A, Probability among nonphysicians (n = 26 640) and physicians (n = 5328). B, Probability among family physicians (n = 2442) and other specialists (n = 1878). Age in years serves as the time scale; HR indicates hazard ratio of physicians having children within the indicated age range, with nonphysicians serving as reference .
eTable 1. Datasets from CPSO and ICES used in the present study
eTable 2. List of Johns Hopkins Aggregated Diagnosis Groups (ADGs), with examples of common included diagnoses
eTable 3. Baseline characteristics of physicians and nonphysicians at the date of licensing
eTable 4. Childbirth in nulliparous physicians and nonphysicians after their actual or simulated licensing date respectively
eFigure 1. Plot of log[-log(survival)] versus log (time) for assessment of the proportional hazards assumption
eFigure 2. Sensitivity analyses comparing the cumulative probability of childbirth
Customize your JAMA Network experience by selecting one or more topics from the list below.
Identify all potential conflicts of interest that might be relevant to your comment.
Conflicts of interest comprise financial interests, activities, and relationships within the past 3 years including but not limited to employment, affiliation, grants or funding, consultancies, honoraria or payment, speaker's bureaus, stock ownership or options, expert testimony, royalties, donation of medical equipment, or patents planned, pending, or issued.
Err on the side of full disclosure.
If you have no conflicts of interest, check "No potential conflicts of interest" in the box below. The information will be posted with your response.
Not all submitted comments are published. Please see our commenting policy for details.
Cusimano MC, Baxter NN, Sutradhar R, et al. Delay of Pregnancy Among Physicians vs Nonphysicians. JAMA Intern Med. 2021;181(7):905–912. doi:10.1001/jamainternmed.2021.1635
Are women physicians more likely to delay childbearing or less likely to have children compared with nonphysicians?
In this population-based retrospective cohort study of 5238 reproductive-aged physicians matched 1:5 to nonphysician counterparts, physicians significantly postponed the initiation of childbearing. Despite this delay, physicians ultimately achieved a similar probability of childbirth as nonphysicians, owing to higher rates of pregnancy at advanced maternal ages; this phenomenon was most pronounced for specialists.
Physicians appear to delay childbearing and may be at increased risk of age-related adverse pregnancy outcomes.
Women physicians may delay childbearing and experience childlessness more often than nonphysicians, but existing knowledge is based largely on self-reported survey data.
To compare patterns of childbirth between physicians and nonphysicians.
Design, Setting, and Participants
Population-based retrospective cohort study of reproductive-aged women (15-50 years) in Ontario, Canada, accrued from January 1, 1995, to November 28, 2018, and observed to March 31, 2019. Outcomes of 5238 licensed physicians of the College of Physicians and Surgeons of Ontario were compared with those of 26 640 nonphysicians (sampled in a 1:5 ratio). Physicians and nonphysicians were observed from age 15 years onward.
Physicians vs nonphysicians.
Main Outcomes and Measures
The primary outcome was childbirth at gestational age of 20 weeks or greater. Cox proportional hazards models were used to examine the association between physician status and childbirth, overall and across career stage (postgraduate training vs independent practice) and specialty (family physicians vs specialists).
All physicians (n = 5238) and nonphysicians (n = 26 640) were aged 15 years at baseline, and 28 486 (89.1%) were Canadian-born. Median follow-up was 15.2 (interquartile range, 12.2-18.2) years after age 15 years. Physicians were less likely to experience childbirth at younger ages (hazard ratio [HR] for childbirth at 15-28 years, 0.15; 95% CI, 0.14-0.18; P < .001) and initiated childbearing significantly later than nonphysicians; the cumulative incidence of childbirth was 5% at 28.6 years in physicians and 19.4 years in nonphysicians. However, physicians were more likely to experience childbirth at older ages (HR for 29-36 years, 1.35; 95% CI, 1.28-1.43; P < .001; HR for ≥37 years, 2.62; 95% CI, 2.00-3.43; P < .001), and ultimately achieved a similar cumulative probability of childbirth as nonphysicians overall. Median age at first childbirth was 32 years in physicians and 27 years in nonphysicians (P < .001). After stratifying by specialty, the cumulative incidence of childbirth was higher in family physicians than in both surgical and nonsurgical specialists at all observed ages.
Conclusions and Relevance
The findings of this cohort study suggest that women physicians appear to delay childbearing compared with nonphysicians, and this phenomenon is most pronounced among specialists. Physicians ultimately appear to catch up to nonphysicians by initiating reproduction at older ages and may be at increased risk of resulting adverse reproductive outcomes. System-level interventions should be considered to support women physicians who wish to have children at all career stages.
Despite increasing gender parity in the physician workforce, a career in medicine is still frequently viewed as a barrier to motherhood.1-4 Women physicians who wish to have children face demanding work hours, limited options for parental leave and child support, and potential stigmatization by peers and superiors.3-5 Women physicians may therefore remain childless or delay childbearing relative to the general population; 50% to 60% report postponing pregnancy to independent practice,6,7 and 25% who attempt conception report infertility.8,9 These factors may place women physicians at risk of age-related adverse reproductive outcomes.10,11
Existing studies examining pregnancy and childbirth in women physicians are almost exclusively self-reported surveys prone to sampling and information bias. To our knowledge, only 1 observational study has described reproductive patterns in women physicians12; the authors found that maternal age at delivery was higher for physicians relative to nonphysicians, but the study did not explore time to childbirth or whether parity, specialty, or training status influenced the trends observed.
Large epidemiologic studies using validated data sources are needed to accurately characterize patterns of childbirth among women physicians. These data would contextualize pregnancy outcomes in this population and directly inform reproductive planning and care. We therefore examined patterns of childbirth in physicians compared to nonphysicians using population-based health administrative data.
We performed a population-based retrospective cohort study of reproductive-aged women in Ontario, Canada, where 14.6 million citizens reside and 40% of Canadian childbirths occur.13 All Ontario residents are eligible for universal health insurance coverage for hospital and physician services. The study protocol was published14 and approved by the Research Ethics Board at St Michael’s Hospital (Toronto, Ontario, No. 18-248) (Supplement 1).
To practice medicine in Ontario, physicians must obtain a medical license from the College of Physicians and Surgeons of Ontario (CPSO), which is the sole regulatory body that grants medical licenses in Ontario. Physicians are first granted a postgraduate education license at completion of medical school and initiation of residency training, and subsequently granted an independent practice license after examination and certification by either the College of Family Physicians of Canada or the Royal College of Physicians and Surgeons of Canada.
To complete this study, we obtained a data set of Ontario physicians licensed by the CPSO and linked this data set to population-based databases held at ICES, a nonprofit research institute and prescribed entity under section 45 of Ontario’s Personal Health Information Protection Act, which authorizes ICES to collect personal health information on all Ontario residents, without consent, for the purpose of health system evaluation and improvement (eTable 1 in Supplement 2). The unique linkage of CPSO data to ICES data enabled identification of physicians and nonphysicians, physician characteristics (eg, specialty, date of licensing to train or practice independently), covariates, and outcomes.14 Data sets were linked using unique encoded identifiers and analyzed at ICES. The study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.
Women were classified as physicians if they had a record of being licensed to practice medicine with the CPSO, either as a postgraduate trainee or independent practitioner. We included all women physicians (aged 15-50 years) who (1) were first licensed between January 1, 1995, and November 26, 2018; (2) were Ontario residents on the date that they were granted their CPSO license; and (3) had been eligible for provincial health insurance since the age of 15 years. These strict criteria enabled selection of women in which all childbirths over the reproductive life span could be accurately detected in ICES databases.
Nonphysician women (aged 15-50 years) were drawn from the ICES Registered Persons Database and randomly assigned a simulated medical licensing date according to the distribution of all licensing dates in physicians. Nonphysicians were included if they (1) were alive on their simulated licensing date; (2) were Ontario residents on that date; and (3) had been eligible for provincial health insurance since the age of 15 years. This approach mirrored the selection of physicians, who by definition were alive on the date of licensing.
We aimed to determine whether patterns of childbirth differed for physicians and nonphysicians over the reproductive life span. To do so, we observed physicians and nonphysicians from the date of their 15th birthday. For each physician, we sampled 5 eligible nonphysicians born in the same year to ensure that groups were balanced on age and era of cohort entry. It should be recognized that a period of immortal time was introduced in this process; however, its extent was similar in physicians and nonphysicians, as neither could die prior to their actual or simulated licensing date, respectively, and it was unlikely to bias results because the cumulative probability of death after licensing was also very low (<0.2%).
The primary outcome was time to childbirth, defined as the time in years from individuals’ 15th birthday to their first subsequent singleton or multiple live birth or stillbirth at gestational age of 20 weeks or greater. This outcome was selected as the best available reflection of an intention to continue pregnancy to term. Childbirths were identified from the ICES-derived MOMBABY database, which links the inpatient admission records of all mothers and newborns in Ontario (eTable 1 in Supplement 2). Because 97% of Ontario births occur in hospitals, use of MOMBABY ensured accurate outcome ascertainment. Data on childbirths were available to March 31, 2019.13
Demographic characteristics were ascertained for physicians and nonphysicians at age 15 years and included residential income quintile, era of cohort entry (1995-2006, 2007-2018), immigration status (immigrant, Canadian-born),15 comorbidities (0, 1-5, 6-9, ≥10), and previous live births (0, 1, ≥2). Residential income quintile is an area-level measure of socioeconomic status derived from Canadian census data on the median reported income for the neighborhood where individuals live.16,17 Comorbidities were categorized into Aggregated Diagnosis Groups on the basis of clinical similarity, chronicity, disability, and likelihood of requiring specialty care, using the Johns Hopkins ACG System, version 10 (Johns Hopkins Healthcare Solutions) (eTable 2 in Supplement 2).18
Physician characteristics were determined at the date of licensing and included career stage (postgraduate trainee, independent practitioner) and specialty (family medicine, other specialists, or specialty not yet determined). Career stage was treated as a time-varying covariate; because only independent practitioners can submit billing claims for health services, we used the first claim submitted to the Ontario Health Insurance Plan as an indicator of the transition from postgraduate trainee to independent practitioner (eTable 1 in Supplement 2).14 Specialty was assigned based on information in the CPSO data set or other ICES databases (eTable 1 in Supplement 2). If physicians had no documented specialty in any available data source and were recent medical graduates (≥2013), indicating ongoing postgraduate training,1 then their specialty of practice was categorized as not yet determined.14
Baseline characteristics (at age 15 years) were compared between groups using standardized differences.19 Standardized differences reflect differences between group means and proportions relative to the pooled SD and are appropriate for comparison of frequency-matched or individually matched samples.19
A Cox proportional hazards model was used to assess whether physician status was associated with time to childbirth. Individuals were observed from age 15 years and censored at death, loss to follow-up (ie, loss of eligibility for provincial health insurance), or end of follow-up (ie, March 31, 2019, as per MOMBABY). We also plotted the cumulative probability of childbirth for physicians and nonphysicians using maternal age as the time scale, both overall and stratified by specialty (family physicians vs other specialists).
The proportional hazards assumption was evaluated with an interaction between physician status and time, and by plotting log[−log(survival)] vs log (time). Both methods demonstrated violation of the proportional hazards assumption (interaction P < .001; eFigure 1 in Supplement 2). We addressed this by fitting a piecewise Cox proportional hazards model at 3 time intervals: (1) age 15 to 28 years; (2) age 29 to 36 years; and (2) age 37 years and older. Overall and piecewise models are presented with hazard ratios (HRs) and 95% CIs. Because hazards were nonproportional, the HRs presented reflect the average relative rate between physicians and nonphysicians over respective time periods.
We performed several sensitivity analyses to confirm that the findings were robust. To account for socioeconomic status as a potential confounder, we restricted the nonphysician comparator group to those in the highest income quintile only. To determine whether patterns of childbirth varied over time and over specific specialty type, we additionally stratified analyses for physicians born from 1976 to 1984 and 1985 to 1994, and for physicians in surgical and nonsurgical specialties.
To account for the fact that most physicians remained nulliparous until their licensing date, and to ensure that groups were as comparable as possible, we performed a supplemental analysis in which nulliparous physicians were matched to a separate group of nonphysicians who also remained nulliparous until their simulated licensing date. In this analysis: (1) each physician was individually matched to 5 eligible nonphysicians on age and year of licensing; (2) physicians and nonphysicians were observed from their actual or simulated date of licensing, respectively; and (3) physician status was modeled as 3-level time-varying exposure (nonphysician, postgraduate trainee, independent practitioner) to address the previously noted nonproportional hazards and enable an assessment of patterns of childbirth by career stage. A Cox proportional hazards model, with a robust variance estimator to account for correlation within individually matched sets, was used to assess whether physician status was associated with time to childbirth; we repeated these analyses with a stratified Cox proportional hazards model as an alternative approach to account for individual matching.20 Analyses were run overall and stratified by specialty (family physicians vs other specialists).
All statistical tests were 2-sided, with P less than .05 and standardized differences 0.1 and greater considered significant.19 Complete case analyses were performed, as data were rarely missing (residential income quintile, <0.5%; physician specialty, 2.0%). Emigration from Ontario was 5% for nonphysicians and 4% for physicians and was the only reason for loss to follow-up. Analyses were performed using SAS, version 9.4 (SAS Institute).
We identified 16 920 physicians of reproductive age who registered with the CPSO between January 1, 1995, and November 26, 2018. After excluding those who had not resided in Ontario since age 15 years, our cohort included 5328 physicians who could be observed over their reproductive life span; 2442 (45.8%) were training or practicing in family medicine, 1878 (35.2%) were training or practicing in other specialties, 900 (16.9%) had not completed training at any point during follow-up and were categorized as specialty not yet determined, and 108 (2.0%) were missing data on specialty. Physicians were successfully frequency matched to 26 640 nonphysicians at age 15 years (Figure 1).
At baseline (age 15 years), women who were ultimately licensed as physicians were more likely to live in high-income urban areas (2092 of 5328 [39.3%] vs 4653 of 26 640 [17.5%]; P < .001), less likely to live in rural areas (390 of 5328 [7.3%] vs 4141 of 26 640 [15.5%]; P < .001), and more likely to be immigrants (881 of 5328 [15.2%] vs 2671 of 26 640 [10.0%]; P < .001) than women who were nonphysicians (Table 1).
Median (interquartile range [IQR]) follow-up was 16.7 (14.7-19.1) years in physicians and 14.8 (11.6-17.9) years in nonphysicians (Table 2). Over the reproductive life span, physicians on average had a decreased rate of childbirth compared with nonphysicians (HR, 0.62; 95% CI, 0.59-0.65; P < .001). In piecewise models, physicians had a markedly decreased rate of childbirth from age 15 to 28 years (HR, 0.15; 95% CI, 0.14-0.18; P < .001), slightly increased rate of childbirth from age 29 to 36 years (HR, 1.35; 95% CI, 1.28-1.43; P < .001), and markedly increased rate of childbirth after age 37 years (HR, 2.62; 95% CI, 2.00-3.43), compared with nonphysicians (Table 3).
Age at initiation of childbearing was later for physicians than nonphysicians: the cumulative probability of childbirth was 5% at age 19.4 years in nonphysicians and 28.6 years in physicians (Figure 2A). However, by age 37 years, the cumulative incidence of childbirth was similar in both groups (62.7% in physicians, 62.1% in nonphysicians; Figure 2A). Median (IQR) age at first childbirth was 27.0 (22.6-30.2) years for nonphysicians and 31.6 (29.8-33.6) years for physicians (P < .001) (Table 2).
After stratifying by specialty, unadjusted rates of childbirth were higher in family physicians (2.83 per 100 person-years) than in specialists (2.42 per 100 person-years). The cumulative incidence of childbirth was also higher in family physicians than in specialists at all observed ages (Table 2; Figure 2B).
After restricting to nonphysicians in the highest income quintile only, physicians still appeared to delay childbirth relative to nonphysicians to a similar degree (eFigure 2A in Supplement 2). After stratifying by era of birth, physicians born recently (1985-1994) had a lower cumulative incidence of childbirth than physicians of the same age but born earlier (1976-1984; log-rank P < .001; eFigure 2B in Supplement 2). After further stratifying by specialty, patterns of childbirth did not differ substantially for surgical and nonsurgical specialists, but both had a lower cumulative incidence of childbirth relative to family physicians at all ages (log-rank P < .001; eFigure 2C in Supplement 2).
Approximately 98% of physicians (n = 5227) were nulliparous at the time of licensing (eTable 3 in Supplement 2). In supplemental analyses comparing these physicians with nulliparous nonphysicians, physicians had a decreased rate of childbirth (HR, 0.79; 95% CI, 0.74-0.85; P < .001) compared with nonphysicians while in postgraduate training, but an increased rate (HR, 2.23; 95% CI, 2.10-2.36; P < .001) while in independent practice (eTable 4 in Supplement 2). Results were similar whether a marginal or conditional approach was used to account for matching.
After stratifying by specialty, specialists had a decreased rate of childbirth compared with nonphysicians while in postgraduate training (HR, 0.71; 95% CI, 0.64-0.70; P < .001) but an increased rate while in independent practice (HR, 2.13; 95% CI, 1.92-2.36; P < .001). In contrast, family physicians had a rate of childbirth comparable to nonphysicians while in postgraduate training (HR, 0.93; 95% CI, 0.81-1.03; P = .17) but an increased rate while in independent practice (HR, 2.18; 95% CI, 2.03-2.35; P < .001) (eTable 4 in Supplement 2).
This population-based retrospective cohort study of more than 5300 physicians suggests that women pursuing a career in medicine delay childbearing relative to the general population. Physicians almost universally remained nulliparous prior to age 28 years but had high rates of childbirth thereafter, particularly on entering independent practice. As a result, physicians were often pregnant at advanced maternal ages, when the risks of infertility and adverse maternal and fetal outcomes are more pronounced.10,11
The present study demonstrates that delay of childbirth in physicians begins early and is directly associated with career stage and specialty. Only 2% of physicians had children before completing medical school; while family physicians went on to have rates of childbirth that were comparable to nonphysicians during residency, specialists continued to have decreased rates of childbirth until they began independent practice. This complements the observations of previous surveys, which have found that mean physician age at first childbirth ranges from 30 to 33 years8,21 and that only 14% to 40% of physicians experience a pregnancy during postgraduate training.6,7,22 To our knowledge, only 1 other observational study has compared reproductive patterns between physicians and nonphysicians. Using the Taiwan National Health Insurance Database, Wang et al12 showed that the median maternal age at admission for childbirth ranged from 31 to 33 years for physicians and 27 to 31 years for nonphysicians from 1996 to 2013; however, this study did not account for parity or assess outcomes by specialty or career stage. The present work not only confirms that physicians delay childbearing, but also identifies the time period when this delay occurs and suggests that the duration of delay may be lengthening over time rather than shortening.
The present study is the first to map the trajectory of childbirth in women physicians using epidemiologic data and directly model rates of childbirth for both postgraduate trainees and independent practitioners in differing specialties. In contrast to previous surveys, we studied a large population-based cohort of physicians licensed to practice in an entire province. By using validated administrative data, we observed individuals over prolonged follow-up and identified childbirths with little risk of misclassification. This work also provides the context required to understand the factors that contribute to adverse reproductive outcomes among physicians. We show that physicians experience childbirth at an advanced maternal age, which is associated with infertility and pregnancy complications.10,11 By identifying when and in whom pregnancy delay is most likely to occur, the findings can inform the development of strategies to support pregnancy and ensure equity in career advancement opportunities for trainees with children.
Several limitations merit consideration. First, to model the reproductive life span, we initiated observation at age 15 years, after exposure status had already been determined. It is possible that early childbirth could have prevented some women from becoming physicians, thus reducing the incidence of childbirth in those who ultimately did complete licensure. To address this, we conducted a supplemental analysis in which we initiated observation at the date of licensing; this confirmed delay of pregnancy in specialists and illustrated when most physicians initiate childbearing. Second, the study cohort does not include women who moved to Ontario after age 15 years. While it is possible that the results may not reflect the experiences of international medical graduates or mobile physicians, we would not anticipate these groups to be any less likely to delay childbirth. Third, we lacked data on relationship status, race/ethnicity, use of assisted reproductive technology, pregnancy intent, and occupation for nonphysicians. Delay of childbirth may also occur in other professions that require prolonged training, such as basic science, law, and engineering. These factors should be studied further; however, the consistency noted even after restricting to high-income nonphysicians and the dramatic shift in rates of childbirth after the completion of postgraduate training suggest that voluntary childlessness or a preference for delay is unlikely to explain the findings. Whether delay of pregnancy is due to career-related concerns,23,24 demanding academic schedules and limited support,4,25-29 or simply personal choice, women physicians in our current system must complete training during their primary reproductive years. Strategies that could be considered to ensure that women physicians can pursue pregnancy if and when they desire include adequate parental leave, remuneration for physicians during parental leave, options for childcare that extend beyond the traditional workday, increased flexibility in both undergraduate and postgraduate training schedules, and a culture of leadership that supports physician mothers and promotes the importance of shared parenting and domestic tasks.4,30
Women physicians appear to postpone childbearing compared with nonphysician counterparts, and this phenomenon is more pronounced in specialists. Although physicians ultimately achieve a similar cumulative probability of pregnancy as nonphysicians, they do so by initiating reproduction at older ages and may be at increased risk of adverse reproductive outcomes. System-level interventions are required to support women physicians who wish to have children at all career stages.
Accepted for Publication: March 15, 2021.
Published Online: May 3, 2021. doi:10.1001/jamainternmed.2021.1635
Corresponding Author: Andrea N. Simpson, MD, Department of Obstetrics and Gynecology, St Michael’s Hospital/Unity Health Toronto, 507-55 Queen St E, Toronto, Ontario, M5C 1R6, Canada (firstname.lastname@example.org).
Author Contributions: Dr Simpson had full access to all of the data in the study and takes responsibility for the integrity of the data and the accuracy of the data analysis.
Concept and design: Cusimano, Sutradhar, McArthur, Vigod, Simpson.
Acquisition, analysis, or interpretation of data: All authors.
Drafting of the manuscript: Cusimano, Sutradhar, Vigod.
Critical revision of the manuscript for important intellectual content: All authors.
Statistical analysis: Sutradhar, McArthur, Ray.
Obtained funding: Baxter, Simpson.
Administrative, technical, or material support: Cusimano, Baxter, Garg.
Supervision: Baxter, Sutradhar, Simpson.
Conflict of Interest Disclosures: Dr Cusimano reported being supported by the American College of Surgeons (ACS) Resident Research Scholarship and the Canadian Institutes of Health Research (CIHR) Vanier Canada Graduate Scholarship. Dr Baxter reported receiving grants from Physician Services Incorporation Foundation during the conduct of the study. Dr Garg reported being supported by the Dr Adam Linton Chair in Kidney Health Analytics and a CIHR Clinician Investigator Award. Dr Vigod reported receiving royalties for authorship from UpToDate Inc outside the submitted work. No other disclosures were reported.
Funding/Support: This study was conducted with grant funding from Physicians’ Services Incorporated (PSI) Foundation. This study was also supported by ICES, which is funded by an annual grant from the Ontario Ministry of Health and Long-Term Care (MOHLTC).
Role of the Funder/Sponsor: The funders had no role in the design and conduct of the study; collection, management, analysis, and interpretation of the data; preparation, review, or approval of the manuscript; and decision to submit the manuscript for publication.
Disclaimer: The analyses, conclusions, opinions, and statements expressed herein are solely those of the authors and do not reflect those of the funding or data sources; no endorsement by ICES or the MOHLTC is intended or should be inferred.
Additional Contributions: The authors thank Peter Tanuseputro, MD, Manish Sood, MD, and Emily Rhodes, MSc, at the Ottawa Hospital Research Institute, for their assistance with data acquisition. They were not compensated for their contributions. Parts of this material are based on data and information compiled and provided by the Canadian Institute for Health Information; Ontario Ministry of Health and Long-Term Care; Immigration, Refugees and Citizenship Canada Permanent Resident Database; Cancer Care Ontario; and Service Ontario. However, the conclusions, opinions, and statements expressed herein are solely those of the authors and not those of the bodies listed. No endorsement by these bodies is intended or should be inferred.
Additional Information: The data set from this study is held securely in coded form at ICES. While data sharing agreements prohibit ICES from making the data set publicly available, access may be granted to those who meet prespecified criteria for confidential access, available at https://www.ices.on.ca/DAS.