Specialties Providing Ambulatory Care and Associated Healthcare Utilization and Quality for Children with Medical Complexity

Abstract

Objectives:

Although children with medical complexity (CMC) have substantial healthcare needs, the extent to which they receive ambulatory care from primary care versus specialist clinicians is unknown. We aimed to determine the predominant specialty providing ambulatory care to CMC (primary care or specialty discipline), the extent to which specialists deliver well-child care, and associations between having a specialty predominant provider and healthcare utilization and quality.

Study Design:

In a retrospective cohort analysis of 2012–2017 all-payer claims data from Colorado, New Hampshire, and Massachusetts, we identified the predominant specialty providing ambulatory care for CMC <18 years. Propensity score weighting was used to create a balanced sample of CMC and assess differences in outcomes, including adequate well-child care, continuity of care, emergency visits, and hospitalizations, between CMC with a primary care versus specialty predominant provider.

Results:

Among 67,218 CMC, 75.3% (n=50,584) received the plurality of care from a primary care discipline. Body system involvement, age >2years, urban residence, and cooccurring disabilities were associated with predominantly receiving care from specialists. After propensity score weighting, there were no significant differences between CMC with a primary care or specialist “predominant specialty seen” (PSS) in ambulatory visit counts, adequate well-child care, hospitalizations, or overall continuity of care. Specialists were the sole providers of well-child care and vaccines for 49.9% and 53.1% of CMC with a specialist PSS.

Conclusions:

Most CMC received the plurality of care from primary care disciplines, and there were no substantial differences in overall utilization or quality based on the predominant specialty seen.

Introduction

Effective primary care is foundational to child health, yet primary care accounts for a small fraction of child-related healthcare expenditures in the United States.^1–3 Recognizing both the importance of primary care and gaps in access, quality, and equity, the National Academies of Sciences, Engineering, and Medicine (NASEM) recently published a report to guide high quality primary care delivery.¹ Much of the guidance outlined in this report aligns with principles of the patient-centered medical home, first introduced by the American Academy of Pediatrics in 1967 to support care for children with special healthcare needs.³ Given the increasing role of specialists in healthcare delivery, the report calls for research to determine the extent to which primary care is delivered by specialists and the effects of this delivery on access and quality.

Children with medical complexity (CMC) – defined as those with complex chronic diseases, functional limitations, and high levels of health system utilization – represent a vulnerable population that may disproportionately experience care fragmentation and poor healthcare quality.^4–8 Although CMC stand to benefit tremendously from high quality primary care given their complex healthcare needs, they may experience unique barriers to primary care access including financial insecurity, receipt of care from multiple specialists that may supplant a perceived need for primary care, and primary care clinician discomfort with medical complexity.^9,10 Additionally, implementation of medical home services and measuring health care utilization have been noted as research priorities for this population.^11,12 Among adults with multimorbidity, having a primary care clinician, instead of a specialist, as the predominant clinician providing healthcare has been shown to reduce healthcare costs and improve care coordination.^13–17 The extent to which CMC receive their ambulatory care predominantly from primary care clinicians as opposed to specialists may similarly be associated with healthcare quality, but research in this area is limited.^18,19

This study aimed to identify the predominant specialty providing ambulatory care to CMC (primary care versus specialty discipline), determine the extent to which well-child care was delivered by specialists, and describe associations between having a specialist predominant provider and measures of utilization and quality.

Methods

Data Sources

This retrospective cohort study analyzed all payer claims data (APCD) from Colorado (CO), New Hampshire (NH), and Massachusetts (MA), provided by the Center for Improving Value in Health Care in CO, NH Comprehensive Health Care Information System and Department of Health and Human Services, and MA Center for Health Information and Analysis, respectively. These states were selected to represent populations with varied rural-urban compositions and because their data can be used for research purposes. Datasets included professional and facility healthcare claims for children and adolescents who were insured by Medicaid or employee-sponsored commercial payers. Unique identifiers linked children who were insured by more than one payer. Each dataset spanned 5 years, including 1/1/2013–12/31/2017 for NH and MA and 10/1/2012–09/30/2017 for CO. The Dartmouth-Hitchcock Institutional Review Board deemed the study exempt from further review and informed consent. Study reporting followed the Strengthening the Reporting of Observational Studies in Epidemiology guidelines.²⁰

Eligibility Criteria

The first 3 years of APCD were used to identify the CMC cohort and to determine, for each CMC, the predominant discipline (primary care vs. specialty) providing ambulatory care. We examined healthcare utilization and quality during the subsequent two years to reduce endogeneity (in which the predictor variable is correlated with the error term) between the “predominant specialty seen” and our utilization and quality measures.²¹ CMC were identified using the Pediatric Medical Complexity Algorithm (PMCA),^22–24 applying previously published conservative APCD coding criteria.²⁵

We excluded those who didn’t have a ZIP code of residence in their state’s dataset (CO, MA or NH) during the study period, CMC >18 years before the end of the cohort creation period, those without ambulatory clinic encounters, and those who died during the cohort creation period. Participants were also required to have >12 months of eligible enrollment and/or claims in both the three-year cohort creation period and the two-year outcome period. To focus on community-dwelling CMC, we excluded those with >100 days of continuous hospitalization or long-term care during the study period.

Patient Attribution to Predominant Specialty Seen

The “predominant specialty seen” (PSS) was determined for each CMC, defined as the specialty of the clinician(s) with whom they had the most clinic visits during the cohort creation period. Visits were limited to those in outpatient settings with physicians, nurse practitioners, or physician assistants, as well as visits to primary care oriented ambulatory clinics such as federally qualified or rural health centers (Supplemental Table 1). Emergency department (ED) visits were excluded but urgent care clinic visits were retained. To focus on face-to-face visits and avoid double-counting encounters, healthcare claims with pathologists or radiologists were excluded, as were claims from anesthesiologists that occurred on the same day as claims from non-anesthesiologist clinicians. For visits that occurred at ambulatory care clinics where the specific clinician providing care couldn’t be confirmed, we limited to visits with Current Procedural Terminology (CPT) codes for office or other outpatient evaluation and management services.

Each clinician was assigned a primary specialty based on their National Provider Identifier (NPI) by linking it to the Centers for Medicare and Medicaid Services National Plan and Provider Enumeration System (NPPES).²⁶ Each clinician was assigned to their most specialized taxonomy listed in the NPPES. For example, a clinician listed with a primary taxonomy of pediatrics and a secondary taxonomy of pulmonology was assigned pulmonology as their specialty. Following assignment of each clinic visit to one of 41 potential specialties (Table 2), the proportion of a CMC’s visits with each specialty was determined, and the specialty with the largest proportion of visits was assigned as the PSS. The PSS was then categorized as either primary care or specialist. In the event of a tie, the PSS was assigned as primary care (n=2,212, 3.9% of cohort). A PSS of primary care included primary care oriented clinics as well the specialties of pediatrics, internal medicine, family medicine, general practice, and osteopathic manipulative medicine without additional subspecialty taxonomies. A specialist PSS included one of 35 unique specialties that were not categorized as primary care.

Table 2.

Primary Specialty Providing Ambulatory Care for Children with Medical Complexity.

	Specialty of PSS
	N	%
Primary Care
General Pediatrics1	39,743	56.84%
Family Medicine	5,554	7.94%
Primary Care Ambulatory Clinic2	4,086	5.84%
General Practice	835	1.19%
General Internal Medicine	353	0.50%
Osteopathic Manipulative Medicine	13	0.02%
Total	50,584	75.25%
Specialists
Adolescent Medicine	2,050	2.93%
Physical Medicine & Rehabilitation	1,503	2.15%
Neurology	1,435	2.05%
Psychiatry	1,320	1.89%
Cardiology	1,240	1.77%
Allergy & Immunology	1,149	1.64%
Otolaryngology	1,113	1.59%
Ophthalmology	990	1.42%
Orthopedic Surgery	989	1.41%
Developmental-Behavioral Pediatrics	874	1.25%
Endocrinology	686	0.98%
Hematology & Oncology	579	0.83%
Gastroenterology	574	0.82%
Pulmonology	329	0.47%
Dermatology	263	0.38%
Urology	245	0.35%
Anesthesiology	167	0.24%
Emergency Medicine3	166	0.24%
Nephrology	158	0.23%
Palliative Care	140	0.20%
General Surgery	105	0.15%
Neonatology	102	0.15%
Medical Genetics	95	0.14%
Infectious Diseases	93	0.13%
Plastic Surgery	69	0.10%
Obstetrics/gynecology	54	0.08%
Sports Medicine	51	0.07%
Rheumatology	40	0.06%
Neurosurgery	26	0.04%
Critical Care	13	0.02%
Other specialty4	16	0.02%
Total	16,634	24.75%

¹General pediatrics includes pediatric clinicians as well as medicine-pediatrics clinicians

²Primary care ambulatory clinics are shown in Supplemental Table 1; a specific clinician specialty was not available for these encounters

³Emergency department visits were excluded; these clinicians provided care in other ambulatory care settings

⁴Other specialties include: Sleep Medicine, Toxicology, and Pain Medicine; these specialties are not reported separately given data use agreement cell suppression rules.

^*The specialties of Child Abuse and Transplant Medicine were also represented in the data; however, these specialties did not result in the designation of “predominant specialty seen” for any children in the cohort.

Clinical and Demographic Characteristics

Baseline characteristics, measured during the 3-year cohort creation period, included age in years at cohort entry, gender, state of residence, primary payer (any Medicaid vs. no Medicaid) and rurality. Residential ZIP codes were linked to Rural-Urban Commuting Area codes to determine whether a child was urban- or rural-residing.²⁷ We reported body system flags used to classify children as CMC, and cooccurring disability was identified using the Children with Disabilities Algorithm.²⁸ Given cell suppression rules associated with our data use agreement, when gender was labelled as unknown, missing, or other, we combined these records with those labelled as male to create a non-female category. No other variables had missing or unknown values.

Healthcare Utilization and Quality of Care

Measures of ambulatory healthcare utilization included the total number of clinic visits, the distribution of these visits across specialties, and the number of unique clinicians (based on unique NPIs) seen by each CMC. Additionally, we determined the proportion of CMC who had claims for well-child care and vaccines, defined using previously established CPT codes and International Classification of Diseases, Tenth Revision Z-codes.²⁹

As a measure of healthcare quality, we calculated continuity of care (COC) using the Bice-Boxerman index; this measure has previously been validated for CMC and endorsed by the National Quality Forum.^4,30,31 The index takes into account the number of visits to a given clinician/NPI relative to the total number of visits; CMC cared for in settings with high turnover of clinicians will have lower COC, as will CMC cared for by many different clinicians in the same or different clinics. The COC ranges from 0 (indicating that a different clinician is seen for each visit) to 1 (indicating all visits were to the same clinician). Consistent with our approach to identify the PSS, we calculated this index separately for primary care and specialist NPIs, thereby generating measures of primary care and specialist continuity. For measure stability and in keeping with past research, each COC measure was limited to those with at least 4 visits during the observation period.^32–34 Keeping with the Continuity of Primary Care for Children with Medical Complexity quality measure, we defined adequate COC as having a COC >0.5.³⁵ Informed by the Bright Futures/American Academy of Pediatrics Recommendations for Preventive Pediatric Health Care, we defined receipt of adequate well-child care as >2 well-child visits in 24 months for CMC with 24 months of enrollment, and >1 well-child visits in CMC with <24 months of enrollment.³⁶ Included were ICD-10 codes starting with Z23 and 6 CPT codes (90460, 90461, 90471, 90472, 90473, 90474). Because the recommended schedule of well-child visits varies based on age, we conducted a sensitivity analysis limited to CMC >3 years, during which period a single well-child visit is recommended annually. Finally, as measures of potentially avoidable healthcare utilization, we examined ED visits not resulting in hospital admission or inter-facility transfer, and unplanned (non-elective) hospitalization at acute care hospitals.^37,38

Statistical Analysis

Given our large sample size, we followed guidance by Austin et al. and calculated standardized differences to determine statistically significant differences in baseline characteristics between CMC with a primary care and specialty PSS, where standardized differences >|0.1| were considered imbalanced.³⁹ Observing imbalance in these characteristics (Table 1), propensity score weighting was used to create a weighted sample of CMC, balanced on baseline demographic and clinical characteristics.³⁹ This was achieved using inverse-probability of treatment weighting (IPTW), defined as the inverse of the predicted probability of a given CMC’s PSS. The predicted probabilities were determined using logistic regression where the outcome was the PSS (primary care or specialty) and the covariates included all baseline characteristics shown in Table 1 (age, gender, primary payor, co-occurring disability, state, rurality of residence, body system involvement). Balance in the observable confounders was again assessed after IPTW using standardized differences. Creation of this balanced cohort ameliorated the need for subsequent regression analyses to identify differences in outcomes between CMC with primary care or specialist PSS.

Table 1.

Demographic and clinical characteristics of children with medical complexity with primary care versus specialty predominant providers of ambulatory care, unweighted and inverse probability treatment weighted.

	Unweighted							Weighted
							Standardized difference1			Standardized difference
	Total		Primary Care		Specialty			Primary Care	Speci alty
	N	%	N	%	N	%		%	%
Sample size	67,218	100.00%	50,584	75.25%	16,634	24.75%		50.05%	49.95%
Age
<2yrs	14,158	21.06%	11,754	23.24%	2,404	14.45%	0.226	21.09%	21.49%	−0.010
2–5yrs	15,829	23.55%	12,017	23.76%	3,812	22.92%	0.020	23.54%	23.48%	0.001
6–11yrs	26,474	39.39%	19,245	38.05%	7,229	43.46%	−0.110	39.37%	39.13%	0.005
12–15yrs	10,757	16.00%	7,568	14.96%	3,189	19.17%	−0.112	16.00%	15.90%	0.003
Gender
Female	27,234	40.52%	20,435	40.40%	6,799	40.87%	−0.010	40.52%	40.49%	0.001
Non-female2	39,984	59.48%	30,149	59.60%	9,835	59.13%	0.010	59.48%	59.51%	−0.001
Primary Payor
Any Medicaid	50,645	75.34%	38,350	75.81%	12,295	73.91%	0.044	75.37%	75.70%	−0.008
Cooccurring Disability	28,692	42.68%	20,612	40.75%	8,080	48.58%	−0.158	42.74%	42.70%	0.001
State
Colorado	14,632	21.77%	10,850	21.45%	3,782	22.74%	−0.031	21.80%	21.72%	0.002
Massachusetts	47,879	71.23%	36,197	71.56%	11,682	70.23%	0.029	71.22%	71.43%	−0.005
New Hampshire	4,707	7.00%	3,537	6.99%	1,170	7.03%	−0.002	6.98%	6.84%	0.006
Rurality
Urban core / Suburban	63,045	93.79%	47,400	93.71%	15,645	94.05%	−0.015	93.81%	93.85%	−0.002
Large / Small town rural	4,173	6.21%	3,184	6.29%	989	5.95%	0.015	6.19%	6.15%	0.002
Body System Disease Classification
Mental Health	41,956	62.42%	31,487	62.25%	10,469	62.94%	−0.014	62.48%	62.87%	−0.008
Progressivecondition	29,931	44.53%	22,384	44.25%	7,547	45.37%	−0.023	44.55%	44.35%	0.004
Pulmonary-respiratory	26,130	38.87%	20,403	40.33%	5,727	34.43%	0.122	38.91%	39.14%	−0.005
Neurological	19,819	29.48%	14,021	27.72%	5,798	34.86%	−0.154	29.55%	29.52%	0.001
Musculoskeletal	8,603	12.80%	5,820	11.51%	2,783	16.73%	−0.150	12.85%	12.82%	0.001
Cardiac	6,053	9.01%	4,195	8.29%	1,858	11.17%	−0.097	9.10%	9.34%	−0.008
Genetic	4,923	7.32%	3,422	6.76%	1,501	9.02%	−0.084	7.44%	7.72%	−0.011
Endocrinological	4,339	6.46%	2,930	5.79%	1,409	8.47%	−0.104	6.52%	6.56%	−0.002
Metabolic	4,318	6.42%	3,337	6.60%	981	5.90%	0.029	6.45%	6.61%	−0.006
Gastrointestinal	4,098	6.10%	2,792	5.52%	1,306	7.85%	−0.093	6.22%	6.45%	−0.009
Genitourinary	2,711	4.03%	2,040	4.03%	671	4.03%	0.000	4.05%	4.11%	−0.003
Immunological	2,219	3.30%	1,487	2.94%	732	4.40%	−0.078	3.35%	3.42%	−0.004
Malignancy	2,000	2.98%	1,231	2.43%	769	4.62%	−0.119	3.01%	3.04%	−0.002
Hematological	1,687	2.51%	1,082	2.14%	605	3.64%	−0.090	2.55%	2.53%	0.001
Renal	1,337	1.99%	897	1.77%	440	2.65%	−0.059	2.01%	2.05%	−0.003
Other3	5,166	7.69%	3,457	6.83%	1,709	10.27%	−0.123	7.76%	7.88%	−0.004

¹Statistically significant differences based on standardized differences <−0.1 or >0.1 are shown in bold.

²Those with unknown gender, n=51 (<0.1%), were combined with male to avoid reporting small cell sizes per the data use agreement.

³Otological, ophthalmological, otolaryngological, craniofacial, and dermatological.

We identified demographic and clinical characteristics associated with having a specialty (versus primary care) PSS using logistic regression; the associated chi-squared statistic was evaluated as a measure of predictive importance of the given characteristic in determining the PSS. To determine associations between the PSS and measures of healthcare utilization and quality, standardized differences of <−0.1 or >0.1 in the IPTW cohort were considered statistically significant.

All analyses were performed using SAS software, Version 9.4 with the logistic regression modeling run using the LOGISTIC procedure and weighting and standardized differences with the PSMATCH procedure.⁴⁰

Results

Characteristics of the Cohort

Among the 67,218 CMC meeting study eligibility criteria (Supplemental Figure 1), the majority (75.3%) received the plurality of their ambulatory care from a primary care discipline (Table 1). In the unweighted comparison of baseline patient characteristics between CMC with a primary care and specialist PSS, we saw statistically significant differences (|standardized differences| > 0.1) in patient age (with CMC < 2 years significantly more likely to have a primary care PSS), cooccurring disability (more likely to have a specialist PSS), and several of the body system disease categories. The PSS distribution across specialties is shown in Table 2. The mean number of study-eligible months in the outcome period was 22.1 months (standard deviation=3.3) for both groups.

After IPTW, all baseline covariates and clinical measures were balanced across the levels of the PSS. About 41% of the weighted sample was female, 76% were insured by Medicaid, 43% had a cooccurring disability, and 94% were urban-residing. Among the most frequent chronic conditions in our cohort, 62.4% had a PMCA-defined mental health condition, 44.5% had a progressive condition, and 38.9% had a pulmonary-respiratory condition.

Factors Associated with Predominant Care by a Specialist

Several demographic and clinical characteristics were associated with having a specialty PSS. As shown in Figure 1a, the largest effect sizes associated with having a specialty PSS were seen among children and adolescents aged 6–15 years relative to those <2 years, and in those with a hematologic condition or malignancy. Cooccurring disabilities and being urban-residing were also significantly associated with having a specialist PSS. Factors with the largest χ² statistics (Figure 1b) similarly included age >2 years, body system involvement (e.g., musculoskeletal condition), and cooccurring disability.

Figure 1. Demographic and Clinical Characteristics Associated with Having a Specialty vs. Primary Care as the Predominant Specialty Seen (PSS) for Ambulatory Care Visits: (a) Odds Ratio for Specialist relative to Primary Care PSS, (b) χ² statistic indicating the predictive importance of the given factors.

*Otological, ophthalmological, otolaryngological, craniofacial, and dermatological.

Healthcare Utilization & Quality

Although there were differences in the total number of ambulatory care visits in the unweighted cohort between CMC with a primary care versus specialist PSS, in the IPTW cohort there were no significant differences observed (Table 3). However, we did observe differences in where CMC received their care. CMC with a primary care PSS had more visits to primary care clinicians than those with a specialty PSS (7.9 vs. 3.9; std diff = 0.406) and most of this care was with pediatric primary care clinicians. We see the inverse of this pattern when looking at specialist visits; CMC with a specialty PSS had a greater number of specialist visits than those with a primary care PSS (10.9 vs. 5.1; std diff = −0.233). The observed differences were primarily driven by differences in visits to medical specialists (8.3 vs. 3.4; std diff = −0.217) and multi-specialty groups (1.44 vs. 0.76; std diff = −0.101), with the rate of visits to surgical specialists not statistically different by PSS.

Table 3.

Healthcare utilization and quality of care during the two-year outcome period for children with medical complexity with primary care versus specialist predominant specialty seen, unweighted and inverse probability treatment weighted.

	Total		Primary Care		Specialty		Standardized difference1	Primary Care		Specialty		Standardized difference1
	N / Mean	% / StD	N / Mean	% / StD	N / Mean	% / StD		N / Mean	% / StD	N / Mean	% / StD
N	67,218		50,5 84	75.25%	16,634	24.75%
Number of ambulatory care visits	13.44	14.16	12.76	12.09	15.49	18.98	−0.172	12.98	14.24	14.87	37.00	−0.067
# visits with primary care	6.91	7.18	7.84	7.45	4.07	5.37	0.581	7.85	8.65	3.94	10.51	0.406
# visits with pediatric primary care	4.98	5.91	5.68	6.15	2.84	4.48	0.529	5.69	7.14	2.75	8.81	0.367
# visits with non-pediatric primary care	1.93	4.68	2.16	5.12	1.23	2.85	0.223	2.16	5.91	1.19	5.60	0.168
# visits with a specialist	6.53	11.43	4.92	8.03	11.42	17.31	−0.482	5.13	9.52	10.93	33.87	−0.233
# visits with medical specialist	4.56	10.07	3.21	6.48	8.67	16.11	−0.444	3.35	7.66	8.32	31.58	−0.217
# visits with surgical specialist	1.05	2.33	0.98	2.13	1.28	2.86	−0.122	1.02	2.52	1.16	5.53	−0.033
# visits with multi-specialty clinic	0.92	3.5 3	0.74	3.1 8	1.47	4.3 9	−0.193	0.76	3.7 8	1.44	8.6 8	−0.101
Number of unique providers of care	4.91	3.80	4.91	3.80	4.94	3.82	0.008	5.00	4.50	4.67	7.30	0.054
Continuity of care (COC)2
Adequate Continuity (n/% COC > 0.5)	15,082	22.44%	11,402	22.54%	3,680	22.12%	0.010	14779	21.97%	16227	24.16%	−0.052
Primary Care	25,770	38.34%	20,633	40.79%	5,137	30.88%	0.208	27337	40.65%	206 28	30.71%	0.209
Specialists	21,066	31.34%	14,846	29.35%	6,220	37.39%	−0.171	19927	29.63%	25281	37.64%	−0.170
COC (0–1 scale)	0.353	0.299	0.353	0.300	0.352	0.293	0.005	0.348	0.344	0.369	0.601	−0.042
Primary Care COC	0.522	0.366	0.521	0.359	0.522	0.394	−0.002	0.520	0.414	0.526	0.786	−0.010
Specialists COC	0.496	0.378	0.501	0.391	0.486	0.342	0.040	0.492	0.452	0.508	0.688	−0.028
Received a vaccine	45,739	68.05%	34,797	68.79%	10,942	65.78%	0.064	46175	68.65%	44569	66.35%	0.049
Adequate well child visits3	47,057	70.01%	36,081	71.33%	10,976	65.99%	0.115	47718	70.95%	45229	67.34%	0.078
Any ED visits not resulting in hospitalization Acute care hospitalizations	32,833	48.85%	24,961	49.35%	7,872	47.32%	0.040	33003	49.07%	32750	48.76%	0.006
Any	6,60 5	9.8 3%	4,53 0	8.9 6%	2,07 5	12.47%	−0.114	6362	9.4 6%	753 2	11.20%	−0.057
Elective	1,718	2.56%	1,081	2.14%	637	3.83%	−0.100	1562	2.32%	2191	3.26%	−0.057
Non-elective	5,545	8.25%	3,862	7.63%	1,683	10.12%	−0.087	5419	8.06%	6150	9.16%	−0.039

¹Statistically significant differences based on standardized differences <−0.1 or >0.1 are shown in bold.

²Limited to those with >4 ambulatory care visits during the outcome period (n=65,424for overall, 59,458for Primary Care, 51,987for specialist).

³Receipt of adequate well-child care defined as >2 well-child visits in 24 months for CMC with 24 months of enrollment, and >1 well-child visits in CMC with <24 months of enrollment.

After IPTW, CMC with a primary care PSS were more likely to have adequate continuity of care (COC > 0.5) with primary care clinicians (40.7% vs. 30.7%; std diff = 0.209) than CMC with a specialist PSS (Table 3). Conversely, CMC attributed to primary care were less likely to have adequate continuity among their specialist clinicians than those attributed to specialists (29.6% vs. 37.6%; std diff = −0.170). However, there were no significant differences in overall CoC between CMC with a primary care or specialty PSS.

While the proportion of CMC receiving adequate well-child care during the follow-up period was not significantly different among CMC with a primary care PSS compared to those with specialty PSS (71.0% vs. 67.3%; std diff = 0.078), we observed differences in where CMC received this care (Figure 2). Among CMC who received well-child care, in the IPTW cohort, those with a primary care PSS received the majority of their well-child care solely from primary care clinicians (83.7%), while specialist clinicians provided well-child care for 44.8% of CMC with a specialty PSS (std diff = 0.887). A small percentage of CMC received well-child care in both primary care and specialist settings (primary care PSS 4.7% and specialty PSS 5.3%). Although the proportions of CMC who received vaccines were similar in the primary care and specialty PSS cohorts, vaccines were significantly more likely to be administered by specialists for CMC with a specialty PSS, and by primary care clinicians for CMC with a primary care PSS.

Figure 2. Distribution of location of receipt of (a) well-child care* and (b) vaccines during two-year outcome period by predominant specialty seen.**.

*excludes 8,312 without a well-child visit during the outcome period

**excludes 21,479 without a vaccine during the outcome period

Other measures of quality and utilization were statistically balanced across the PSS status in the IPTW samples. In our sensitivity analysis examining whether the proportion of CMC receiving adequate well-child care during the follow-up period differed in the restricted cohort of CMC aged 3–17 years, no significant differences were observed (primary care PSS: 68.8% versus specialty PSS 65.6%; std diff=0.068).

Discussion

In this analysis of all-payer claims data from CO, NH, and MA, we found that three-quarters of CMC had the plurality of their ambulatory care provided by a primary care discipline, most often general pediatrics. Age was the demographic factor most strongly associated with having a specialty PSS, with type of chronic condition and co-occurring disability also strongly associated with the PSS. Adolescents 12–15 years of age were more likely to have a specialty PSS; the larger role of primary care clinicians in ambulatory care delivery for younger CMC is not altogether surprising given the higher number of well-child care visits recommended during early childhood. Accounting for baseline demographic and clinical characteristics, we observed no significant differences in healthcare quality between CMC with a specialty versus primary care PSS, with the exception of discipline-specific continuity of care measures. These findings differ substantially from previous research applying similar methods with adult populations, in which having a primary care “predominant provider” was associated with better quality and lower costs.^13–17 Substantial differences between pediatric and adult medical complexity, and between pediatric and adult systems of care, may explain these disparate findings.

In outlining recommendations to strengthen access to high quality primary care, the 2021 NASEM report called for research to evaluate the extent to which primary care is delivered by specialists, and the effects of such healthcare delivery on access and quality.¹ Although the primary care patient-centered medical home is widely endorsed by professional societies, a national survey of primary care pediatricians found that more than 40% believed subspecialists provided the best medical home for CMC.⁹ We found no significant differences in receipt of adequate well-child care between CMC with a primary care or specialist PSS, but did observe substantial differences in where this care was provided. These results are consistent with past research showing that routine or preventative visits by known patients are the most frequent type of visit to pediatric specialists.⁴¹ There were no significant differences in other established measures of healthcare quality between CMC who received the plurality of their ambulatory care from specialist or primary care disciplines, suggesting that the PSS is not a key factor contributing to these outcomes.

Past work, including caregiver self-reported surveys and analyses of health systems data, has demonstrated suboptimal continuity of care for children with complex and/or chronic conditions.^4,42,43 A threshold of 0.5 on the Bice-Boxerman COC index has been endorsed by the Pediatric Quality Measures Program as a measure of adequate continuity of primary care for CMC.³⁵ By this definition, we found that only one-third of CMC had adequate continuity with a primary care clinician, and that a significantly greater of proportion of CMC with a primary care PSS had adequate primary care continuity. In contrast, patterns of specialist continuity were the inverse. Taken together with observed patterns of well-child care provision, it appears as though specialists are assuming central roles in the provision of primary care some CMC.

The results of this study should be interpreted in the context of several strengths and limitations. The large sample size and inclusion of both Medicaid- and commercially-insured CMC is a strength, although analysis of APCD from 3 states is not nationally representative and those without insurance are excluded. We observed a higher proportion of urban-residing children in our data compared to the general US child population, which stems from the large representation of Massachusetts residents. As a result, the relative role of specialists in providing care to CMC may be over-estimated. Importantly, we were unable to determine why some patients predominantly visited specialists or whether clinician preference, family preference, or patient needs were key drivers of care patterns. Our measures of quality and utilization reflect only those services covered by participating payers, which may have underestimated healthcare utilization. Relatedly, the quality measures are neither disease- nor age-specific and the number of measures that can be evaluated using APCD is limited. Our measures used to evaluate traditional aspects of primary care delivery, including vaccine administration and well child care, were not intended to determine the quality or appropriateness of this care, but to identify the disciplines providing this care for CMC. Additionally, we cannot assert a causal relationship between PSS and resource use in this observational study. While we strove to minimize the confounding effects of CMC observable characteristics, some CMC may have visited more specialists and had higher utilization because of characteristics that we could not observe. Finally, we were unable to determine the proportion of specialist care that was provided in the context of multidisciplinary complex care programs, increasingly prevalent at academic medical centers, and patterns of care may have changed since 2017.^44,45

In conclusion, the majority of CMC received the plurality of their care from a primary care discipline, with older age and type of complex condition most strongly associated with predominantly receiving care from specialists. This study also shows that specialists frequently take on traditional primary care roles, such as vaccine administration and well-child care, for a substantial fraction of CMC. We observed few differences in healthcare utilization and quality based on the predominant specialty providing care. Going forward, it will be important to better understand how primary care clinicians and specialists balance care for CMC over the life course, what models of care optimize quality and patient outcomes, and to develop more robust measures of quality for this population.

What’s New?

Children with medical complexity (CMC) often require subspecialty care. This study demonstrates that most CMC receive the plurality of ambulatory care from a primary care discipline (versus a specialty), with no significant differences in healthcare utilization or quality.

Abbreviations

APCD

all-payer claims data

CMC

children with medical complexity

CO

Colorado

COC

continuity of care

CPT

Current Procedural Terminology

ED

emergency department

IPTW

inverse-probability of treatment weighting

MA

Massachusetts

NASEM

National Academies of Sciences, Engineering, and Medicine

NH

New Hampshire

NPI

National Provider Identifier

NPPES

National Plan and Provider Enumeration System

PSS

predominant specialty seen