Timing of High-Risk Asthma Specialist Enrollment and Severe Asthma Exacerbations in Children

Abstract

Background:

Although specialist care reduces acute healthcare utilization among children with uncontrolled asthma, the timeliness of enrollment and consequences of delayed specialty care remain poorly characterized.

Objective:

To characterize correlates of delayed enrollment in specialty asthma care and evaluate whether pre-post changes in childhood severe asthma exacerbation (SAE) incidence differed by patient characteristics.

Methods:

We conducted a pre-post, quasi-experimental longitudinal cohort study of 229 children (<11 years) with moderate-to-severe asthma enrolled in a high-risk asthma (HRA) clinic between January 2010 and December 2021, with follow-up through 2024. Early childhood asthma risk burden was quantified using predefined risk factors. Cox proportional hazards models examined correlates of delayed enrollment. Piecewise generalized linear mixed-effects models estimated changes in SAE odds before and after HRA enrollment, adjusting for demographic characteristics and time-varying treatments.

Results:

The cohort was 66% male and 61% African American, with a mean (SD) age of 4 (3) years at HRA enrollment. Despite earlier asthma diagnosis, children with a high (vs low) burden of early childhood risk factors experienced delayed HRA enrollment and 2–3 years of increasing annual SAE odds before enrollment (adjusted OR per year 1.54; 95% CI, 1.31–1.80). HRA enrollment was associated with a 36% reduction in SAE odds (adjusted OR 0.64; 95% CI, 0.51–0.80). Reductions were greater among children who initiated inhaled corticosteroids within 12 months of diagnosis (adjusted OR 0.61; 95% CI, 0.48–0.78).

Conclusions:

Enrollment in specialty asthma care was associated with reduced SAE risk; however, high-risk children experienced substantial delays in HRA enrollment despite escalating exacerbations. Early risk stratification may facilitate timely enrollment into specialty care and improve outcomes.

Graphical Abstract

INTRODUCTION

Asthma affects nearly 6% of children in the United States and remains a leading cause of pediatric emergency department (ED) visits and hospitalizations, accounting for more than 790,000 ED visits annually and nearly one-quarter of pediatric hospitalizations.^1,2 The burden of severe asthma exacerbations (SAEs) is driven by persistent airway inflammation and hyperresponsiveness, which increase susceptibility to common triggers such as viral infections and allergens. Although children with severe asthma represent only 4% - 5% of the pediatric asthma population,^3–5 they account for a disproportionate share of morbidity and healthcare costs.^6,7

Clinical guidelines recommend referral to a pulmonologist or allergist for children with frequent exacerbations, recent hospitalization, or need for high-intensity therapy.^8,9 Specialist care can clarify diagnosis, optimize treatment, and facilitate access to advanced therapies, including biologics.^8,10,11 However, referral to specialist care remains inconsistent. More than half of children hospitalized for asthma do not receive specialist follow-up within two years,¹² and fewer than half of children with prior SAEs ever see a specialist.^13,14 Even when guidelines are available, adherence to referral recommendations is suboptimal.¹⁵

Although prior studies suggest that specialist involvement reduces exacerbations,^16–19 little is known about whether delays in enrollment contribute to escalating exacerbation trajectories, or whether the benefits of specialist care vary according to early-life risk or timing of controller therapy initiation. We hypothesized that (1) children with a high burden of early childhood asthma risk factors would experience delayed specialist care enrollment despite worsening exacerbation risk, and (2) earlier enrollment in specialist care - particularly in conjunction with early inhaled corticosteroid initiation - would be associated with greater reductions in SAEs.

To test these hypotheses, we conducted a pre-post, quasi-experimental longitudinal cohort study to characterize the correlates of delayed specialist care and evaluate whether pre-post changes in childhood severe asthma exacerbation (SAE) incidence differed by patient characteristics.

METHODS

Study design and data sources

We used a pre-post, quasi-experimental, longitudinal design to generate the study cohort from the Cerner electronic health record (EHR) data warehouse of Indiana University Health (IU Health) system–affiliated institutions and clinics. IU Health provides primary and subspecialty care to over 1.2 million Indiana residents. To capture encounters outside IU Health, the Indiana Network for Patient Care (INPC) -which includes IU Health and over two-thirds of healthcare institutions in Indiana - was queried for diagnoses, medication orders, and prescriptions from inpatient, outpatient, and emergency department encounters.

Data were extracted using structured queries with standardized codes, including ICD-9-CM/ICD-10-CM for diagnoses, RxNorm for medications, and CPT codes for procedures. Extracted variables included demographics, encounter characteristics, diagnosis codes, and medication orders, and relevant clinical measures during the study period. Records were de-duplicated and cleaned to ensure coding consistency and temporal alignment.

IU Health and INPC records were linked within the INPC infrastructure using deterministic matching on unique identifiers and key demographics, supplemented by probabilistic algorithms to resolve discrepancies. Matched records were consolidated to create longitudinal patient-level data and reduce the likelihood of missing encounters outside the IU Health system. In addition, because referral order dates were inconsistently documented in the EHR, timing of specialist care was operationalized using age and time to enrollment in the multidisciplinary high-risk asthma (HRA) clinic. First specialist encounter and HRA enrollment were evaluated separately as healthcare milestones.

HRA Clinic Intervention

Patients are referred to the multidisciplinary HRA clinic by their treating pulmonologist or allergist based on standardized clinical criteria, including prior intensive care unit admission for asthma, recurrent asthma-related hospitalizations (≥2/year), recurrent emergency department visits (≥3/year), persistent airflow obstruction on spirometry, or frequent oral corticosteroid courses (>3/year). Poor adherence or psychosocial barriers alone are insufficient for referral in the absence of at least one clinical severity criterion.

Standard pulmonary and allergy specialist care across INPC-participating institutions typically consists of follow-up visits every 3–6 months with rotating nursing support and limited dedicated social work availability. In contrast, the HRA clinic provides more frequent follow-up (typically every 6–8 weeks) and multidisciplinary care involving a certified asthma educator, clinical psychologists, dedicated social work support, and pediatric pulmonologists. Care focuses on asthma education, adherence, psychosocial assessment, barrier reduction, and treatment optimization, including evaluation for biologic therapies. Additional details regarding HRA clinic structure and implementation have been previously described.¹⁷

Study Cohort

Our study cohort included children ≤11 years old with asthma who were referred to and enrolled in a HRA specialty clinic between January 2010 and December 2021 (Figure E1). Linked EHR data were available from 2010–2024; therefore, restricting HRA enrollment to 2010–2021 ensured a minimum of 3 years of longitudinal follow-up after enrollment. Children born before 2010 were excluded because complete ascertainment of early-life risk factors and longitudinal follow-up were not consistently available due to transitions in IUH EHR systems. Patients were followed longitudinally through December 2024 using linked EHR data from participating Indiana Network for Patient Care (INPC) institutions. A patient’s HRA enrollment date served as the index date for all downstream analyses and defined the pre-HRA period (1–5 years before enrollment) and post-HRA period (1–3 years after enrollment).

Risk of severe asthma exacerbation (SAE) pre-post HRA enrollment

Time to HRA enrollment was evaluated from three reference points: (1) first documented healthcare encounter in the INPC, (2) incident asthma diagnosis, and (3) first pulmonology or allergy specialist encounter. Because all study patients ultimately enrolled in the HRA clinic, no patients were right-censored. Severe asthma exacerbation (SAE) was defined according to ATS/ERS criteria as an asthma-related hospitalization or emergency department visit.²⁰

The primary outcome was the occurrence of ≥1 SAE during the pre- versus post-HRA enrollment periods. Oral corticosteroid prescriptions not associated with a hospitalization or emergency department visit were excluded because indication for treatment of an acute exacerbation could not be reliably determined from the EHR. Annualized SAE measures were calculated using complete 12-month observation periods before and after HRA enrollment; however, patients enrolled before their first birthday contributed follow-up only to the post-HRA period.

Covariates

Early childhood asthma risk factors:

Table E1 summarizes ICD-9/ICD-10 codes and key terms used to identify study covariates. Demographic variables included age at incident asthma diagnosis, sex, race, and parental history of asthma. Asthma-related treatment history included age at first prescription of inhaled corticosteroids (ICS), ICS plus long-acting β-agonists (ICS+LABA), biologics, and oral corticosteroids. Early childhood asthma risk burden was quantified using our previously validated Passive Digital Marker (PDM) score, which incorporates EHR-derived risk factors assessed at ≤3 years of age, including parental asthma, eczema, wheezing, pneumonia, bronchiolitis, and allergic sensitization.²¹

To account for heterogeneity in allergy profiles and asthma treatment patterns, hierarchical clustering on principal components (HCPC)²² was applied to allergy and treatment variables to identify patient subgroups with similar sensitization and treatment trajectories. Allergy clusters were derived from reported allergies and allergen-specific IgE results, while treatment clusters were based on age and duration of ICS, ICS+LABA, and biologic therapy. Cluster membership was subsequently evaluated as a potential modifier of HRA-associated changes in SAEs.

Longitudinal models additionally adjusted for calendar-year fixed effects and a time-varying COVID-19 indicator (2020–2021 vs all other years) to account for secular and pandemic-related changes in healthcare utilization. Markers of fragmented asthma care included the frequency of SAEs before incident ICS use and the number of distinct asthma-related healthcare institutions visited before HRA enrollment.

Statistical analysis

Descriptive statistics were used to summarize demographic and clinical characteristics before and after HRA enrollment.

Kaplan–Meier curves were used to characterize time from birth, first documented healthcare encounter within the INPC, incident asthma diagnosis, and first specialist appointment to HRA enrollment. Time to HRA enrollment was further evaluated using Cox proportional hazards models adjusted for demographic characteristics and early childhood risk factors. Although all patients ultimately enrolled in HRA care, Cox models were used to estimate relative differences in time to enrollment while accommodating varying observation times and multivariable adjustment. Adjusted hazard ratios (aHRs) and 95% confidence intervals (CIs) were estimated for correlates of earlier versus later HRA enrollment.

To evaluate changes in severe asthma exacerbations (SAEs) before and after HRA enrollment, we used piecewise generalized linear mixed-effects models (GLMMs) with a logit link and a knot at the HRA enrollment year (index year = 0). Our primary GLMM outcome was a binary indicator of ≥1 SAE per year. Models included separate pre- and post-HRA slopes, an immediate level change at enrollment, and random intercepts to account for within-patient correlation. Covariates included demographic characteristics, calendar year, and PDM risk classification. Time-varying covariates included asthma treatments (ICS, ICS+LABA, and biologics) and periods of strict COVID-19 mitigation measures (2020–2021). Covariate-by-time interactions were used to assess whether pre-post changes in SAE odds differed across patient subgroups. Incomplete person-years were retained using available follow-up under maximum likelihood estimation assumptions. Covariates for confounding control were selected using a directed acyclic graph (DAG) framework.²⁴

Prespecified subgroup analyses were performed by sex, race, PDM risk classification, allergy clusters, treatment clusters, and asthma severity.

Sensitivity analyses evaluated potential impact of unmeasured confounding using E-values.²⁵ Here, an E-value of 1 indicates that no unmeasured confounding is required to explain away the observed association, whereas larger E-values indicate increasing robustness to potential unmeasured confounding.

Recurrent SAE events were analyzed using an Andersen–Gill model,²⁶ which treated each SAE as a separate event and incorporated time-varying exposures to HRA enrollment and asthma-related treatments (ICS, ICS+LABA, and biologics). SAEs occurring ≥1 week apart were considered distinct events. As an alternative approach to address time-varying confounding and informative observation bias, we also fit a marginal structural model using inverse probability weighting for HRA enrollment and visit frequency.

Final models were selected based on Akaike and Bayesian Information Criteria, with lower values indicating better model fit. Statistical significance was defined as a two-sided p-value <0.05. Analyses were conducted using R version 4.4.3.

RESULTS

Participant Characteristics

The cohort included 229 children with asthma enrolled in the HRA clinic between 2010 and 2021 (Figure E1). Overall, 66.4% were male, 60.7% were Black or African American, and 8.7% were Hispanic/Latino; 89.1% had Medicaid insurance (Table 1). The mean (SD) age was 2 (2) years at incident (first EHR-documented) asthma diagnosis and 4 (2) years at HRA enrollment. Based on early-life (≤3 years) risk factors, 69% were classified as high PDM risk (i.e., predicted risk of asthma diagnosis at ≥5 years). Higher PDM risk was associated with earlier incident asthma diagnosis (Figure E2A) but not earlier ICS initiation (Figure E2B). In the year before the index year, 97.4% and 54.1% had ≥1 prescription for ICS and ICS+LABA, respectively. Prior to ICS initiation, 64% of the patients had at least one SAE. Similarly, prior to HRA enrollment, most (62%) patients had a history of two or more distinct asthma-institutions/clinics from which they received asthma related care. Overall, a patient had a median follow-up of 10 years (IQR, 7–12).

Table 1.

Demographic and Clinical Characteristics of the HRA study cohort

Characteristic	Categories/Statistic	N = 229
Sex	Female	77 (33.6%)
	Male	152 (66.4%)
Race	White	78 (34.1%)
	Black/African American	139 (60.7%)
	Others1	12 (5.2%)
Ethnicity	Non-Hispanic/Latino	209 (91.3%)
	Hispanic/Latino	20 (8.7%)
Age at Specialist Care enrollment (Years)	Mean (SD)	4 (3)
Asthma Diagnosis (Age: Years)	Mean (SD)	2 (2)
Pre-school age asthma	Yes	181 (79.0%)
Early childhood asthma risk factors used to generate the PDM score 2	Allergy/Allergy sensitization	167 (72.9%)
	Eczema	117 (51.1%)
	Wheezing	199 (86.9%)
	Wheezing without cold	187 (81.7%)
	Parental asthma	153 (66.8%)
	Bronchiolitis	83 (36.2%)
	Pneumonia	77 (33.6%)
Tobacco Exposure	Yes	201 (87.8%)
PDM score	Low (0–39)	33 (14.4%)
	Moderate (40–79)	38 (16.6%)
	High (80–115)	158 (69.0%)
Asthma Severity	Mild	46 (20.1%)
	Moderate/Severe	79 (34.5%)
	Unspecified	104 (45.4%)

¹Others - Race: Asian 2 (0.9%); Bi-Racial 1 (0.4%); Other 9 (3.9%)

²All PDM-related variables are determined ≤3 years old

Children diagnosed before age 5 years experienced longer time from incident asthma diagnosis to HRA enrollment compared with those diagnosed at ≥5 years (Figure 1A; log-rank p<0.001). After adjustment for PDM risk and demographics, the likelihood of early HRA enrollment (<5 years from diagnosis) was 92% lower among those diagnosed before age 5 (aHR 0.08; 95% CI, 0.03–0.19; p<0.001).

Figure 1.

Kaplan-Meir Plot: Time from asthma diagnosis to specialist care enrollment by age of asthma diagnosis (A) and PDM risk (B)

Moderate and high PDM risk were also associated with delayed enrollment (Figure 1B; log-rank p=0.0019). Compared with low risk, moderate and high PDM risk were associated with 52% and 47% lower likelihood of earlier enrollment (aHR 0.48; 95% CI, 0.30–0.77; p=0.002; and aHR 0.53; 95% CI, 0.36–0.77; p<0.001, respectively). Higher PDM risk was associated with shorter time to enrollment when measured from birth but not from other healthcare contact points (Figure E3).

Hierarchical clustering identified three allergy clusters (explaining 58.1% of variance) and three pre-HRA asthma-treatment clusters (39.0% of variance) (Figures E4–E5; Tables E2–E5). Allergy clusters differentiate children with low environmental sensitization from those with broader aeroallergen or food sensitization. Treatment clusters reflected escalation stage before and after HRA enrollment and were included in subsequent analyses.

Severe Asthma Exacerbations (SAEs) Pre and Post HRA Enrollment

Annualized SAE incidence increased progressively in the years preceding HRA enrollment (19% to 82%) and declined sharply afterward (43% in year 1; 41% in years 2–3) (Figure 2A-B). Overall, HRA enrollment was associated with a 36% reduction in SAE odds (adjusted OR 0.64; 95% CI, 0.51–0.80), adjusting for PDM risk, prior ICS/ICS+LABA use, and demographics. Further covariate adjustment considering calendar year showed a slightly more pronounced post HRA SAE reduction (adjusted OR 0.61; 95%CI: 0.48, 0.76). No comparable reduction was observed after incident specialist appointment (Figure E6; adjusted OR 1.28; 95%CI, 1.09–1.52).

Figure 2.

Severe asthma exacerbations (SAEs) Pre-Post High Risk Asthma (HRA) Clinic enrollment

Although not statistically significant, the COVID-19 mitigation period (2020–2021) was associated with lower pre-HRA SAE odds (adjusted OR: 1.10; 95%CI: 0.89, 1.39) and greater post-HRA reductions (adjusted OR 0.50; 95% CI, 0.35, 0.71) compared with periods before or after implementation of COVID-related preventive measures (adjusted OR: pre HRA – 1.81; 95%CI: 1.48, 2.21 and post HRA - 0.68; 95%CI, 0.68, 0.91).

Heterogeneity of HRA Associations

Pre-HRA SAE trajectories differed by age at asthma diagnosis (interaction p<0.001), with higher pre-HRA odds among children diagnosed before age 5 (Figure 2C). Post-HRA associations varied by both age at diagnosis and age at enrollment (three-way interaction p=0.0011): earlier HRA enrollment was associated with lower annual SAE odds, particularly among children diagnosed at ≥5 years (Figure 2D; Table 2). Post-HRA reductions in SAE odds also varied by timing of ICS initiation (interaction p=0.0547), but not by initiation of ICS+LABA or biologics. Only children initiating ICS within 12 months of diagnosis experienced significant post-HRA reductions in SAEs (adjusted OR 0.61; 95% CI, 0.48–0.78). Markers of fragmented asthma care - including more frequent SAEs before ICS initiation and a greater number of asthma-related care institutions before HRA enrollment - were associated with larger post-HRA reductions in SAE odds (Table E6). No significant association modification was observed by PDM risk, allergy clusters, asthma severity, sex, race, or tobacco exposure (p>0.05).

Table 2.

Piecewise Generalized Linear Mixed Model Results: Severe Asthma Exacerbations Pre-Post HRA Initiation

		OR (95%CI) Pre-HRA Slope		OR (95%CI) Post-HRA Slope
Characteristic	Subgroup	Crude model*	Adjusted model1	Crude model*	Adjusted model1
Overall		1.57 (1.36, 1.80)	1.54 (1.31, 1.80)	0.64 (0.52, 0.77)	0.64 (0.51, 0.80) #
Sex	Female	1.48 (1.17, 1.88)	1.53 (1.16, 2.02)	0.73 (0.53, 0.99)	0.71 (0.49, 1.01)
	Male	1.60 (1.34, 1.91)	1.68 (1.36, 2.07)	0.59 (0.46, 0.76)	0.56 (0.42, 0.74) #
Race	White	1.82 (1.32, 2.51)	1.82 (1.26, 2.64)	0.54 (0.37, 0.78)	0.53 (0.35, 0.80) #
	Black	1.45 (1.24, 1.70)	1.52 (1.27, 1.84)	0.72 (0.56, 0.91)	0.68 (0.52, 0.89) #
	Other	3.56 (1.12, 11.3)	4.23 (1.28, 14.0)	0.28 (0.09, 0.87)	0.25 (0.07, 0.84) #
Age of Incident Asthma Diagnosis	0–4 years	1.60 (1.37, 1.87)	1.72 (1.43, 2.05)	0.69 (0.56, 0.84)	0.65 (0.52, 0.81) #
	5–11 years	1.35 (0.82, 2.22)	1.36 (0.74, 2.51)	0.07 (0.01, 0.38)	0.07 (0.01, 0.49)
PDM Risk	Low	1.63 (0.77, 3.42)	1.47 (0.63, 3.45)	0.15 (0.04, 0.58)	0.17 (0.04, 0.76)
	Moderate	1.38 (1.01, 1.88)	1.56 (1.09, 2.24)	0.63 (0.39, 1.02)	0.58 (0.33, 1.00) #
	High	1.59 (1.35, 1.87)	1.72 (1.42, 2.10)	0.71 (0.57, 0.88)	0.66 (0.51, 0.84) #
Allergy Cluster	1	1.53 (1.32, 1.77)	1.61 (1.35, 1.91)	0.62 (0.51, 0.76)	0.59 (0.47, 0.75) #
	2	2.69 (1.17, 6.16)	4.10 (1.59, 10.6)	0.76 (0.31, 1.87)	0.58 (0.22, 1.49)
	3	0.83 (0.29, 2.43)	1.06 (0.28, 3.93)	1.13 (0.27, 4.79)	1.10 (0.25, 4.74)
Asthma Severity	Mild	1.92 (1.38, 2.67)	1.86 (1.29, 2.68)	0.69 (0.45, 1.06)	0.69 (0.43, 1.12)
	Moderate/Severe	1.30 (1.07, 1.58)	1.39 (1.11, 1.75)	0.68 (0.50, 0.93)	0.63 (0.44, 0.91) #
	Unspecified	1.73 (1.33, 2.26)	1.82 (1.81, 1.83)	0.57 (0.42, 0.78)	0.55 (0.55, 0.55) #
Tobacco Exposure	Yes	1.97 (1.58, 2.47)	1.98 (1.54, 2.55)	0.60 (0.45, 0.78)	0.59 (0.43, 0.80) #
	No	1.28 (1.07, 1.54)	1.41 (1.13, 1.76)	0.67 (0.51, 0.89)	0.62 (0.45, 0.85) #
Time from incident asthma diagnosis to ICS initiation	<12 Months	1.55 (1.32, 1.82)	1.62 (1.34, 1.95)	0.64 (0.51, 0.80)	0.61 (0.48, 0.78) #
	12–23 Months	1.84 (1.06, 3.20)	1.97 (1.01, 3.81)	0.78 (0.44, 1.40)	0.75 (0.39, 1.46)
	24+ Months	1.44 (1.02, 2.04)	1.61 (1.03, 2.49)	0.63 (0.66, 1.20)	0.56 (0.27, 1.18)
Pre-HRA Medication cluster	1	1.39 (1.16, 1.64)	1.43 (1.18, 1.74)	0.63 (0.48, 0.83)	0.62 (0.45, 0.84) #
	2	2.05 (1.56, 2.70)	2.20 (1.59, 3.04)	0.63 (0.47, 0.84)	0.59 (0.42, 0.82) #
Post-HRA Medication cluster	1	1.33 (1.03, 1.72)	1.49 (1.11, 1.99)	0.70 (0.48, 1.02)	0.65 (0.42, 0.98) #
	2	1.79 (1.46, 2.20)	1.83 (1.44, 2.33)	0.66 (0.52, 0.85)	0.64 (0.49, 0.85) #
	3	1.26 (0.86, 1.86)	1.31 (0.84, 2.04)	0.51 (0.25, 1.04)	0.49 (0.22, 1.06)

Table shows the results of the adjusted piecewise GLMM results in each of the listed subgroups.

^*The crude models include only time-related variables for estimating pre- and post-slopes, with no covariate adjustments.

^#Pre vs post adjusted slopes were significantly different. The adjusted models include PDM score and age of HRA enrollment for estimating pre- and post-slopes; model for overall sample is adjusted by PDM score and age of high-risk enrollment.

Sub-group Analyses

Pre-HRA SAE odds increased with PDM risk, whereas post-HRA reductions were greatest among low-risk children, who also had lower allergy burden (Table 2; Tables E2–E3). In treatment clusters 1 and 2, SAE odds increased before and declined after HRA enrollment. The largest post-HRA reductions were observed in cluster 3, characterized by lower medication exposure and minimal ICS+LABA or biologic use (Table 2; Tables E4–E5). Pre-HRA SAE odds increased with asthma severity and were highest among children with unclassified severity. Post-HRA reductions were modestly greater among children with moderate/severe versus mild asthma. Associations between HRA enrollment and reduced SAE odds were otherwise consistent across sex, race, and tobacco exposure strata (Table 2).

Time-Varying Associations of HRA Enrollment and Asthma treatments

In Andersen–Gill recurrent event models, HRA enrollment was associated with lower SAE intensity (aHR 0.52; 95% CI, 0.47–0.58) after adjustment for demographics and treatment history. Initiation of ICS was associated with the greatest reduction in recurrent SAE risk (aHR 0.40; 95% CI, 0.37–0.43), followed by ICS+LABA (aHR 0.51; 95% CI, 0.46–0.56) and biologics (aHR 0.65; 95% CI, 0.45–0.92) (Figure 3; p<0.01 for trend).

Figure 3.

Forest Plot: Time-varying drivers of severe asthma exacerbations during study follow

MSM Counterfactual Trajectories

In marginal structural model analyses, weighted counterfactual trajectories demonstrated consistently lower predicted SAE probabilities among HRA-enrolled children than among non-enrolled children (Figure 4). Although SAE risk declined over time in both groups, reductions were greater among HRA-enrolled children, with divergence of the trajectories occurring within 2–3 years of asthma diagnosis and persisting throughout follow-up.

Figure 4.

Marginal Structural Model (MSM) Adjusted Longitudinal Trajectories. Estimated population-average (marginal) trajectories for severe asthma exacerbations over a 9-year follow-up period after an incident asthma diagnosis, comparing the HRA enrolled state (teal) to the non-HRA enrolled counterfactual (red). Estimated proportion of severe asthma exacerbations were derived from a weighted generalized estimating equation (GEE) utilizing inverse probability weights to adjust for time-varying clinical confounding and informative observation frequency.

Sensitivity Analyses

E-value analyses indicated that an unmeasured confounder would need to be associated with both HRA enrollment and SAEs by an odds ratio of at least 2.50 to fully explain the observed association (E-value = 2.50; lower confidence bound = 1.85).

DISCUSSION

Our findings provide real-world evidence that children with the highest burden of early childhood asthma risk factors experience substantial delays in enrollment into specialized asthma care despite a disproportionately high and increasing burden of severe asthma exacerbations (SAEs). Although these children were diagnosed with asthma earlier, HRA enrollment was delayed by an average of 2–3 years while annualized SAE rates continued to increase. Earlier intervention - both in terms of inhaled corticosteroid (ICS) initiation and HRA enrollment was associated with greater reductions in SAE risk, whereas reductions were attenuated among children receiving more intensive therapies, likely reflecting greater baseline disease severity. Marginal structural model analyses provided complementary evidence to the primary GLMM and recurrent-event analyses, demonstrating consistently lower predicted SAE risk among HRA-enrolled children after accounting for time-varying confounding. To our knowledge, this is the first study to systematically characterize delays in HRA enrollment and heterogeneity in post-enrollment reductions in SAE risk.

The delayed enrollment of high-PDM-risk children is concerning. Prior studies have reported that fewer than 50% of children with asthma receive specialist consultation within one year of experiencing multiple SAEs.^13,14 Our findings extend this literature by showing that children with the greatest early-life asthma risk not only experience a higher exacerbation burden but also delayed ICS initiation and HRA enrollment. These delays may represent missed opportunities for earlier intervention. Children with elevated PDM risk may represent a distinct, difficult-to-treat phenotype that is less responsive to standard stepwise management alone. Early risk stratification and proactive case-finding may facilitate timely initiation of controller therapy and specialty care.

We also observed larger post-HRA reductions in SAE risk among children with markers of fragmented asthma care, including frequent SAEs before ICS initiation and care across multiple healthcare institutions. These findings suggest that structured HRA care may be particularly beneficial for children experiencing discontinuity in asthma management by improving care coordination, adherence support, and guideline-directed treatment. Identifying such patients may help target specialty care resources toward those most likely to benefit.

Importantly, specialist appointment alone was not associated with reductions in SAE risk, whereas enrollment in the multidisciplinary HRA clinic was associated with meaningful reductions in both SAE incidence and recurrence. Although universal early referral and enrollment into specialty care may not be feasible in resource-constrained settings, risk-stratification approaches based on PDM risk or exacerbation history may help identify children most likely to benefit from early intensive specialty care. The observed reductions in recurrent SAEs associated with ICS, ICS+LABA, and biologic initiation further support the value of earlier specialty involvement before disease progression necessitates advanced therapies.

Several limitations warrant consideration. As an observational study, residual confounding cannot be excluded. Although post-enrollment improvements may partially reflect care optimization strategies, regression to the mean, or changes in care intensity, the absence of comparable reductions following specialist appointment alone and the consistency of findings across sensitivity analyses suggest that these factors are unlikely to fully explain the observed associations. Because referral order dates were inconsistently documented in the EHR, our findings should be interpreted as evaluating timing of HRA enrollment rather than referral timing directly. Additionally, several subgroup and interaction analyses involved relatively small strata and should be considered exploratory pending validation in larger multicenter cohorts. Outcomes and covariates were derived from EHR data and are therefore subject to documentation error and missingness.²⁸ We excluded oral corticosteroid prescriptions from the SAE definition because indication for acute exacerbation treatment could not be reliably confirmed. Finally, although the study period overlapped with the COVID-19 pandemic, adjustment for calendar year and pandemic-related factors did not materially alter the observed associations.

Despite these limitations, this study leverages longitudinal, linked EHR data to examine real-world asthma outcomes among children receiving specialty care. The consistency of findings across multiple analytic approaches - including GLMM, Andersen–Gill, marginal structural model, and E-value analyses - supports the robustness of the observed associations. Furthermore, E-value analyses indicated that a moderately strong unmeasured confounder would be required to fully explain the association between HRA enrollment and reduced SAE risk.

In summary, enrollment in a structured multidisciplinary asthma specialty clinic was associated with meaningful reductions in severe asthma exacerbations, particularly when ICS therapy and specialty care were initiated earlier in the disease course. Integrating early childhood asthma risk stratification into routine care may facilitate timely enrollment of high-risk children into specialty programs and improve long-term outcomes.

HIGHLIGHTS.

1. What is already known about this topic?

Specialist care reduces emergency department visits and hospitalizations among children with uncontrolled asthma. Current guidelines recommend referral for poor control or frequent exacerbations, but the consequences of delayed specialty care enrollment remain unclear.

2. What does this article add to our knowledge?

High-risk children experienced 2–3 years of delayed enrollment into specialty asthma care despite escalating severe asthma exacerbations (SAEs). HRA enrollment was associated with a 36% reduction in SAE odds, with greater reductions among children who initiated inhaled corticosteroids within 12 months of diagnosis.

3. How does this study impact current management guidelines?

Early risk stratification may facilitate timely enrollment of high-risk children into specialty asthma care. Proactive specialty care involvement - rather than referral after repeated exacerbations - may reduce preventable SAEs and improve implementation of guideline-based asthma management.

List of Abbreviations

AG

Andersen–Gill

aHR

adjusted hazard ratio

aOR

adjusted odds ratio

CI

confidence interval

ED

emergency department

EHR

electronic health record

GINA

Global Initiative for Asthma

GLMM

generalized linear mixed-effects model

HRA

high-risk asthma

ICS

inhaled corticosteroid

ICS+LABA

inhaled corticosteroid plus long-acting beta-agonist

INPC

Indiana Network for Patient Care

IUH

Indiana University Health

IQR

interquartile range

OCS

oral corticosteroid

PDM

Passive Digital Marker

SAE

severe asthma exacerbation

SD

standard deviation

DATA SHARING

De-identified individual participant data will be available upon reasonable request to the corresponding author, subject to institutional data use agreements and ethical approvals.