 356,921 (NA) | 282,971 (36,744) | −73,950 | NA | NA |
| State | ||||||
| CA (n = 68) | 289,238 (29,260) | 294,994 (23,446) | 284,122 (33,088) | −10,872 | 0.127 | 1.000 |
| DC (n = 10) | 295,908 (23,627) | 298,268 (34,400) | 293,548 (7670) | −4719 | 0.772 | 1.000 |
| FL (n = 42) | 294,351 (27,820) | 294,416 (23,551) | 294,265 (33,408) | −151 | 0.986 | 1.000 |
| GA (n = 17) | 319,058 (32,346) | 319,825 (34,609) | 318,640 (32,781) | −1186 | 0.945 | 1.000 |
| IA (n = 11) | 320,121 (24,285) | 301,620 (46,313) | 324,233 (19,092) | 22,613 | 0.254 | 1.000 |
| IA (n = 11) | 272,173 (25,790) | 273,239 (25,549) | 271,393 (26,441) | −1846 | 0.816 | 1.000 |
| LA (n = 12) | 296,251 (17,389) | 301,183 (21,010) | 292,729 (15,012) | −8454 | 0.433 | 1.000 |
| MI (n = 13) | 236,276 (21,628) | 227,035 (12,182) | 257,069 (25,351) | 30,034 | 0.013 | 0.293 |
| MO (n = 12) | 307,597 (27,778) | 309,846 (13,395) | 307,147 (30,361) | −2699 | 0.907 | 1.000 |
| NC (n = 25) | 313,179 (25,293) | 315,183 (29,347) | 310,174 (18,694) | −5008 | 0.638 | 1.000 |
| NE (n = 10) | 276,223 (27,616) | 263,850 (27,480) | 294,782 (16,242) | 30,933 | 0.080 | 1.000 |
| NY (n = 49) | 249,624 (29,747) | 249,791 (26,562) | 249,499 (32,410) | −291 | 0.973 | 1.000 |
| OH (n = 14) | 299,764 (33,543) | 289,257 (29,162) | 303,968 (35,679) | 14,711 | 0.481 | 1.000 |
| OR (n = 11) | 311,123 (28,283) | 302,943 (27,128) | 325,438 (27,662) | 22,494 | 0.221 | 1.000 |
| PA (n = 24) | 279,124 (38,573) | 267,139 (30,406) | 285,117 (41,654) | 17,977 | 0.292 | 1.000 |
| RI (n = 11) | 290,943 (14,375) | 280,763 (23,562) | 293,206 (12,539) | 12,443 | 0.291 | 1.000 |
| SC (n = 17) | 296,819 (35,212) | 287,942 (21,219) | 299,550 (38,810) | 11,608 | 0.581 | 1.000 |
| TN (n = 21) | 312,659 (32,255) | 318,842 (37,369) | 305,858 (25,733) | −12,984 | 0.370 | 1.000 |
| TX (n = 48) | 304,066 (37,351) | 293,000 (28,850) | 312,672 (41,301) | 19,672 | 0.070 | 1.000 |
| UT (n = 22) | 290,320 (24,226) | 285,164 (34,422) | 293,889 (14,263) | 8725 | 0.420 | 1.000 |
| VA (n = 10) | 312,843 (26,773) | 294,049 (13,492) | 320,898 (27,608) | 26,849 | 0.156 | 1.000 |
| WI (n = 20) | 289,357 (22,036) | 288,131 (22,124) | 289,882 (22,814) | 1752 | 0.876 | 1.000 |
| Other (n = 120) | 300,894 (31,297) | 302,208 (30,799) | 300,211 (31,726) | −1997 | 0.742 | 1.000 |
NA values indicate that there were not enough observations to calculate a mean, standard deviation, or p value.
Bonferroni corrected p value, calculated by multiplying the p value by the number of categories for that variable (corrected p values are capped at 1.000).
Univariate regression
Table 3 shows the results of univariate simple linear regressions for each of the twelve variables. Statistically significant unadjusted effects on salary were observed in several categories: chief status (+$33,673 [p = 5.5e-11]), full-time paygrade (+$28,472, p = 2.0e-23), and Central and Northeast VISN regions compared to the Southern reference (Central, −$18,148 [p = 1.8e-7]; Northeast, −$31,759 [p = 4.1e-16]) were associated with the most significant effects. To a lesser degree, the Western VISN region also had a significant effect (−$10,500 [p = 0.003]). Additional significant effects included the following: complexity levels 1c (−$10,141 [p = 0.009]) and 2 (+$16,721 [p = 0.004]) compared with the reference 1a; years since residency (+$478 for each additional year [p = 2.6e-4]); h-index (+$310 for each additional point [p = 0.03]); neuro-ophthalmology (−$18,320 [p = 0.005]) and oculoplastics (−$11,336 [p = 0.03]) compared with the reference comprehensive; assistant professor (−$7131 [p = 0.04]) and professor (+$10,776 [p = 0.02]) compared to the reference “none.” Lastly, there were significant differences in the following states compared with the reference “other”: California (−$11,655), Georgia (+$18,165), Iowa (+$19,228), Illinois (−$28,721), Michigan (−$64,617), Nebraska (−$24,671), New York (−$51,269), Pennsylvania (−$21,769) (p < 0.05 for each).
Table 3.
Univariate regression information for each variable as a covariate for salarya.
| Variable | Beta | P value |
|---|---|---|
| Sex | ||
| Male | Ref | — |
| Female | −4410 | 0.116 |
| Chief status | ||
| No | Ref | — |
| Yes | 33,673 | 5.52E-11 |
| Pay grade | ||
| Part-time | Ref | — |
| Full-time | 28,472 | 1.96E-23 |
| Service line | ||
| Surgery | Ref | — |
| Eye | −3692 | 0.185 |
| Fellowship | ||
| No | Ref | — |
| Yes | −1872 | 0.5 |
| VISN region | ||
| Southern | Ref | — |
| Central | −18,148 | 1.85E-07 |
| Northeast | −31,759 | 4.12E-16 |
| Western | −10,500 | 0.00313 |
| Complexity Level | ||
| 1a | Ref | — |
| 1b | 684 | 0.849 |
| 1c | −10,141 | 0.00906 |
| 2 | 16,721 | 0.00354 |
| 3 | 20,635 | 0.0608 |
| Years since residency | 478 | 0.000262 |
| Years at VHA | 192 | 0.393 |
| H-index | 310 | 0.0312 |
| Subspecialty | ||
| Comprehensive | Ref | — |
| Cornea | 2207 | 0.624 |
| Glaucoma | −5007 | 0.228 |
| Medical retina | 6488 | 0.361 |
| Neuro-ophth | −18,320 | 0.00542 |
| Ocular oncology/pathology | 18,159 | 0.242 |
| Oculoplastics | −11,336 | 0.0318 |
| Peds/adult strab | −12,011 | 0.204 |
| Surgical retina | 6418 | 0.166 |
| Uveitis | 2499 | 0.785 |
| Academic rank | ||
| None | Ref | — |
| Instructor | −11,594 | 0.193 |
| Assistant prof | −7131 | 0.0412 |
| Associate prof | 1821 | 0.649 |
| Adjunct prof | 8060 | 0.267 |
| Professor | 10,776 | 0.0161 |
| Chair | 9415 | 0.588 |
| State | ||
| Other | Ref | — |
| CA | −11,655 | 0.01 |
| DC | −4985 | 0.61 |
| FL | −6543 | 0.22 |
| GA | 18,165 | 0.0187 |
| IA | 19,228 | 0.0404 |
| IL | −28,721 | 4.79E-08 |
| LA | −4642 | 0.606 |
| MI | −64,617 | 3.28E-13 |
| MO | 6704 | 0.457 |
| NC | 12,286 | 0.0605 |
| NE | −24,671 | 0.0119 |
| NY | −51,269 | 1.43E-22 |
| OH | −1129 | 0.893 |
| OR | 10,229 | 0.275 |
| PA | −21,769 | 0.00111 |
| RI | −9950 | 0.288 |
| SC | −4075 | 0.597 |
| TN | 11,765 | 0.0947 |
| TX | 3172 | 0.532 |
| UT | −10,573 | 0.126 |
| VA | 11,950 | 0.222 |
| WI | −11,537 | 0.108 |
Reference category for each variable is indicated. For categorical variables, the beta value indicates the average change in salary when going from that variable’s reference category to the indicated category. For continuous variables, it indicates the average change in salary when that variable is increased by one. All beta values are in USD. P values are reported to 3 significant digits.
Table S2 (https://www.aupojournal.org/cgi/viewcontent.cgi?filename=0&article=1310&context=jao&type=additional&preview_mode=1) provides the regression information when evaluating for the interaction effect between each covariate and sex. A significant interaction effect was observed with the Northeast VISN region (−$15,991 for females [p = 0.042]), the state of Texas (−$21,670 for females [p = 0.037]), the chair academic rank (+$83,382 for females [p = 0.037], although the sample size for this subgroup was only three males and one female), and years at VHA (+$1393 for each additional year for females [p = 0.002]). Unexpectedly, the years at VHA:sex interaction brought out an initial sex effect of −$15,666 for females compared with the male reference (p = 0.0008). This relationship is shown in Figure S4 (https://www.aupojournal.org/cgi/viewcontent.cgi?filename=0&article=1310&context=jao&type=additional&preview_mode=1).
Pairwise differences between nominal variables
For the five nominal variables, pairwise testing indicated significant differences in salary for four VISN region comparisons (southern > central/northeast/western; western > northeast), two complexity level comparisons (level 2 > levels 1a/1c), one academic rank comparison (professor > assistant professor), and 43 state comparisons. Each of the average differences and adjusted p values for these 50 significant comparisons are reported in Table S3 (https://www.aupojournal.org/cgi/viewcontent.cgi?filename=0&article=1310&context=jao&type=additional&preview_mode=1). These results offer a more in-depth statistical summary of the comparative effects of individual variables on salary.
Multivariate regression model
Each of the variables that had a p value < 0.05 or a sex interaction effect with a p value < 0.05 were included into a baseline multivariable regression model. These included sex, chief status, paygrade, VISN region, years since residency, h-index, complexity level, subspecialty, academic rank, years at VHA, years at VHA:sex interaction, and VISN region:sex interaction. It is important to note that if an interaction effect is included in a multivariate regression model, the contributing variables must also be included as individual effects. It is not necessary for these components to be statistically significant for their interaction effect to be meaningful. The state variable was manually excluded because it was confounded with VISN region. The academic rank:sex interaction was not included because it was only significant within the chair subgroup, which only had a sample size of three males and one female. The baseline model was optimized into a final multivariate model which is displayed in Table 4. It uses the following covariates: sex, chief status, paygrade, VISN region, complexity level, academic rank, years at VHA, and the sex:years at VHA interaction. The model has an R2 of 0.347 and an Akaike Information Criterion (AIC) of 13,199. The beta for sex after accounting for all mentioned variables is −$9726 for females compared with the male reference (p = 0.018).
Table 4.
Final multivariate regression using a subset of individually significant covariates and interaction effects that were selected using bidirectional stepwise variable selection.a
| Variable | Beta | P value |
|---|---|---|
| Intercept | 292,205 | 2.47E-301 |
| Sex | ||
| Male | Ref | — |
| Female | −9726 | 0.0181 |
| Chief status | ||
| No | Ref | — |
| Yes | 25,249 | 1.93E-07 |
| Pay grade | ||
| Part-time | Ref | — |
| Full-time | 21,606 | 4.40E-13 |
| VISN region | ||
| Southern | Ref | — |
| Central | −14,231 | 1.58E-05 |
| Northeast | −31,334 | 2.85E-15 |
| Western | −11,542 | 6.07E-04 |
| Complexity level | ||
| 1a | Ref | — |
| 1b | 5000 | 0.121 |
| 1c | −9362 | 0.00915 |
| 2 | 27,590 | 1.19E-06 |
| 3 | 16,774 | 0.103 |
| Academic rank | ||
| None | Ref | — |
| Instructor | −3837 | 0.606 |
| Assistant prof | 723 | 0.821 |
| Associate prof | 8665 | 0.0172 |
| Adjunct prof | −7174 | 0.277 |
| Professor | 13,973 | 0.00119 |
| Chair | 30,322 | 0.0359 |
| Years at VHA | −158 | 0.547 |
| Sex Female: Years at VHA | 733 | 0.0679 |
Reference values are indicated. The intercept value is the estimated average salary for a physician when all reference values are true. All beta values are in USD. P values are reported to 3 significant digits.
Discussion
Among the ophthalmologists at the VHA for whom salary data was available, the initial observation of a 59–41 male-to-female ratio demonstrates a slight underrepresentation in female physicians. This is most apparent among the population who finished residency 27+ years ago, where there were almost four times as many males than females. This is consistent with findings that the sex-bias has decreased in the physician workforce since 1970.13
Of note, our data indicates that the ratio of male-to-female ophthalmologists is more balanced in the VHA than in the population of US-practicing ophthalmologists in the American Academy of Ophthalmology (AAO), where 2021 self-reported sex data indicated a ratio of 74 males to 26 females.14 Additionally, distribution of males and females are equitable across subspecialties, yet glaucoma skews female, and surgical retina heavily skews male. To note, this trend follows the findings of the 2020 American Board of Ophthalmology diplomates distribution of males and females in ophthalmic subspecialties.15
When observed as a crude comparison, sex alone had an insignificant effect on salary, but this is a common observation regarding sex and salary, because other confounding variables often mask underlying differences. Thus, it is more notable that there is a significant interaction effect between sex and years at the VHA as well as the northeast VISN, both of which reveal decreased female salary. The significant interaction effects between sex and VISN region has also been observed for VHA dermatologists in a study showing that VHA dermatologists make significantly more than their female counterparts in the Midwest.8 Although our trend was noted in the Northeast, we observed a similar interaction effect within the central VISN region, indicating a $11,096 lower average salary among females, but the effect was not statistically significant (p = 0.11; Table S2) (https://www.aupojournal.org/cgi/viewcontent.cgi?filename=0&article=1310&context=jao&type=additional&preview_mode=1). The sex:years at VHA interaction effect was particularly surprising, because neither covariate was significantly associated with salary when analyzed alone. However, evaluating their effects together not only uncovered a significant interaction effect, but also a significant effect of sex alone. When years at the VHA is equal to 0, the average female salary is $15,666 lower than the average male salary, which agrees with the findings in Table 2 regarding higher male salary in 0—2 years and 3—7 years at the VHA. Of note, this disparity is not present for physicians where years at VHA is >7 years. Because this is a cross-sectional study, we cannot infer the cause behind this inequity, but we note that male ophthalmologists who started at the VHA as recently as 2014—2021 tended on average to make a higher salary than their female counterparts. This trend of males starting at a higher salary was also seen in 2008 data for physicians in New York State when controlling for specialty, practice setting, work hours, and other characteristics.16
With regard to covariates other than sex in the linear regression, it is not surprising to see chief status, years since residency, h-index, and the professor academic rank all associated with higher salary.7–9 Years at VHA was not associated with higher pay, yet years since residency was. This could potentially indicate that years of total experience is more important when salary is considered as opposed to “loyalty” to the VHA. Another noteworthy point is that completion of a fellowship was not associated with a higher salary―but potentially with a lower average salary in the cases of neuro-ophthalmology and oculoplastics. Sub-specialists in ophthalmology may therefore avoid the VHA because of the opportunity cost of a decreased salary as opposed to other practice settings.
Significant differences in salary by VISN region is unexpected, with all regions having a significantly lower salary than the Southern region―the Northeast region considerably so. Increased demand may be contributing to higher salaries in the south, although there could be a variety of reasons for this trend. Additionally, full-time status appears to have a significant positive increase in when compared with the full-time equivalent effort of ophthalmologists who work part-time. This may be an important consideration for physicians exploring how best to split their work effort outside VHA.
The multivariate model shown in Table 4 is intended to be inferential, giving broad insights into what factors, taken together, may play a role in ophthalmologist salary at the VHA. This model can serve as guidance for future VHA physicians and can be a tool for the VHA as they seek to further eliminate both conscious and unconscious biases. Specifically, it may be important to adjust policy to make consistencies across starting salary while also considering cost of living and demand for ophthalmologists or certain subspecialists in the region. Additionally, these data may be value for those who may seek to weigh potential opportunity costs related to differences in VHA compensation vis-à-vis those at academic medical centers/university site, private practice, and private equity—owned practices. There may be additional factors to weigh, for example, the role of physician-scientist, as research-related compensation may be higher in the VHA versus the National Institutes of Health (e. g., VHA career development awards are typically higher than NIH career development awards in awarded USD for physicians) at early stages in one’s career but compensation and total research funding per investigator may be lower later in the career (e.g., VHA Merit award is typically lower than an NIH R01 grant in awarded USD).
There are several limitations to this study. A potential bias arises from the fact that 297 (32 %) of the 940 listed ophthalmologists did not have salary data available. Although initially concerning, FederalPay.org’s retrieval of salary data is directly from the government, thereby indicating that only active members are in this database. While our analyses were based on active ophthalmologists at the VHA, there may be some staff ophthalmologists who work at the VHA but who do not receive federal salary, such as through agreements with academic centers. This may lead to the discrepancy between our number of VHA ophthalmologists and the number found on the VHA’s Ophthalmology Service website.17 The cross-sectional nature of this study does not allow us to make conclusions about changes in physician salary over time―we are solely describing the current VHA ophthalmologist landscape.
The present study serves as an initial probe into VHA ophthalmologist pay equity. Further studies should be performed to better understand salary disparities in VHA ophthalmologist subpopulations. For instance, an analysis of longitudinal data of ophthalmologist salary could shed light on trends that are not visible in a single cross-sectional study. Additionally, it may be insightful to explore whether there is an impact on salary based on race or ethnicity. Nevertheless, this work serves as a reference resource across geographic locales, sex, and subspeciality that may be useful to ophthalmology trainees, chairs in academic ophthalmology, and VHA administrators―physician and nonphysician alike.
Supplementary Material
Financial Support
This work was supported by the Surgery Service of the Veterans Administration Ann Arbor Healthsystem, United States. We thank Chief of the Surgery Service, Dr. Nicholas Osborne, for support of the project. RCR was supported by National Eye Institute (R01EY030989, P30EY007003), and Research to Prevent Blindness (Departmental Grant to the University of Michigan Kellogg Eye Center and Career Advancement Award), Beatrice and Reymont Paul Foundation, March Hoops to Beat Blindness, the Taubman Institute, the Leonard G. Miller Endowed Professorship and Ophthalmic Research Fund at the Kellogg Eye Center, and the Grossman, Elaine Sandman, Marek and Maria Spatz (endowed fund), Greenspon, Dunn, Avers, Boustikakis, Sweiden, and Terauchi research funds.
Conflict of interest
Dr. Rao is a consultant for Astellas, Regeneron, and Sumitomo Dainippon.. Dr. Rao receives grant support from Luxa Biotechnology. None of the other authors have any financial disclosures to report.
Footnotes
Ethics information
None.
References
- 1.Chun-Hoon W 5 Fast Facts: The Gender Wage Gap. U.S. Department of Labor; 2023. https://blog.dol.gov/2023/03/14/5-fast-facts-the-gender-wage-gap#:~:text=On%20average%2C%20men%20are%20paid,paid%20%241%2C998%20and%20women%20%241%2C546. Accessed December 19, 2023. [Google Scholar]
- 2.Ness RB, Ukoli F, Hunt S, et al. Salary equity among male and female internists in Pennsylvania. Ann Intern Med. 2000; 133:104–110. [DOI] [PubMed] [Google Scholar]
- 3.Jagsi R, Griffith KA, Stewart A, Sambuco D, DeCastro R, Ubel PA. Gender differences in salary in a recent cohort of early-career physician-researchers. Acad Med. 2013;88: 1689–1699. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 4.Jena AB, Olenski AR, Blumenthal DM. Sex differences in physician salary in US public medical schools. JAMA Intern Med. 2016;176:1294–1304. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 5.Gilbert SB, Allshouse A, Skaznik-Wikiel ME. Gender inequality in salaries among reproductive endocrinology and infertility subspecialists in the United States. Fertil Steril. 2019;111:1194–1200. [DOI] [PubMed] [Google Scholar]
- 6.Emami-Naeini P, Lieng MK, Chen J. Sex differences in salaries of academic ophthalmologists in the United States. JAMA Ophthalmol. 2022;140:519–522. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 7.Dermody SM, Litvack JR, Randall JA, Malekzadeh S, Maxwell JH. Compensation of otolaryngologists in the Veterans Health Administration: is there a gender gap? Laryngoscope. 2019;129:113–118. [DOI] [PubMed] [Google Scholar]
- 8.Do MH, Lipner SR. Contribution of sex on compensation of Veterans Affairs-affiliated dermatologists: a cross-sectional study. Int J Womens Dermatol. 2020;6:414–418. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 9.Maxwell JH, Randall JA, Dermody SM, et al. Gender and compensation among surgical specialties in the Veterans Health Administration. Am J Surg. 2020;220:256–261. [DOI] [PubMed] [Google Scholar]
- 10.Veterans Health Administration of the United States of America. Summary of VHA facility complexity model. https://www.vendorportal.ecms.va.gov/FBODocument-Server/DocumentServer.aspx?DocumentId=2793591&FileName=VA118-16-R-1059-A00002002.docx. Accessed 19 December 2023.
- 11.Veterans Health Administration of the United States of America. Veterans Health administration Veteran Integrated Services Networks (VISNs). Accessed 19 December 2023. https://www.va.gov/HEALTH/visns.asp.
- 12.Hirsch JE. An index to quantify an individual’s scientific research output. Proc Natl Acad Sci USA. 2005;102: 16569–16572. https://pnas.org/doi/full/10.1073/pnas.0507655102. Accessed January 21, 2024. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Pelley E, Carnes M. When a specialty becomes “women’s work”: trends in and implications of specialty sex segregation in medicine. Acad Med. 2020;95:1499–1506. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Woreta FA, Gordon LK, Knight OJ, Randolph JD, Zebardast N, Pérez-González CE. Enhancing diversity in the ophthalmology Workforce. Ophthalmology. 2022;129:e127–e136. [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Steren BJ, Yee P, Rivera PA, Feng S, Pepple K, Kombo N. Sex distribution and trends of ophthalmology subspecialties, 1992–2020. Am J Ophthalmol. 2023;253:22–28. [DOI] [PubMed] [Google Scholar]
- 16.Lo Sasso AT, Richards MR, Chou CF, Gerber SE. The $16,819 pay gap for newly trained physicians: the unexplained trend of men earning more than women. Health Aff. 2011;30: 193–201. [Google Scholar]
- 17.US Department of Veterans Affairs VHA. Veterans affairs. Ophthalmology Services; June 25, 2014. https://www.patient-care.va.gov/Ophthalmology/Ophthalmology_Services.asp#:~:text=The%20Ophthalmology%20Service%20is%20an,facilities%20throughout%20the%20United%20States. Accessed November 22, 2024. [Google Scholar]
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.