Challenges in Studying the Pediatric Hospital Medicine Workforce

The coining of the term “hospitalist” in 1996 started the exciting development of parallel subspecialties within adult and pediatric medicine. Pediatric Hospital Medicine (PHM) programs have exponentially grown since then, and PHM was officially recognized by the American Board of Medical Specialties in 2016. One of the hallmarks of PHM as a field is the broad clinical scope that hospitalists cover—including but not limited to acute care wards, intensive care units, emergency departments, delivery rooms, sedation services, and within many other subspecialties such as hematology/oncology and cardiology. Pediatric hospitalists in the United States work across the gamut of hospital sizes and settings (rural, suburban, urban). Workforce surveys have traditionally helped define clinical practice patterns, staffing models, and compensation. As the field of PHM grows, we anticipate they will additionally be critical in evaluating disparities within the workforce, investigating the currently projected increased demands on pediatric hospitalists with Accreditation Council for Graduate Medical Education changes to pediatric residency curricula, closures of smaller inpatient pediatric programs, and the culmination of the practice pathway for PHM board certification. Continued definition of the scope of PHM as a field and the experiences of those in it are critical. However, foundational challenges exist in the identification of all pediatric hospitalists, resulting in workforce studies with significant limitations in their methodological approaches.The largest challenge in surveying the PHM workforce stems from the lack of an accurate denominator that includes most, or all, pediatric hospitalists. Not one list exists of all pediatric hospitalists, or even PHM programs, throughout the country. In other subspecialties such as cardiology or gastroenterology, nearly all subspecialists interface with the American Board of Pediatrics for their subspecialty certification; however, this does not provide a denominator for all of PHM, as not all pediatric hospitalists are board eligible or certified. Three professional societies exist for PHM: the American Academy of Pediatrics (AAP) Section on Hospital Medicine (SOHM), the Academic Pediatric Association (APA), and the Society for Hospital Medicine (SHM). Although membership in one or more of these societies is common, it is impractical to capture how many practicing hospitalists exist in the United States who are not part of one of these groups.A second compounding challenge is that, historically, there has been suboptimal engagement from the target population in workforce study research. Many published studies have response rates of 50% or less, whereas others are unable to determine a response rate given the lack of denominator.Although prior workforce studies have provided integral snapshots into PHM as a field, the lack of a denominator and limited engagement raise concerns surrounding bias, accuracy, generalizability, and precision of data obtained. Here, we review prior studies and lessons learned and offer potential approaches to future workforce research.Recognizing these challenges, there have been several approaches and ideas to reach the broadest number of hospitalists. Perhaps the most time-consuming and thorough attempt was sponsored by the AAP SOHM and conducted under the leadership of Clota Snow, MD, nearly a decade ago. Dr Snow and her team used a list of hospitals from the American Hospital Directory and then attempted to contact each of them either through personal contacts or contacting the hospital directly (by phone, email, or fax, with contacts including the charge nurse, unit secretary, or pediatric acute care unit). From this, they contacted over 7000 hospitals and created a database of 826 hospitals with PHM services (personal communication with Clota Snow, MD). This was a Herculean effort, landmark for our field, and unfortunately, outdated within a year. Future attempts to use that list by the SOHM resulted in countless undeliverable emails.Many surveys have used the AAP SOHM’s listservs, which include a main listserv and multiple subcommittee listservs (Advanced Practice Providers, Community Hospitalists, Division Directors, Global Health, LGBTQIA+, Med-Peds, Neonatal Hospitalists, PHM Educators, Provider Wellness, Surgical Care, Women in PHM). Listserv members at the time of prior surveys did not have to be members of either the AAP or SOHM to join the mailing lists; this changed in March 2025. As of January 2025, the main listserv included 4130 members (personal communication with Dana Bennett-Tejes, MA). The recruitment strategies in these surveys are shown in Table 1, along with response rate data. Response rates, even when able to be calculated, have overall been limited.As we consider workforce studies, there are 2 foundational groups often considered as potential recipients: division (or group) directors and individuals in PHM. For division directors, several approaches have been used. Some studies have used the AAP SOHM’s Division Directors listserv, but this group is a self-selection of leadership from PHM programs.1,2 Some of its limitations include not representing all institutions, not capturing all hospitalist role types within a single institution, and having multiple representations from a single institution. This listserv has attempted to establish a database of division directors by asking listserv participants to add their names and information to an online repository. As of February 2025, this database included 165 programs (personal communication with Rachel Peterson, MD), but it is unclear how many programs are still missing from this database and/or if any leadership has changed. Allan et al used a purposeful sample of university-based programs through a list of Liaison Committee on Medical Education–accredited medical schools and by performing online searches of division leadership, but this sampling strategy only provided data from university sites.3 Although university sites commonly have partnerships with community sites, this still threatens to miss a significant piece of the community-based workforce. Recently, Osburn and colleagues used a combination of both the Division Directors’ listserv and recruitment at a national meeting.4 Although this does have a denominator (in the 300s normally), the methodology may result in more than one respondent per program or, more likely, miss programs whose leaders are not active on these listservs or present at the national meeting.For surveying individuals, prior studies used the listservs, convenience samples solicited through personal connections for qualitative work, or a convenience sample of program leaders to distribute to members of their programs.5,6 The AAP SOHM society-sponsored workforce studies accessed the membership lists and the listservs of the 3 professional societies; 2 subsequent workforce studies then used the respondents from the 2022 study as their own study sample.7–9 By leveraging the AAP SOHM, APA, and SHM membership lists in addition to listservs, these studies captured the largest denominators thus far of PHM workforce research. They also used professional survey development and distribution services, allowing for unique IDs to be assigned to all emails. These lists give the promise of a “holy grail” list of programs and individuals, but several issues remain. First, the societies are reticent to share their member lists with researchers in order to protect members from survey fatigue. Additionally, these lists are not inclusive of pediatric hospitalists who are not members of the professional societies. Even the largest SOHM-sponsored study had a denominator of just under 3000, which is smaller than the estimated 3000 to 4400 pediatric hospitalists in the United States and only barely higher than the 2537 who are board certified in PHM as of 2022, indicating pediatric hospitalists are still missing from this sampling methodology.10,11Ultimately, these studies have all done what they can within the restrictions of a specialty that does not have a primary list of all members or programs. What can be done next? At the programmatic level, the most obvious solution (and often requested) is to create a list of all pediatric hospitalist programs. Yet, the prior work by Dr Snow suggests that this is a difficult task to achieve and nearly impossible to sustain. Continued maintenance of the Division Directors database may be a reasonable option to identify PHM programs, but there needs to be investment into keeping it current and outreach in obtaining information from programs that are currently missing. It is very likely that this current list overrepresents university-based PHM groups compared to the field at large. Alternative sampling frames such as PHM fellowship programs or the Children’s Hospital Association’s Advancing Research, Technology, and Innovation Initiative carry the advantage of known denominators, although may only be indirectly able to answer questions regarding the PHM workforce.With the continuing advances of artificial intelligence, another consideration is to use machine learning tools to identify and extract contact information for pediatric hospitalists from hospital websites, institutional directories, or professional networking sites.Regarding individuals, the largest list right now of those who identify as pediatric hospitalists is through the AAP SOHM. Although it is understandable that they do not want to give out that membership list freely, they could consider following the example of the Association of Pediatric Program Directors’ Surveys Enhancing Educator Knowledge Network, which reviews and approves surveys to go to its membership. SHM’s annual survey, State of Hospital Medicine Report, is an example of a repeated rigorous analysis, although limitations for our context include providing data by group rather than individual members of the workforce, a smaller pediatric sample size, a limited response rate, and restricted accessibility of results to SHM members. This represents an area of potential collaboration across the 3 professional societies in PHM. These societies have the precedent of shared hosting of the annual PHM national meeting and leverage the strengths of each society toward meeting the needs of hospitalists at large. Leaders of the professional societies previously proposed a PHM dashboard, including descriptive data on each PHM group’s workforce.12 This dashboard was not implemented but shows a shared value across the professional societies for tracking this information. Marek et al noted that the “work of repeated [workforce] surveys” can considerably be transitioned from “academy inquiry to professional society service,” supporting that these societies can and should consider taking on the responsibility of reproducible, rigorous workforce analysis.13The AAP SOHM has experience developing research networks, such as the Pediatric Acute and Critical Care Quality Network, formerly known as the Value in Inpatient Pediatrics Network, which develops and leads multisite studies on the application of evidence-based practice through toolkit dissemination, data sharing, and research support. The AAP SOHM, alone or in conjunction with the other societies, could spearhead the development of such a research network, soliciting programs to not only join but also to allow their teams to participate in various studies. Notably, and not to be underestimated, any endeavor including professional societies as a vehicle of identifying hospitalists may exacerbate the existing reality of exclusion, as a proportion of hospitalists, because of personal or systemic reasons, are not engaged within a professional society. Such an endeavor would have to be highly intentional about being inclusive of various program types, sizes, and geography as well as board certification status and clinical practice settings. Additionally, although the funneling of workforce studies may improve the rigor of data collection, care should ensure that this does not lead to a funneling of researchers allowed access to investigate the PHM workforce.Another potential approach to future workforce surveys could be the use of a panel survey that involves gathering workforce data from the same sample of individuals over several years. Although efforts to have the sample of hospitalists as representative as possible might be extensive, the advantages are numerous. Panel surveys can study changes within individuals or groups and are particularly useful in understanding individual change and causal relationships such as labor market attitudes, activities, and dynamics—all areas of research yet untapped for the PHM workforce. This methodology, however, requires continued follow-up with the panelists to keep them engaged and minimize panel attrition.Increased engagement, including both diversified and targeted recruitment methods, is fundamental to accurately reflect the workforce at large. This is most feasible only through increased funding for this research. Thus far, participant incentives have most commonly been limited to access to data (for division directors) and goodwill (for individuals), with only 1 study in our sample including financial incentives for participants. Personalized invitations, multimodal reminders for study completion, and formal survey administration tools can help increase engagement. Funding can additionally help support external validation of participants’ responses, such as cross-checking with hospital credentialing departments. Although we noted earlier that professional societies within PHM can spearhead workforce study efforts, these societies can also offer competitive grant funding for interested individuals to elevate the rigor of studies conducted.Lastly, although limitations will always exist for any workforce study, rigor in survey construction and reporting allows for the best interpretation of obtained data. Consultation with, and inclusion of, experts in survey methodology within the research team can help ensure methodological soundness and feasibility for widespread participation. Careful selection of sampling methodology should include intentional evaluation of who may be systematically excluded. Differences between hospitalists who are systematically excluded from the sample frame to those in it can potentially lead to bias as much as nonresponding sample members do. Some information about those excluded generally allows researchers to gauge if bias might be present in related survey statistics and which direction it might take. A frequently used solution that may minimize the systematical exclusion of hospitalists includes the use of a multiple sampling frame design which then necessitates the identification of duplicate listings. Relevant information on all sample members should ideally be included to assess nonresponse bias. Although not all nonresponse is biasing for survey statistics, it is important to understand under what conditions bias occurs and therefore which information to collect from all sample members. Information that is most useful is the one that is hypothesized to be causally linked to the survey content and the respondents’ answers as well as the decision to respond to the survey. Because this information is rarely collected for a sampling frame in advance of a survey study in question, nonresponse bias studies often rely on comparing demographic and other related information between respondents and nonrespondents.14 This information has to usually be requested before the survey, so considerations regarding nonresponse bias should be made before data are collected. Other methods assessing potential nonresponse bias include a follow-up survey among nonrespondents focusing only on the key questions or analyzing the differences between early and late responders to the survey. Thoughtful survey construction with inclusive language and answer options to fully reflect the breadth of PHM are essential, in addition to formal testing through cognitive interviews prior to study administration. Limiting survey questions to essential length and unbiased presentation can help with both accuracy and engagement. Journals with large pediatric hospitalist audiences have, appropriately, been home to most prior workforce studies. Author guidelines and peer review processes within these journals can help identify best practices for workforce survey research. Recommendations can include the inclusion use standardized reporting guidelines, such as the Checklist for Reporting Results of Internet E-Surveys or the Consensus-Based Checklist for Reporting of Survey Studies, to optimize reporting quality.15,16Despite these challenges, it is imperative that workforce studies in PHM continue even when the methodology is imperfect. They are foundational to the growth and sustainability of the field and provide benchmarking data for both upcoming and existing programs.17 Interpreting these studies with caution knowing their limitations still adds integral knowledge valuable to hospitalists, programs, and hospital administration groups. It is time for the subspecialty to think outside the current box on how we can thoughtfully capture and represent hospitalists who may be excluded with the present sampling methodologies, consider partnerships with our professional societies to optimize our sampling frames, and increase engagement and rigor through funded research. Onwards!