Pediatric Medical Subspecialist Use in Outpatient Settings

Key Points Question How have rates of outpatient pediatric medical subspecialist use changed overall and by payer type? Findings In this cross-sectional study that used electronic health records and health plan claims data, annual rates of subspecialist use ranged from 8% (Medicaid beneficiaries) to 21% for primary care users within academic health systems. Annual rates significantly increased for children with commercial insurance, while rates decreased for children with Medicaid coverage. Meaning These findings suggest that differences in the use of pediatric medical subspecialists by type of payer may reflect differences in demand, health need, or access to care.


Introduction
Pediatric medical subspecialists provide medical care to children whose health problems are uncommon or atypical, have not responded well to therapy typically delivered in primary care settings, or require technical expertise such as advanced procedures for diagnosis or treatment. 1bspecialists serve an essential role in the health care system by providing cognitive and procedural consultative services to patients and primary care physicians and longer-term primary-specialty shared care for individuals with chronic conditions or conditions with high medical complexity, 2 although effectiveness of coordination across the primary-specialty care interface is uneven. 3,4Most pediatric medical subspecialists complete a pediatric residency and then receive additional fellowship training in a subspecialty.Other subspecialists such as pediatric dermatologists complete a residency (eg, dermatology) and then complete a pediatric specialization (eg, pediatric dermatology) during their fellowship.
Understanding whether the number and distribution of pediatric medical subspecialists are meeting the needs of the nation's children is challenging because of a paucity of evidence regarding their frequency of use.Available data that estimated rates of primary care physician referral and subspecialist use by children predate the year 2000. 1,5Perceptions of an undersupply of pediatric subspecialists stemming from long wait times for appointments for some subspecialties, long distances to access care, and insurance-related limitations on subspecialty access have been difficult to confirm due to lack of available data. 6,7 address this evidence gap, we conducted a study using 3 complementary data sources to quantify children's use of pediatric medical subspecialties from 2011 to 2021.Electronic health record data from large pediatric medical centers and administrative data from Medicaid and commercial plans were evaluated.This study was performed in support of a National Academies of Sciences, Engineering, and Medicine (NASEM) Committee on the Pediatric Subspecialty Workforce and Its Impact on Child Health and Well-Being that produced a consensus study report. 8The recently released report-funded by a coalition of sponsors-recommends actions to ensure equitable access to high-quality pediatric subspecialty care to advance the health and health care of infants, children, and adolescents.
Both the NASEM report and the present study focused on pediatric medical subspecialists, excluding pediatric surgery, anesthesiology, psychiatry, radiology, and pathology practitioners.Our overall goal was to characterize the rates and temporal patterns of outpatient pediatric medical subspecialty use and determine the most common types of conditions managed to inform action on strengthening the pediatric subspecialty system of care.A secondary goal was to contrast rates of pediatric subspecialists and internal medicine-and adult-trained (IMA) subspecialists for Medicaid beneficiaries.

Methods
The research reported in this repeated cross-sectional study used deidentified data and was designated as nonhuman subjects research and exempt from review and the requirement for The study sample consisted of patients with at least 1 outpatient visit to a general pediatrician that occurred from January 1, 2010, to December 31, 2021, for patients younger than 21 years.To reduce the number of patients with missing race and ethnicity data and to evaluate possible disparities in subspecialist use by race and ethnicity, we used the Centers for Disease Control and Preventionrecommended approach for a combined race and ethnicity variable in which ethnicity was used to define Hispanic category and race was used for non-Hispanic categories. 10Race and ethnicity were self-reported by parents and older youths.If the patient had public insurance (ie, Medicaid or Children's Health Insurance Program [CHIP]) at any time during a 1-year observation period, that was recorded as their payer status; otherwise, if they had commercial insurance, that was listed as payer status, and all others were listed as unknown or self-pay.

HIRD Dataset
The HIRD includes claims data submitted for payment for services to individuals enrolled in 14 commercial health plans that are part of Elevance Health (formerly Anthem Blue Cross and Blue Shield), as well as member enrollment and professional and facility claims.Medical claims are subject to the quality control, inspection, and validation procedures performed by the plans for payment processing.This analysis used enrollment files and fully adjudicated claims with service dates between January 1, 2011, and December 31, 2021.

T-MSIS Dataset
The T-MSIS is a national Medicaid and CHIP dataset from states, territories, and Washington, DC.The first available year of T-MSIS is 2016, and the study used data from January 1, 2016, through December 31, 2019.Other service files were used to identify clinicians providing outpatient evaluation and management services.Annual Provider files were used to match and fill in the missing National Provider Identifier when available.The final analysis included 44 states, Washington, DC, and Puerto Rico and excluded Arkansas, California, Delaware, Indiana, Minnesota, and Pennsylvania because of inadequate data quality.The T-MSIS database has known data quality challenges, which are analyzed and reported by the Centers for Medicare & Medicaid Services. 11A patient was included in the study sample if they were enrolled in Medicaid at any time during a given year and were younger than 19 years at the end of that year.The younger upper limit age cut point for the T-MSIS was chosen to reflect Medicaid eligibility criteria.

Pediatric Medical Subspecialties
The study examined use of 25 subspecialties in outpatient settings.This included 14 pediatric subspecialties for which the American Board of Pediatrics (ABP) provides board certification: adolescent medicine, pediatric cardiology, child abuse pediatrics, pediatric critical care medicine, developmental-behavioral pediatrics, pediatric emergency medicine, pediatric endocrinology, pediatric gastroenterology, pediatric hematology and oncology, pediatric infectious diseases, neonatal-perinatal medicine, pediatric nephrology, pediatric pulmonology, and pediatric rheumatology.Pediatric critical care medicine and neonatal-perinatal medicine were included in

JAMA Network Open | Pediatrics
Pediatric Medical Subspecialist Use in Outpatient Settings overall outpatient specialist use rates, but because these specialties provide so few outpatient services, they were excluded from specialty-specific calculations.Pediatric hospital medicine was excluded because of the study's focus on outpatient care.
We also included 5 subspecialties for which the ABP cosponsors board certification, though primarily administered by other specialty boards: hospice and palliative medicine, medical toxicology, sleep medicine, sports medicine, and pediatric transplant hepatology.An additional 6 subspecialties that are not sponsored or cosponsored by the ABP but are commonly recognized as pediatric medical subspecialties were also included: pediatric allergy and immunology, child neurology, obesity medicine, pediatric dermatology, clinical genetics, and pediatric rehabilitation medicine.The T-MSIS analysis also examined IMA medical subspecialties that corresponded to the included pediatric subspecialties, excluding those with no counterpart (eg, adolescent medicine, developmental-behavioral pediatrics, and neonatal-perinatal medicine).

Statistical Analysis
The primary rate calculation was the annual number of children with at least 1 completed visit to a subspecialist in an outpatient setting per population.Notably, this measure excludes patients for whom a referral to a subspecialist was made but the visit was not completed.Use rates excluded visits in emergency departments or inpatient settings.
Denominator definitions differed across the 3 data sources.The PEDSnet denominators included individuals with a visit to a general pediatrician during the index or prior year and therefore constituted a primary care use-based sample.The HIRD denominators required that individuals have at least 6 months of plan enrollment in a given year.The T-MSIS calendar year denominators required any duration of enrollment in Medicaid or CHIP.
The PEDSnet analysis included annual rates of patients seeing multiple types of subspecialiststhat is, having at least 1 visit with 2 or more of the 25 subspecialty types.The PEDSnet analysis also compared children insured by Medicaid and those with commercial insurance as well as differences by self-identified race and ethnicity.
To calculate rates of change over time, the aggregate data (denominator and numerator in each year) were fit as a function of year using a quasi-Poisson regression, which is a Poisson regression including an allowance for potential extra variance (overdispersion) in the data.The regression coefficient for year is the log of the mean rate ratio between consecutive years.The rate ratio was extracted as the exponential of the coefficient, and then the percentage change per year was the rate ratio minus 1 expressed as a percentage.For example, a coefficient of 0.18 is exponentiated to get a rate ratio of 1.20, which gets recorded as a mean annual percentage change of 20%.We computed 95% CIs on the annual changes in rate, and those that did not cross zero were deemed statistically significant.
To describe the most common health conditions managed by different subspecialties, we used diagnostic data from PEDSnet, combining all study years.Diagnoses were combined into a cluster if they were for the same condition (eg, all diagnoses for primary and secondary hypertension were combined into a single hypertension cluster).For each health condition, we computed the proportion of patients seen by the subspecialty who were assigned the code(s) for the condition at least once.Analyses were performed using R, version 4.3.2(R Project for Statistical Computing).

Results
In the study sample, there was marked growth in the size of the outpatient subspecialty sample among PEDSnet (academic medical centers) from a low level of 493 628 in 2011 to a high level of 858 551 in 2021 (Table 1).Denominators were stable for HIRD (commercial health plans) at
Both PEDSnet and HIRD showed a pandemic-related decline in use of subspecialists in 2020 with a return to baseline in 2021.There was a trend toward increasing subspecialist use in the HIRD decline (95% CI, −0.7% to −4.3%) per year; Hispanic patients and those identifying as multiracial each had a −1.9% per-year decline (95% CIs, −1.2% to −2.7% per year for Hispanic and −1.0% to −2.7% per year for multiracial patients); and non-Hispanic Black patients had a −0.5% decline (95% CI, -0.8% to 1.9%) per year.
We computed temporal trends in subspecialist use for 13 common pediatric subspecialties (Table 2).Across all 3 data sources, pediatric rheumatology was in the top 2 subspecialties for annual rate increases.Most of the subspecialties experienced some level of increase, although the absolute levels differed by data source.To address the apparent inconsistency between flat overall rates, particularly for PEDSnet and T-MSIS, and increasing individual subspecialty rates, we used PEDSnet data to compute the annual rates of patients seeing more than 1 type of subspecialist each year.
Across the study period, the rates of multiple use of subspecialists rose from 4.7% in 2011 to 6.4% in 2021.Pediatric hematology and oncology 0.9 (−0.2 to 1.9) 7.9 Using the T-MSIS data from 2019, we contrasted subspecialist use rates for 11 common subspecialties with both pediatric-and IMA-trained physicians (Table 3).Except for hematology and oncology, the absolute number of subspecialists with at least 1 child visit was higher for IMAcompared with pediatric-trained physicians.The IMA subspecialists more commonly provided care for adolescents rather than younger children.Rates of use of subspecialists were higher for pediatric subspecialists compared with their IMA counterparts except for allergy and immunology.
We identified the 3 most commonly managed health conditions for 13 pediatric subspecialties using PEDSnet data (Table 4).These 3 conditions accounted for about one-third or more patients for 8 of 13 subspecialties.Three conditions accounted for 72% of patients seen by pediatric allergists and immunologists and 50% seen by pediatric gastroenterologists.Many of the conditions across all the subspecialties were common health problems.

Discussion
Among children in this cross-sectional study, 8.6% of Medicaid beneficiaries, 10.4% of those commercially insured, and 21.3% of those whose primary care is received in academic health systems used pediatric medical subspecialty care each year.The rate estimates varied between data sources, which may be due to differences in access to subspecialty care by payer, geographic accessibility differences across regions of the US, type of health system, and a range of other factors that influence access and use of care.There was a small increase in rates of subspecialty use among children with commercial but not Medicaid insurance.These results do not provide any information on appropriateness of the subspecialist use, which means that it is unclear whether the temporal trend differences by payer are due to variation in demand, health needs, access, or some combination of these factors.
The PEDSnet academic medical center data demonstrate that these institutions have experienced an absolute increase in use for subspecialty care since 2011 due largely to more children receiving most of their care (primary and subspecialty) within their systems.Even though we restricted the PEDSnet sample to patients who obtain primary care from those health systems, children receiving care in academic medical centers may have higher complexity and thus more intense service use.Third, PEDSnet has devoted extensive effort to validating physician subspecialty, working with each member institution to examine and improve the quality of this data element.It is possible that the other 2 datasets had lower rates because some subspecialty care may have been billed using a generic pediatric medical group code rather than a specific subspecialty group practice.The HIRD subspecialty care data relied on the clinician taxonomy documented on the claim submitted for payment.The T-MSIS analysis relied on primary specialty in the National Plan and a Diagnosis data were obtained from visits made to subspecialists from 2011 to 2021 and thus reflect the health conditions as diagnosed by the subspecialists.
informed consent by the Children's Hospital of Philadelphia Institutional Review Board (IRB) for PEDSnet, the NASEM IRB for the Healthcare Integrated Research Database (HIRD), and the George Washington University IRB for the Transformed Medicaid Statistical Information System (T-MSIS).JAMA Network Open | Pediatrics Pediatric Medical Subspecialist Use in Outpatient Settings This study followed the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) reporting guideline.PEDSnet Dataset Electronic health record data in PEDSnet are obtained from hospital information systems and standardized to the Observational Medical Outcomes Partnership common data model. 9Participating institutions included Ann & Robert H. Lurie Children's Hospital of Chicago, Chicago, Illinois; Children's Hospital Colorado, Aurora; Children's Hospital of Philadelphia, Philadelphia, Pennsylvania; Cincinnati Children's Hospital Medical Center, Cincinnati, Ohio; Nationwide Children's Hospital, Columbus, Ohio; Nemours Children's Health, Wilmington, Delaware, and Orlando, Florida; Seattle Children's Hospital, Seattle, Washington; and Stanford Children's Health, Stanford, California.
approximately 5 million and for T-MSIS (Medicaid or CHIP) at approximately 35 million.The PEDSnet dataset included proportionally more infants, while the T-MSIS dataset had a greater share of older youth.

Figure .
Figure.Annual Outpatient Pediatric Medical Subspecialist Use Rates, 2011 to 2021 Abbreviations: HIRD, Healthcare Integrated Research Database; T-MSIS, Transformed Medicaid Statistical Information System.

Table 1 .
Characteristics of Study Samples a Abbreviations: EHR, electronic health record; HIRD, Healthcare Integrated Research Database; NA, not applicable; T-MSIS, Transformed Medicaid Statistical Information System.aUnless otherwise indicated, data are expressed as No. (%) of patients.bA total of 193 090 patients (11.8% of total) in the PEDSnet sample and 8 908 462 (25.6% of total) in the T-MSIS sample had missing race and ethnicity data.These patients were excluded from the race and ethnicity distributions.cA total of 196 553 patients (24.1% of total) in the PEDSnet sample had unknown payer data in 2019.These patients were excluded from the payer distribution.

Table 2 .
Mean Annual Change in Outpatient Subspecialist Use Rates for Selected Subspecialties by Data Source

Table 3 .
Use of Pediatric vs Internal Medicine-and Adult-Trained Subspecialists in T-MSIS Dataset a 12 other words, while the subspecialist use rate per 1000 children has not increased, the number of children receiving that care has increased dramatically.Moreover, in academic medical centers, the need for multiple subspecialties for the same child has risen, pointing toward greater medical complexity among patients served.Data from inpatient settings12indicate that the prevalence of complex diagnoses among children grew 45% from 2009 to 2019; most of these children also require outpatient services.
There are several reasons why the subspecialty use rates were highest in PEDSnet.First, the HIRD and T-MSIS datasets included children enrolled in health care plans, which included both users and nonusers of health care, while PEDSnet was a use-based cohort.Second, the HIRD and T-MSIS datasets included children who sought care in all types of settings, both community and academic institutions, while PEDSnet included children obtaining care from academic medical centers only.

Table 4 .
Most Commonly Managed Health Conditions by Pediatric Medical Subspecialty in Outpatient Settings in PEDSnet Dataset a