This cross-sectional study examines trends in individual physician work hours and their contribution to clinical workforce changes over a 20-year period.
Key Points
Question
How have physician work hours, including total work hours contributed by the physician workforce, changed since 2001?
Findings
In this cross-sectional study of 87 297 monthly surveys of physicians from 17 599 unique households, average weekly hours worked by individual physicians declined by 7.6% from 2001 to 2021, driven by a decrease among men, particularly fathers, while mothers’ hours increased. Total weekly hours contributed by the physician workforce per capita grew at less than half the rate of US population growth, while advanced practice professional workforce hours rose considerably over the same period.
Meaning
Over 2 decades, physicians’ average weekly work hours consistently declined, more among men than women, with both positive and negative implications for gender equity; lagging growth in hours contributed by the physician workforce per capita was offset by a growing advanced practice professional workforce.
Abstract
Importance
Physician work hours are an underexplored facet of the physician workforce that can inform policy for the rapidly changing health care labor market.
Objective
To examine trends in individual physician work hours and their contribution to clinical workforce changes over a 20-year period.
Design, Setting, and Participants
This cross-sectional study focused on active US physicians between January 2001 and December 2021 who were included in the Current Population Survey. Outcomes for physicians, advanced practice professionals (APPs), and nonphysician holders of doctoral degrees were compared, and generalized linear models were used to estimate differences in time trends for weekly work hours across subgroups.
Main Outcomes and Measures
Physician and APP workforce size, defined as the number of active clinicians, 3-year moving averages of weekly work hours by individual physicians, and weekly hours contributed by the physician and APP workforce per 100 000 US residents.
Results
A total of 87 297 monthly surveys of physicians from 17 599 unique households were included in the analysis. The number of active physicians grew 32.9% from 2001 to 2021, peaking in 2019 at 989 684, then falling 6.7% to 923 419 by 2021, with disproportionate loss of physicians in rural areas. Average weekly work hours for individual physicians declined by 7.6% (95% CI, −9.1% to −6.1%), from 52.6 to 48.6 hours per week from 2001 to 2021. The downward trend was driven by decreasing hours among male physicians, particularly fathers (11.9% decline in work hours), rural physicians (−9.7%), and physicians aged 45 to 54 years (−9.8%). Physician mothers were the only examined subgroup to experience a statistically significant increase in work hours (3.0%). Total weekly hours contributed by the physician workforce per 10 000 US residents increased by 7.0%, from 13 006 hours in 2001 to 2003 to 13 920 hours in 2019 to 2021, compared with 16.6% growth in the US population over that time period. Weekly hours contributed by the APP workforce per 100 000 US residents grew 71.2% from 2010 through 2012 to 2019 through 2021.
Conclusions and Relevance
This cross-sectional study showed that physician work hours consistently declined in the past 20 years, such that physician workforce hours per capita lagged behind US population growth. This trend was offset by rapid growth in hours contributed by the APP workforce. The gap in physician work hours between men and women narrowed considerably, with diverging potential implications for gender equity. Increasing physician retirement combined with a drop in active physicians during the COVID-19 pandemic may further slow growth in physician workforce hours per capita in the US.
Introduction
During the COVID-19 pandemic, the health care workforce in the US has experienced substantially increased turnover and shortages across all professions, including physicians. National surveys during the pandemic have found that nearly 1 in 4 physicians intended to quit in the upcoming 2 years, while 1 in 3 physicians, particularly women, planned to reduce their work hours. Given prepandemic data showing that 40% of physicians, including more than 50% of female physicians, have high levels of burnout, many have raised concerns about the resilience of the physician workforce and the implications for health care workforce policy.
An underexplored facet of the physician workforce, in the context of a rapidly changing labor market, is physicians’ work hours. It cannot be assumed that physician work hours are stable given that prior research found a sharp drop in physician work hours from 1996 to 2008 after decades of stability. While there is vigorous debate about the nature and magnitude of a physician shortage, discussion often focuses on the absolute number of physicians in the US rather than the intensity of work contributed by individual physicians. There has been little evidence examining recent trends in physician work hours, especially in light of the major disruptions to the health care system since 2020 and the potential contributions of other health care professionals such as nurse practitioners (NPs) and physician assistants (PAs).
Changes in physician work hours could influence multiple policy debates. The first is the debate around projecting physician supply as older physicians, who historically work more hours, retire and are replaced by younger physicians and advanced practice professionals (APPs) such as NPs. Another active debate is the effect of burnout, which is positively associated with work hours, on the physician workforce. Recent surveys suggest that many physicians plan to reduce their work hours, but the degree to which physicians have actually cut back on working hours is unknown. Third, there is continued focus on gender disparities in physician salary and work expectations. Female physicians earn substantially less, perform more household and childcare duties, and receive more clinical requests than their male counterparts. Changing trends in work hours by gender have important implications for gender disparities in the physician workforce. Finally, changes in the physician workforce also influence the ongoing policy debate around the role of NPs and PAs, and persistent disputes around expanding scope of practice for these APPs. To address this evidence gap, we used 2 decades of federal survey data on the labor workforce from 2001 to 2021 to examine changes in physician work hours across multiple demographic groups and changes in the size and total hours contributed by the physician and APP workforces.
Methods
Data Sources
We used nationally representative data from the monthly Current Population Survey (CPS) from January 2001 to December 2021. The CPS is a federally administered survey by the US Census Bureau of approximately 60 000 households per month that includes individual-level data on all civilian household members older than 16 years who are not institutionalized. Households are interviewed monthly for 8 months in the course of 1 year. During the COVID-19 pandemic, the CPS survey mode transitioned from a mix of in-person and telephone interviewing to telephone-only interviewing. This caused the monthly survey response rate to drop from 81% to 84% in 2019 to 65% to 80% in March 2020 through December 2021. Per Harvard University’s institutional policy, institutional review board approval and written informed consent were not required for research using publicly available data.
Identifying Clinicians and Comparison Groups
Using occupational data in the CPS, we applied the following 3 criteria to define respondents as physicians: (1) self-identifying as a physician (eMethods 1 in the Supplement), (2) working greater than zero hours, and (3) employment in a hospital or clinical setting. Because the CPS does not identify resident physician trainees, we used respondents’ age of younger than 35 years as a proxy for being a resident, an approach which captures the majority of resident physicians.
For comparison, we examined APPs and advanced degree holders besides physicians. We defined PAs and NPs as (1) CPS respondents self-identifying as a PA or NP (eMethods 1 in the Supplement), (2) working greater than zero hours, and (3) working in a hospital or clinical setting. The CPS began collecting data on NPs starting in 2011, so data on this group were unavailable for the first decade of the study period. Finally, we used nonphysician professionals as an additional comparison group, defined as all holders of doctorate or professional school degrees, excluding physicians (eMethods 1 in the Supplement).
Study Outcomes
We estimated the workforce size for physicians, NPs, and PAs using nationally representative weights from the CPS. We also calculated a national estimate for weekly hours contributed by the physician workforce per 100 000 US residents (hereafter, physician workforce hours) by multiplying physician workforce size and average hours worked per week, then dividing by US residents in increments of 100 000 using annual data on the US population from The World Bank. We used the same procedure to calculate the workforce size of the other professional groups that we examined (PAs, NPs, and nonphysician holders of doctoral degrees).
We estimated average weekly work hours for individual physicians using the “total hours worked last week” variable in the CPS survey instrument. Survey respondents were asked to report the total hours they worked at all jobs during the week that contained the 12th day of the month. The CPS also asks respondents about their “usual” hours worked weekly, but because this was missing for 11% of physicians, we used the “total hours worked last week” variable as previous studies have done. The resulting data aligned with those from the American Time Use Survey, a detailed, but much smaller, survey on how individuals in the US spend their time administered to a subset of CPS respondents (eTable 1 in the Supplement).
Study Covariates
We analyzed weekly work hours for individual physicians across subgroups by multiple covariates: sex; parents vs nonparents (defined as respondents with any child younger than 18 years living in the household); mothers vs fathers (ie, parents by self-identified sex); residents (physicians younger than 35 years); attendings by age group (35-44, 45-54, 55-64, and >65 years); full-time vs part-time employee (<30 hours per week, as per the Internal Revenue Service definition); urban vs rural (urban physicians were those residing in a metropolitan area; all others were rural); census region (Northeast, South, Midwest, and West); foreign vs domestic birthplace (as defined by the CPS); and race (classified as American Indian and Pacific Islander [categorized together owing to small sample sizes], Asian, Black, and White by respondent self-report). We also captured single-earner vs dual-earner households (>1 individual in the household working any hours for pay).
Statistical Analysis
For unadjusted analyses of trends over time, we estimated 3-year moving averages of individual work hours, workforce size, and workforce hours per capita using survey weights from the CPS to produce nationally representative estimates. We used 3-year moving averages to capture long-term trends and smooth year-to-year variability across the 20-year study period. We additionally conducted a pandemic-focused analysis using annual averages in the 2019 to 2021 period, comparing workforce size and weekly work hours in 2019 vs 2021. Unadjusted standard errors for percentage change over time were calculated using a formula provided by the US Census Bureau (eMethods 2 in the Supplement).
We examined the effect of the pandemic period on the time trend of weekly work hours for individual physicians. We used ordinary least squares to calculate the linear predicted time trend in work hours in the absence of the pandemic (excluding 2020-2021 from the analysis), then compared the predicted time trend with the observed time trend, adjusting for all covariates listed in Table 1.
Table 1. Physician Workforce Demographics, 2001-2021a.
| Characteristic | Respondents, % | P valueb | ||
|---|---|---|---|---|
| 2001-2003 | 2010-2012 | 2019-2021 | ||
| Age, y | ||||
| 35-44 | 30.1 | 26.0 | 25.7 | <.001 |
| 45-54 | 30.1 | 25.7 | 22.6 | |
| 55-64 | 14.1 | 20.3 | 18.7 | |
| ≥65 | 5.5 | 9.3 | 12.2 | |
| Sex | ||||
| Female | 29.2 | 33.5 | 38.4 | <.001 |
| Male | 70.9 | 66.5 | 61.7 | |
| Race | ||||
| American Indian and Pacific Islanderc | 0.3 | 1.2 | 1.5 | <.001 |
| Asian | 13.9 | 15.6 | 17.7 | |
| Black | 4.4 | 5.1 | 6.2 | |
| White | 81.5 | 78.2 | 74.7 | |
| Parental statusd | ||||
| Parent | 43.4 | 39.9 | 38.7 | <.001 |
| Nonparent | 56.6 | 60.1 | 61.3 | |
| Parent | ||||
| Mother | 11.1 | 13.5 | 15.4 | <.001 |
| Father | 32.3 | 26.4 | 23.2 | |
| Work status | ||||
| Full time | 92.1 | 91.9 | 92.0 | .85 |
| Part timee | 8.0 | 8.1 | 8.0 | |
| Other | ||||
| Foreign born | 25.1 | 26.1 | 25.0 | .12 |
| Residentf | 20.2 | 18.5 | 20.8 | .26 |
| Dual full-time earner householdg | 37.9 | 39.6 | 50.9 | <.001 |
| Residence | ||||
| Urban | 85.6 | 89.0 | 90.1 | <.001 |
| Rural | 14.2 | 10.6 | 9.5 | |
| Census region | ||||
| Northeast | 29.0 | 27.4 | 21.9 | <.001 |
| Midwest | 24.2 | 21.6 | 18.9 | |
| South | 27.6 | 27.8 | 35.3 | |
| West | 19.2 | 23.2 | 23.8 | |
Data from 2001 to 2003 vs 2019 to 2021. Pearson χ2 tests were used to test statistical significance.
Categorized together owing to small sample sizes.
Parents are defined as individuals with dependent minors in their households.
Part-time employment is defined as at least 30 hours per week, as per the definition used by the Internal Revenue Service.
Residents are defined as physicians younger than 35 years.
Single-income households are defined as having only 1 individual reporting working any hours for pay, whereas dual-income households were those that reported more than 1 individual working any hours for pay.
We separately examined time trends in weekly work hours by individual physicians for subgroups by patient demographics defined previously, comparing subgroups to each other within categories. We used generalized linear models with robust standard errors clustered at the household level to test the statistical significance of differences in time trends within subgroup categories. A continuous measure of work hours per week was the independent variable in these models. For each subgroup category, we implemented a model that included an interaction term in which a categorical variable indicating subgroup was interacted with a linear term for year to yield a statistical test of differences in time trends between female and male respondents. Models were adjusted for all characteristics in Table 1. The models tested a 2-sided statistical significance at α = .05. Analyses were conducted using Stata statistical software, version 17.0 (StataCorp).
Results
Study Sample
The physician sample included 87 297 monthly surveys of physicians from 17 599 unique households over 20 years. The demographic characteristics of physicians evolved over the course of the 20-year study period (Table 1). There were substantial increases in the proportion of female physicians (29.2% in 2001-2003 to 38.4% in 2019-2021; P < .001) and non-White physicians (18.6% in 2001-2003 to 25.4% in 2019-2021; P < .001). The share of physicians older than 65 years more than doubled during the study period, increasing from 5.5% to 12.2% (P < .001). The proportion of physicians working part time (<30 hours per week) remained stable at 8% throughout the study period (P = .85).
Overall Trends in Workforce Size and Weekly Hours Worked
The physician workforce grew 32.9% over the course of the study period, from a 3-year moving average of 711 483 in 2001 through 2003 to 945 320 in 2019 through 2021 (Figure 1A). However, the 3-year moving average of physician weekly work hours dropped by 7.6% (95% CI, −9.1% to −6.1%) from 52.6 to 48.6 hours per week between 2001 and 2021. Of this decrease, 3.0% (95% CI, −4.5% to −1.5%) occurred over the most recent decade (2010-2012 to 2019-2021) (Table 2 and Figure 2). The adjusted linear time trend for weekly work hours decreased by 0.17 (95% CI, −0.19 to −0.15) hours per year from 2001 to 2021. Exclusion of the pandemic years 2020 and 2021 from the analysis did not statistically significantly change the time trend (−0.16 hours; 95% CI, −0.18 to −0.14 hours).
Figure 1. Trends in Size of the Clinician Workforce and Weekly Hours Contributed per 100 000 Population, 2001-2021.
Includes individuals who self-identify as physicians, nurse practitioners, and physician assistants in the Current Population Survey, as well as the sum of all clinicians. Each data point represents a 3-year moving average (ie, the estimate for 2002 represents an average of the raw data from 2001-2003). Estimates are weighted using sampling weights. Population sizes are estimated by dividing the sampling by 12 and summing weights within a given year and clinician group (eg, physicians). The weighted workforce size estimates are multiplied by the average work hours per year for a given clinician group, divided by the US population, in increments of 100 000 for that year.
Table 2. Trends in Average Weekly Work Hours Among Physicians and Physician Subgroupsa.
| Characteristic | No.b | Average weekly work hours | Relative change across time periods, % (95% CI) | P value for yearly time trendc | ||||
|---|---|---|---|---|---|---|---|---|
| 2001-2003 | 2010-2012 | 2019-2021 | 2010-2012 to 2019-2021 | 2001-2003 to 2019-2021 | 2010-2021 | 2001-2021 | ||
| All physicians | 36 137 | 52.6 | 50.1 | 48.6 | −3.0 (−4.5 to −1.5) | −7.6 (−6.1) | NA | NA |
| Age, y | ||||||||
| 35-44 | 9890 | 52.6 | 50.8 | 47.9 | −5.7 (−7.5 to −3.9) | −8.9 (−10.6 to −7.2) | .03 | <.001 |
| 45-54 | 9486 | 52.0 | 50.0 | 46.9 | −6.2 (−8.0 to −4.4) | −9.8 (−11.5 to −8.1) | .17 | .004 |
| 55-64 | 6366 | 48.6 | 49.4 | 48.3 | −2.3 (−4.4 to −0.3) | −0.7 (−3.2 to 1.7) | .29 | .05 |
| ≥65 | 3211 | 38.5 | 39.3 | 39.8 | 1.3 (−2.5 to 5.2) | 3.5 (−1.3 to 8.4) | Reference | Reference |
| Sex | ||||||||
| Female | 12 120 | 48.8 | 46.6 | 46.8 | 0.4 (−1.4 to 2.2) | −4.1 (−5.9 to −2.2) | .09 | .001 |
| Male | 24 017 | 54.1 | 51.8 | 49.7 | −4.0 (−5.2 to −2.9) | −8.1 (−9.2 to −7.0) | Reference | Reference |
| Race | ||||||||
| American Indian and Pacific Islanderd | 1173 | 51.4 | 51.5 | 45.8 | −11.1 (−18.8 to −3.4) | −10.9 (−18.8 to −2.9) | .25 | .02 |
| Asian | 4830 | 52.1 | 49.4 | 48.6 | −1.6 (−3.9 to 0.7) | −6.7 (−9.7 to −3.7) | .80 | .85 |
| Black | 1884 | 53.6 | 51.3 | 49.7 | −3.1 (−4.4 to −1.9) | −7.4 (−8.7 to −6.1) | .30 | .40 |
| White | 28 250 | 54.4 | 51.4 | 49.5 | −3.8 (−5.4 to −2.1) | −9.0 (−10.6 to −7.5) | Reference | Reference |
| Parental status | ||||||||
| Parent | 14 729 | 52.8 | 50.5 | 47.3 | −6.8 (−8.2 to −5.4) | −10.6 (−11.9 to −9.2) | .04 | .09 |
| Nonparent | 21 408 | 52.4 | 49.9 | 49.4 | −0.9 (−2.2 to 0.4) | −5.6 (−7.0 to −4.3) | Reference | Reference |
| Parent | ||||||||
| Mother | 4805 | 42.7 | 44.4 | 43.9 | −1.1 (−3.8 to 1.6) | 3.0 (−0.2 to 6.1) | .047 | <.001 |
| Father | 9924 | 56.3 | 53.5 | 49.6 | −7.4 (−9.0 to −5.8) | −11.9 (−13.4 to −10.4) | Reference | Reference |
| Training status | ||||||||
| Resident | 7184 | 59.5 | 55.5 | 55.5 | 0.0 (−2.1 to 2.1) | −6.8 (−8.8 to −4.8) | .02 | .97 |
| Attending | 28 979 | 50.7 | 48.9 | 46.5 | −4.8 (−5.9 to −3.7) | −8.2 (−9.3 to −7.2) | Reference | Reference |
| Work status | ||||||||
| Full time | 33 241 | 55.6 | 52.9 | 51.1 | −3.4 (−4.2 to −2.5) | −8.1 (−8.9 to −7.3) | .34 | <.001 |
| Part time | 2896 | 17.5 | 17.5 | 17.5 | 0.0 (−4.5 to 4.5) | 0.0 (−4.5 to 4.5) | Reference | Reference |
| Country of origin | ||||||||
| Foreign born | 9174 | 51.8 | 50.5 | 49.0 | −3.0 (−4.9 to −1.1) | −5.4 (−7.3 to −3.5) | .92 | .93 |
| US born | 26 963 | 52.9 | 50.0 | 48.4 | −3.2 (−4.3 to −2.1) | −8.5 (−9.6 to −7.4) | Reference | Reference |
| Household income | ||||||||
| Dual earner | 21 855 | 53.6 | 51.3 | 49.7 | −3.1 (−4.4 to −1.9) | −7.4 (−8.7 to −6.1) | .63 | .54 |
| Single earner | 14 282 | 54.4 | 51.4 | 49.5 | −3.8 (−5.4 to −2.1) | −9.0 (−10.6 to −7.5) | Reference | Reference |
| Residence | ||||||||
| Urban | 31 864 | 52.6 | 50.0 | 48.7 | −2.7 (−3.7 to −1.7) | −7.5 (−8.5 to −6.5) | .01 | .32 |
| Rural | 4273 | 52.0 | 51.3 | 47.0 | −8.4 (−11.6 to −5.2) | −9.7 (−12.8 to −6.7) | Reference | Reference |
Abbreviation: NA, not applicable.
Weekly work hours are self-reported hours worked in the previous week by physician respondents in the Current Population Survey. All demographic characteristics are self-reported.
Sample sizes are unweighted totals for the 3 time periods described (2001-2003, 2010-2012, and 2019-2021).
Statistical testing to compare subgroups to reference groups was completed using general linear regression models that included an interaction term in which a categorical variable indicating subgroup was interacted with a linear term for year to yield a statistical test of differences in time trends. Models were adjusted for all characteristics in Table 1.
Categorized together owing to small sample sizes.
Figure 2. Trends in Average Weekly Work Hours for Individual Physicians, 2001-2021.
Self-reported hours worked in the previous week by physician respondents in the Current Population Survey. Each data point represents a 3-year moving average (ie, the estimate for 2002 represents an average of the raw data from 2001-2003). Estimates are weighted using sampling weights.
In contrast with physician trends, the PA workforce grew by 86.0% from 2001 to 2021, and the NP workforce increased by 110.5% from 2011 to 2021, with both workforces continuing to grow during the first 2 years of the pandemic (Figure 1A). Hours for PAs decreased 5.8% (95% CI, −8.9% to −2.8%) during the study period, from 40.0 to 37.7 hours per week (eTable 2 in the Supplement). Weekly hours for NPs increased slightly over 2011 to 2021, from 36.6 to 37.9 hours per week (relative increase, 3.5%; 95% CI, 1.2%-5.8%).
When combining physician workforce size and hours worked, weekly hours contributed by the physician workforce per 100 000 residents increased by 7.0% from 13 006 in 2001 to 2003 to 13 920 in 2019 to 2021 (Figure 1B), compared with 16.6% growth in the total US population from 2001 to 2021. Between 2010 and 2012 and 2019 and 2021, the decade when complete data on both physician and APP hours were available, physician workforce hours increased 11.0% (12 536 hours in 2010-2012 to 13 920 hours in 2019-2021), while APP workforce hours per 100 000 residents grew 71.2% from 2598 in 2010 to 2012 to 4448 in 2019 to 2021. Weekly hours contributed by the clinician workforce overall (physicians plus APPs) per 100 000 residents grew by 21.4% (15 134 hours in 2010-2012 to 18 369 hours in 2019-2021).
Trends by Subgroup in Weekly Work Hours for Individuals
There was substantial heterogeneity in trends for weekly work hours by individuals across different demographic subgroups of physicians (Table 2). Male physicians’ weekly working hours decreased more than female physicians’ hours over the study period (−8.1% vs −4.1%; P = .001). In 2001, male physicians worked about 5 more hours per week than female physicians (54.1 hours vs 48.8 hours; a difference of 5.3 hours), but the male-female gap narrowed by nearly half by the end of the study period (49.7 hours vs 46.8 hours; a difference of 2.9 hours; Figure 3). Physician fathers substantially decreased their working hours in comparison with physician mothers over the 20-year study period (−11.9% vs 3.0%; P < .001; Table 2 and Figure 3), a change that narrowed the work hour gap between mothers and fathers from 13.6 hours to 5.7 hours across 2001 to 2021. Restricting the analysis to physician parents with children younger than 6 years, we found similar results: a 9.6% decrease in weekly work hours for fathers and 3.0% increase for physician mothers (P < .001).
Figure 3. Trends in the Weekly Work Hours for Individual Physicians and Nonphysician Holders of Doctoral Degrees by Gender and Parental Status, 2001-2021.
Self-reported hours worked in the previous week by physician respondents in the Current Population Survey. Each data point represents a 3-year moving average. Estimates are weighted using sampling weights. Mothers and fathers are defined by sex and the presence of dependent minors in the household.
Examining differences by age, physicians between the ages of 45 to 54 years had the largest decrease in work hours, with a 9.8% decrease in work hours in the past 20 years (52.0 hours to 46.9 hours per week; Table 2). Physicians older than 65 years had an increase in average work hours during the study period, though the change was not statistically significant due to a small sample size. Average weekly hours for residents dropped sharply in 2003 (eFigure 1 in the Supplement), likely due to new restrictions on duty hours. Since 2003, resident hours have been stable.
Trends in physician work hours between physicians of different races were largely similar (Table 2). Full-time physicians reduced weekly hours substantially over the study period (−8.1%; 95% CI, −8.9% to −7.3%), a decline from 55.6 hours to 51.1 hours per week, whereas part-time physicians’ working hours remained stable. Physicians from single-income vs dual-income households decreased work hours to a similar degree (−7.4% vs −9.0%; P = .34). In the past 10 years, hours among rural physicians declined more than urban physicians (−9.7% vs −7.5%; P = .01).
Nonphysician holders of doctoral and professional degrees also had decreases in weekly working hours, though the decline was smaller in magnitude compared with physicians (eTable 3 in the Supplement). Over the 20-year study period, weekly work hours for this group dropped 5.8% (95% CI, −6.2% to −5.4%), from 43.7 hours to 41.2 hours per week. Among nonphysician professionals, weekly work hours for men decreased more than weekly hours for women (−6.4% vs −2.0%; P < .001; Figure 3).
Trends During the COVID-19 Pandemic
The number of clinically active physicians peaked in 2019 at 989 684, then fell by 66 265 in 2021 to 923 419 (6.7% relative decrease; eTables 4 and 5 in the Supplement). There was a smaller proportion of rural, White, and US-born physicians clinically active in 2021 compared with 2019. Average weekly work hours for individual physicians dipped substantially during the first 3 months of the pandemic (April to June 2020), then rebounded and returned to the prepandemic time trend for the remainder of the study period (eFigure 2 in the Supplement). Hours contributed by the physician workforce per 100 000 residents were 14 798 in 2019 but dropped to 13 498 in 2021, which was associated with the decrease in the number of active of physicians between 2019 and 2021.
Discussion
Across 2 decades, individual physicians’ work hours consistently decreased with variable magnitude by age, sex, and family structure. Falling physician weekly hours appeared to be associated with full-time physicians (as opposed to a change in part-time physicians), younger physicians, and male physicians—particularly fathers—while mothers were 1 of the few demographic groups with an increase in work hours. The declining trend in weekly work hours did not have a statistically significant association with the pandemic, suggesting that pandemic-related factors, such as reduced demand for outpatient and elective services, was not a major driver of the findings. However, these findings may be partly explained by secular changes among all US residents with advanced graduate degrees. Decreasing work hours meant that total weekly hours contributed by the physician workforce per 10 000 US residents increased at half the rate of population growth over the past 20 years. However, this slow growth was offset by rapid expansion of the APP workforce size, which led to a near doubling of the growth in clinician workforce hours per capita (ie, total hours contributed by both physicians and APPs) in the decade from 2010 to 2021.
Interpretation of this pattern depends in part on projections of patient demand over the same period as well as the substitutability of physician and APP workforce hours, both debated topics without established consensus. To the extent that APP work hours can substitute for physician time, the expansion of the APP workforce appears to play a key role in filling the gap between patient demand and physician supply. However, as the US population grows and medical technology expands, the physician workforce exit that began in 2020 may portend more pronounced physician shortages in local areas, particularly in rural North America.
One striking trend in these findings was the large decline in the work hour gap between male and female physicians, especially between physician fathers and mothers. This gap narrowed by more than half over the course of the study period, largely associated with a decline in work hours by fathers, with a small increase in work hours among mothers. This narrowing may partially address the well-documented gender gap in physician compensation, but only in regard to total hours worked, as opposed to revenue generated per hour. Falling work hours for men and fathers, among both physicians and nonphysicians with doctoral-level degrees, may reflect broadly changing attitudes toward work-life balance and perhaps greater sharing of domestic duties among the professional-level workforce. The increase in work hours among physician mothers, including those with children younger than 6 years, may suggest improvement over time in workplace accommodations for physician mothers. But the narrowing of the gender gap in physician work hours does not necessarily imply marked improvement in household gender equity in work-life balance. The gender gap in household labor has been declining since the 1960s, but women still do considerably more housework and childcare than men. The pandemic brought further challenges, with many physician mothers struggling with increased burden of responsibilities at work and at home, leading many to consider leaving clinical medicine.
The present results have potentially important implications for physician workforce policy. While the workforce has become more diverse, a further increase in the number of Black, other minority, and female physicians is needed for the physician workforce to reflect the demographic diversity of the US population. In addition, growth in the size of the physician workforce over the past 2 decades has been the key factor that has allowed physician hours to not fall further behind population growth. The impending retirement of a large cohort of older physicians may accelerate the decline in physician work hours. While the APP workforce is still small compared with the physician workforce, the rapidly increasing number of APPs is offsetting the decreasing trend in physician hours, an important shift in the composition of the US clinical workforce. If physicians across the US continue to leave the workforce faster than new trainees graduate, policy makers and leaders in medical education will likely need to allocate resources to enlarge the training pipeline for both physicians and APPs. These findings highlight the potential importance of state legislation to expand the scope of practice for APPs, particularly to maintain health care access in vulnerable areas of the country.
Limitations
This study has limitations. First, the CPS lacks detailed information on physician specialty; although the CPS did start distinguishing between surgeons and medical doctors in the past 2 years, this distinction was too recent to use in this longitudinal study. Second, the response rate to the CPS dropped during the pandemic due to changes in CPS procedures, including cessation of in-person interviews. However, this mainly affected less educated, lower-income respondents, a group that was mostly not included in this study of physician workers. Third, while the CPS does ask respondents to report “usual hours” worked per week, we used actual hours worked in the week prior to the survey data collection. The CPS-based estimates of actual hours worked may differ from other data sources, due in part to differences in how work hours are defined. However, the present results closely matched trends in the American Time Use Survey, a smaller survey using different methods. Fourth, the CPS only started collecting data on NPs in 2011 and in prior years grouped NPs with nurses. This limited the ability to analyze data on APPs for the full duration of the study period.
Conclusions
This cross-sectional study shows that physician work hours have consistently declined in the past 20 years. Growth in physician workforce hours per capita lagged behind US population growth, but this trend was offset by rapid growth in the APP workforce. The gap in physician work hours between men and women narrowed considerably, though the implications for gender equity could be mixed. Increasing physician retirement combined with a drop in active physicians during the COVID-19 pandemic may further slow growth in physician workforce hours per capita in the US.
eMethods 1. Occupation Codes
eTable 1. Benchmarking Weekly Work Hours by Individual Physicians in the CPS to the American Time Use Survey (ATUS)
eMethods 2. Formula for Calculating the Standard Error of the Percent Change Over Time
eTable 2. Advanced Practice Practitioners
eFigure 1. Average Weekly Work Hours among Resident Physicians, 2001-2021
eTable 3. Trends in Average Weekly Work Hours Among Non-Physician Workers with Doctoral Degrees by Selected Subgroups
eTable 4. Weighted and Unweighted Sample Sizes for Physician Workforce, 2019-2021
eTable 5. Demographic Changes in Physician Workforce, 2019-2021
eFigure 2. Average Weekly Work Hours by Quarter, Individual Physicians 2016-2021
References
- 1.Impact of the COVID-19 pandemic on the hospital and outpatient clinician workforce: challenges and policy responses. Department of Health and Human Services . May 3, 2022. Accessed November 10, 2022. https://aspe.hhs.gov/sites/default/files/documents/9cc72124abd9ea25d58a22c7692dccb6/aspe-covid-workforce-report.pdf
- 2.Wager E, Telesford I, Hughes-Cromwick P, Amin K, Cox C. What impact has the coronavirus pandemic had on health employment? Peterson-KFF Health System Tracker . August 24, 2022. Accessed November 10, 2022. https://www.healthsystemtracker.org/chart-collection/what-impact-has-the-coronavirus-pandemic-had-on-healthcare-employment/
- 3.Galvin G. Nearly 1 in 5 health care workers have quit their jobs during the pandemic. Morning Consult . October 4, 2021. Accessed November 10, 2022. https://morningconsult.com/2021/10/04/health-care-workers-series-part-2-workforce/
- 4.ANA urges US Department of Health and Human Services to declare nurse staffing shortage a national crisis. American Nurses Association . September 1, 2021. Accessed November 10, 2022. https://www.nursingworld.org/news/news-releases/2021/ana-urges-us-department-of-health-and-human-services-to-declare-nurse-staffing-shortage-a-national-crisis/
- 5.Frank E, Zhao Z, Fang Y, Rotenstein LS, Sen S, Guille C. Experiences of work-family conflict and mental health symptoms by gender among physician parents during the COVID-19 pandemic. JAMA Netw Open. 2021;4(11):e2134315. doi: 10.1001/jamanetworkopen.2021.34315 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 6.Shanafelt TD, Sinsky C, Dyrbye LN, Trockel M, West CP. Burnout among physicians compared with individuals with a professional or doctoral degree in a field outside of medicine. Mayo Clin Proc. 2019;94(3):549-551. doi: 10.1016/j.mayocp.2018.11.035 [DOI] [PubMed] [Google Scholar]
- 7.Staiger DO, Auerbach DI, Buerhaus PI. Trends in the work hours of physicians in the United States. JAMA. 2010;303(8):747-753. doi: 10.1001/jama.2010.168 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 8.Gudbranson E, Glickman A, Emanuel EJ. Reassessing the data on whether a physician shortage exists. JAMA. 2017;317(19):1945-1946. doi: 10.1001/jama.2017.2609 [DOI] [PubMed] [Google Scholar]
- 9.Eden J, Berwick D, Wilensky G, eds. Graduate Medical Education That Meets the Nation’s Health Needs. National Academies Press; 2014. [PubMed] [Google Scholar]
- 10.Orr R. The U.S. has much to gain from more doctors. Niskanen Center . August 4, 2021. Accessed November 10, 2022. https://www.niskanencenter.org/the-u-s-has-much-to-gain-from-more-doctors/
- 11.Dyrbye LN, Shanafelt TD, Sinsky CA, et al. Burnout among health care professionals: a call to explore and address this underrecognized threat to safe, high-quality care. National Academy of Medicine . July 5, 2017. Accessed November 10, 2022. https://nam.edu/burnout-among-health-care-professionals-a-call-to-explore-and-address-this-underrecognized-threat-to-safe-high-quality-care/
- 12.Rittenberg E, Liebman JB, Rexrode KM. Primary care physician gender and electronic health record workload. J Gen Intern Med. 2022;37(13):3295-3301. doi: 10.1007/s11606-021-07298-z [DOI] [PMC free article] [PubMed] [Google Scholar]
- 13.Whaley CM, Koo T, Arora VM, Ganguli I, Gross N, Jena AB. Female physicians earn an estimated $2 million less than male physicians over a simulated 40-year career. Health Aff (Millwood). 2021;40(12):1856-1864. doi: 10.1377/hlthaff.2021.00461 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 14.Ganguli I, Sheridan B, Gray J, Chernew M, Rosenthal MB, Neprash H. Physician work hours and the gender pay gap—evidence from primary care. N Engl J Med. 2020;383(14):1349-1357. doi: 10.1056/NEJMsa2013804 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 15.Robeznieks A. Why stopping scope creep is about protecting patients. American Medical Association . September 7, 2022. Accessed November 10, 2022. https://www.ama-assn.org/practice-management/scope-practice/why-stopping-scope-creep-about-protecting-patients
- 16.Smolko B, Gables LM. AAPA response to AMA’s #StopScopeCreep campaign. American Academy of Physician Associates . Published November 2, 2020. Accessed November 10, 2022. https://www.aapa.org/news-central/2020/11/aapas-response-to-amas-stopscopecreep-campaign/
- 17.Joint statement on the scope of nurse practitioners. American Association of Nurse Practitioners . November 12, 2020. Accessed November 10, 2022. https://www.aanp.org/advocacy/advocacy-resource/aanp-and-the-np-roundtable-joint-statements/joint-statement-on-the-scope-of-nurse-practitioners
- 18.Current Population Survey: Complete Technical Documentation. US Census Bureau . Updated September 20, 2022. Accessed November 10, 2022. https://www.census.gov/programs-surveys/cps/technical-documentation/complete.html
- 19.Rothbaum J. How does the pandemic affect survey response: using administrative data to evaluate nonresponse in the Current Population Survey Annual Social and Economic Supplement. US Census Bureau . September 17, 2022. Accessed November 10, 2022. https://www.census.gov/newsroom/blogs/research-matters/2020/09/pandemic-affect-survey-response.html
- 20.Population, total—United States. The World Bank . Accessed November 10, 2022. https://data.worldbank.org/indicator/SP.POP.TOTL?end=2020&locations=US&start=1960&view=chart
- 21.Identifying full-time employees. Internal Revenue Service . Updated September 29, 2022. Accessed November 10, 2022. https://www.irs.gov/affordable-care-act/employers/identifying-full-time-employees
- 22.US Census regions and divisions of the United States. US Census Bureau . Accessed November 10, 2022. https://www2.census.gov/geo/pdfs/maps-data/maps/reference/us_regdiv.pdf
- 23.Current Population Survey interviewing manual. US Census Bureau . April 2015. Accessed November 10, 2022. https://www2.census.gov/programs-surveys/cps/methodology/intman/CPS_Manual_April2015.pdf
- 24.Bianchi SM, Sayer LC, Milkie MA, Robinson JP. Housework: who did, does or will do it, and how much does it matter? Soc Forces. 2012;91(1):55-63. doi: 10.1093/sf/sos120 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 25.Kotila LE, Schoppe-Sullivan SJ, Dush CMK. Time in parenting activities in dual-earner families at the transition to parenthood. Fam Relat. 2013;62(5):795-807. doi: 10.1111/fare.12037 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 26.Lyu HG, Davids JS, Scully RE, Melnitchouk N. Association of domestic responsibilities with career satisfaction for physician mothers in procedural vs nonprocedural fields. JAMA Surg. 2019;154(8):689-695. doi: 10.1001/jamasurg.2019.0529 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 27.Linos E, Halley MC, Sarkar U, et al. Anxiety levels among physician mothers during the COVID-19 pandemic. Am J Psychiatry. 2021;178(2):203-204. doi: 10.1176/appi.ajp.2020.20071014 [DOI] [PubMed] [Google Scholar]
- 28.Halley MC, Mathews KS, Diamond LC, et al. The intersection of work and home challenges faced by physician mothers during the coronavirus disease 2019 pandemic: a mixed-methods analysis. J Womens Health (Larchmt). 2021;30(4):514-524. doi: 10.1089/jwh.2020.8964 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 29.McDade W. Competency and the aging physician: CME report 5-A-15. American Medical Association . Accessed November 10, 2022. https://www.cppph.org/wp-content/uploads/2016/02/AMA-Council-on-Medical-Education-Aging-Physician-Report-2015.pdf
- 30.Xue Y, Smith JA, Spetz J. Primary care nurse practitioners and physicians in low-income and rural areas, 2010-2016. JAMA. 2019;321(1):102-105. doi: 10.1001/jama.2018.17944 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 31.Barnes H, Richards MR, McHugh MD, Martsolf G. Rural and nonrural primary care physician practices increasingly rely on nurse practitioners. Health Aff (Millwood). 2018;37(6):908-914. doi: 10.1377/hlthaff.2017.1158 [DOI] [PMC free article] [PubMed] [Google Scholar]
- 32.Lai AY, Skillman SM, Frogner BK. Is it fair? how to approach professional scope-of-practice policy after the COVID-19 pandemic. HealthAffairs . June 29, 2020. Accessed November 10, 2022. https://www.healthaffairs.org/do/10.1377/forefront.20200624.983306/full/
Associated Data
This section collects any data citations, data availability statements, or supplementary materials included in this article.
Supplementary Materials
eMethods 1. Occupation Codes
eTable 1. Benchmarking Weekly Work Hours by Individual Physicians in the CPS to the American Time Use Survey (ATUS)
eMethods 2. Formula for Calculating the Standard Error of the Percent Change Over Time
eTable 2. Advanced Practice Practitioners
eFigure 1. Average Weekly Work Hours among Resident Physicians, 2001-2021
eTable 3. Trends in Average Weekly Work Hours Among Non-Physician Workers with Doctoral Degrees by Selected Subgroups
eTable 4. Weighted and Unweighted Sample Sizes for Physician Workforce, 2019-2021
eTable 5. Demographic Changes in Physician Workforce, 2019-2021
eFigure 2. Average Weekly Work Hours by Quarter, Individual Physicians 2016-2021