The 2021 Pediatric Hospital Medicine Workforce: Results of a National Survey of Program Leaders Free

The study was conducted in the late winter of 2021. It was a survey-based, cohort study of individuals who identify as PHM leaders, as part of a larger survey of pediatric hospitalists conducted by the American Academy of Pediatrics. The larger survey had 3 parts: one designed for all pediatric subspecialists, a second specific to pediatric hospitalists, and a third given only to those who identified themselves as PHM program, division, or section leaders. The 20 questions on this third section form the basis for this workforce analysis. The institutional review board at the American Academy of Pediatrics approved this study as exempt.

The PHM leader survey was developed through expert consensus after review of previous PHM workforce surveys (Supplemental Fig 1). The survey sought to better understand the staffing models, services covered, current work hours, and approaches to high census. The research team consisted of 2 members of the AAP staff with experience with surveys of pediatric subspecialists as well as a representative of each of the sponsoring societies of PHM (the AAP SOHM, APA, SHM). All members of the research team have experience in survey methodology and/or pediatric workforce studies. After consensus was achieved on all parts of the survey, it was pilot tested by 14 pediatric hospitalists recruited by the research team. The survey was adjusted for clarity on the basis of their responses and direct feedback.

To achieve the most generalizable results, a personalized link to the anonymous survey was sent to the membership lists of the AAP SOHM, APA Hospital Medicine Special Interest Group, and the Pediatric members of SHM after the removal of duplicate E-mail addresses. An E-mail was also sent to the listservs of all 3 societies to recruit additional pediatric hospitalists who were not members. People were invited to E-mail one of the authors (HBF) if they did not receive an invitation and wanted to complete the survey. The individuals were then sent their own personal link. Reminder emails with links were sent to the society members list that had not responded every week for 4 weeks. The listserv E-mail was also posted weekly for 4 weeks. A total of 3077 personal links were sent to individuals.

Data analysis was performed using SPSS 18.0 (IBM SPSS Statistics, IBM corporation). Descriptive statistics, including frequency distributions and measures of central tendency, were used to summarize survey responses. Bivariate relationships were tested for statistical significance utilizing χ² or t tests as appropriate for each comparison.

The response rate to the overall PHM survey was 36% (1103 of 3077). Nine hundred eighty-eight respondents were active physicians working primarily in inpatient settings. Of these, 198 identified as program leaders, with 32.8% identifying as division chief or director, 15.2% as section chief or director, 7.6% as program director, and 51.5% as medical director. Organizationally, 41.9% of programs were university-based, whereas 57.1% were community-based, with 25% at freestanding children’s hospitals, 34.4% at children’s hospitals within a larger hospital system, and 40.6% in pediatric units within a hospital. Of the 198 programs, one-half (50.5%) covered only community sites (community programs), 21.2% covered university sites (university programs), and another 21.2% covered both community and university sites (combined programs). Employment model had similar variability, with 19.7% university-owned, 48.7% hospital-owned, 20.2% in a contracted group, and 11.4% in other models (see Table 1).

From a staffing perspective, programs averaged 10.4 total FTEs across sites, which translated to an average of 14.9 individual employees. The average number of FTEs and of individual employees varied significantly (P < .001) by type of site, with totals lowest for community programs (8.8 FTEs, 11.5 employees), followed by university programs (10.5, 19.2), and highest for combined programs (16.2, 21.6). Most of the employees were physicians, with the overall average number of FTEs for advanced practice providers at 1.2. On average, 6.1 individuals per site were 100% clinical effort, whereas 8.8 were <100% clinical effort. Proportion working <100% clinical effort varied significantly (P < .001). This proportion was lowest for community programs, but even these sites reported that more than one-half were at <100%. Of those not working 100% clinical effort, the median individual %FTE dedicated to nonclinical activities was 20%, with those working at community programs having a lower median (20) than those working at university programs (25) or combined programs (25); this difference was not statistically significant. The majority of programs (87%) had hired 2 or fewer replacement clinicians over the past 3 years, with the median across all types of programs being 1.

Pediatric hospitalists provide coverage for numerous service lines in hospitals. Nearly one-third of university programs (30.8%) offered surgical co-management. All other service lines were covered by <20% of university programs, with 18.7% covering pediatric subspecialty units, 16.7% covering newborn nursery, and <10% covering Level 2 and 3 NICU, delivery services, PICU, pediatric care in the ED, inpatient rehabilitation, or urgent care. Community programs had a much more diverse service line profile, with 2/3 covering newborn nursery (67.2%) and over one-half offering pediatric consults in the emergency department (ED) (53.5%). A notable number of community programs provided surgical co-management (45.5%), delivery room services (39.4%), Level 2 NICU (32.3%), intermediate care on the general pediatric unit (29.8%), or pediatric subspecialty units (20.2%). Sedation was covered by 13.1% of university programs and 16.2% of community programs (Table 2).

The median annual hours for a 1.0 clinical FTE was 1849 across all types of programs. Notably, for university programs (n = 42), the median 1.0 FTE hours was lower than the community programs (n = 101) (1800 vs 1900 hours, P < .05). For combined programs (n = 44) the median was 1800. Across program type, nocturnists were used by the minority of programs (n = 27, 13.6%), and all nocturnists worked fewer hours per year than nonnocturnists, ranging from 83.7% to 89% of the total 1.0 FTE hours reported for nonnocturnists. Weekendists were uncommon across all program types, and they averaged a similar reduction in total annual hours. The median numbers of weekend shifts (defined as any shift on a Saturday or Sunday) were 27 for university programs, 34 for community programs, and 29 for combined programs. Considering 24-hour in-house coverage, 42.5% of university programs had this coverage, compared to 64.4% of community programs, and 64.3% of combined programs (P < .05) (Table 3).

With regard to staffing and support, directors were asked about patient caps, back-up systems, and coverage expansion. Patient caps and back-up activation were considered for inpatient and nursery teams, both resident-covered and nonresident-covered. Forty-six (23%) programs responded that they have patient caps, with the average patient cap across all types of programs for resident-covered inpatient teams higher than nonresident covered inpatient teams (16 vs 13). All program types reported median caps between 14 to 16 patients for resident-covered inpatient teams. No community programs had resident-covered nursery teams, and only 2 programs had noncovered nursery team caps; 3 university and 2 combined programs had noncovered nursery team caps.

Of the 198 programs responding, 30.8% reported having a back-up or jeopardy system for significant census increases. Just over one-half of university programs (51.3%) and combined programs (56.7%) had back-up or jeopardy systems, whereas only one-quarter (25.8%) of community programs did (P < .001). Census levels that activated back up were divided, similarly to patient caps, by resident-covered inpatient teams, noncovered inpatient teams, resident-covered nursery teams, and noncovered nursery teams. For inpatient floor coverage, all types of programs had higher census activation levels on resident-covered inpatient floor teams compared to noncovered teams, with activation levels across program types ranging from medians of 15 to 16 for resident-covered compared to 12.0 to 13.5 for noncovered teams.

Finally, 48.7% of university programs reported expanding inpatient coverage during busier seasons, with 18 as the median number of weeks that had added coverage, whereas 31.6% of community programs did (median 20 weeks), and 57.9% of combined programs (median 20 weeks) (P < .05).

This study continues to enhance our understanding of the current workforce models of pediatric hospitalists in the United States. Viewed in context of previous workforce studies in PHM, it helps create an updated picture of work hours, staffing needs, and services provided across practice sites. As a young and rapidly evolving specialty, it is essential to track the workforce and practice models of PHM so that it continues to be a desirable field of practice and one that has sustainable career options.

This PHM workforce study is, to our knowledge, the first that is inclusive of community- and university-based programs, while also being able to delineate between programs that cover only university sites, only community sites, and combined sites. Previous studies either looked only at one type of group or designated them by ownership model. This study also has the highest response rate of any previous PHM workforce studies, with 198 responding program leaders.

PHM is still predominantly staffed by physicians, and the median amount of clinical time worked is 80%. Although our data suggest that over one-half of all FTEs work <100% clinical effort, we were unable to distinguish part-time work from buyout for nonclinical roles. This proportion was higher for university programs (80%) or combined programs (69%), whereas community programs had just over one-half (56%) of their individual hires working <100% clinical effort. The average percent time in nonclinical responsibilities, among those <100% clinical, was similar across site types, though slightly lower among community programs. It is a positive sign that programs needed few replacement hires over the past 3 years, suggesting that there is some stability in staffing, although this could have been impacted by the pandemic, and associated hiring freezes.

As the field evolves, it is essential to get a picture of best practices for work hours, workload, and effective models. In the context of previous studies, our results support the emergence of a consistent median of annual work hours for an FTE. For university and combined programs, the median annual clinical hours for 1.0 FTE were 1800 (IQR 1650–1970) in our study. This is the same as noted in both the 2018 university study by Fromme et al⁵  and the most recent Society of Hospital Medicine survey in 2020.⁸  This study also confirms that community programs tend to work slightly more clinical hours annually with 1900, which is similar to the 1882 hours noted by Alvarez and colleagues in their study of community programs in 2019.⁶  This difference between university and community program work hours may reflect differences in census volumes, patient complexity, total compensation, employment model, additional expected uncompensated work, or geography. More studies need to be done to explore the differences and potential reasons for higher clinical work hours at community programs, especially as concerns for career sustainability and burnout are considered. Turnover may be a late sign that workload is too high, at either type of program.

Service line coverage variation was a notable finding of our study. The notion that PHM programs are considerably different from each other is not new. This study demonstrates just how different programs can be, and how university programs are generally different from community programs. When looking at service lines beyond standard inpatient unit coverage, community programs in our study covered newborn nursery, deliveries, Level II NICU, and sedation at similar levels to the groups in 2019 in the Alvarez study.⁶  Our responding programs covered Level III NICUs more than in 2019 (15.2% of sites vs 1%), as well as ED urgent care (15.7% covered ED and 7.1% covered urgent care vs 6% covered either in 2019). Conversely, a smaller percentage of community sites in this study provided ED consults (53.5% vs 89%).

It is difficult to compare university site service lines to previous studies, because previous studies have either not looked at that level of detail^4,5  or the data does not differentiate between community and university programs.⁸  Our data support that university programs have fewer service line responsibilities outside of the general inpatient unit, which is not surprising since there are often unique surgical and subspecialist teams, as well as pediatric emergency departments as opposed to emergency departments without pediatric coverage.

A few things stood out regarding workload beyond work hours and service lines. Seasonal expansion, census caps, and back-up are all methods for reducing workload burden and improving patient safety. Compared to the Fromme et al⁵  and Alvarez et al⁶  studies, more university and community programs had seasonal expansion of coverage (49% university and 32% community vs 32% university and 14% community). More university programs had census caps (53% vs 39%), whereas community program use of caps stayed relatively stable. Back-up was seen in a slightly lower percentage of university and community programs in our study, although in all it was available in ∼50% of university programs, and roughly 25% to 33% of community programs. And although few newborn nursery services have census caps, inpatient floor program caps in our study (medians 12–16) reflect the recommended census limits from previous studies.^6,9 

In summary, the data from this workforce study has provided an important reference point in the progression of workforce data in PHM. As a young specialty, it is important to benchmark work hours and models so that individual programs and program leaders can advocate for their groups. The majority of programs in our study were based in the community, with the most frequent hospital setting being a pediatric unit within a hospital, and nearly one-half being hospital-owned. Just over one-half of our respondents were medical directors, not division or section chiefs. As a field, we do not know how many PHM programs exist in the United States, although one previous project found that 65% of responding programs were considered community hospitals (C. Snow, personal communication, January 30, 2018). Workforce studies need to continue to look at programs both in aggregate for larger trends, but also through the lens of organizational setting.

Our study has several limitations. First, it was conducted one year into the SARS-CoV-2 pandemic, so it could have reflected trends impacted by the pandemic. Second, program leaders were asked to self-identify through a larger study, so there is no way to confirm that the primary leader completed the survey, although we asked only those who were in the lead position to respond. This could also have led to >1 response from an institution. We attempted to make the questions as clear and narrow as possible, so that respondents knew what to answer, but the variance in programs often makes questions not perfectly suited for each setting.

Current data on clinical work hours for pediatric hospitalists are consistent with recent, smaller studies, suggesting that the current national median for clinical work hours at university programs is 1800 annual hours for a 1.0 FTE. Hospitalists at community programs often work more clinical hours than university programs, and more commonly provide a broader range of service lines, including delivery services, as well as Level II and III NICU, intermediate care, and surgical co-management more frequently. Compared to previous studies, we found greater use of seasonal expansion at all programs, while university programs are more likely to have census caps. More studies are needed to explore the differences of community and university program work models.

The authors acknowledge the AAP Section on Hospital Medicine Division Directors Subcommittee, whose early brainstorming served as a foundation for this project.