BACKGROUND:Children with uncontrolled, moderate-to-severe asthma often have type 2 disease with allergic sensitization and/or elevated serum IgE. OBJECTIVE:To examine dupilumab efficacy in children aged 6 to 11 years with/without allergen sensitization. METHODS:In VOYAGE (NCT02948959), 408 children were randomized to receive dupilumab 100/200 mg every 2 weeks (by weight) or placebo for 52 weeks. This post hoc analysis included 336 children from VOYAGE with type 2 inflammation (eosinophils ≥150 cells/μL or fractional exhaled nitric oxide ≥20 ppb) and known baseline allergen sensitization status (based on total serum IgE and perennial allergen-specific IgE levels). The primary outcome assessed was annualized severe exacerbation rate; we also assessed change from baseline in pre- and postbronchodilator percent predicted forced expiratory volume in 1 second (ppFEV1), Interviewer-Administered 7-Item Asthma Control Questionnaire (ACQ-7-IA) score, and total and allergen-specific IgE levels. RESULTS:Of 336 children with type 2 inflammation and known baseline allergen sensitization status, 75 (22%) were nonsensitized, 58 (17%) monoallergen sensitized, and 203 (60%) multiallergen sensitized. Dupilumab versus placebo reduced asthma exacerbations by 45%, 75%, and 60% in the non-, mono-, and multisensitized subgroups (P = .1286, .0376, and .0004), respectively, by week 52, with no apparent interaction between treatment group and sensitization status (Pint = .48). Similar improvements were observed across sensitization subgroups for pre- and postbronchodilator ppFEV1, ACQ-7-IA, and total IgE levels. CONCLUSION:In children with type 2 asthma, dupilumab showed significant efficacy, reducing exacerbations and improving other outcomes in monoallergen- and multiallergen-sensitized patients, and numerical but nonsignificant effects in nonsensitized patients.

BACKGROUND:Asthma classified as "mild" is commonly assumed to have a low symptom burden and limited impact on quality of life. OBJECTIVE:To characterize patient-reported asthma burden among adults with mild asthma identified using electronic health record-based criteria. METHODS:A total of 3787 adults aged 18-85 years completed a survey assessing asthma burden. Asthma control and health-related quality of life (HRQOL) were measured using the Asthma Impairment and Risk Questionnaire (AIRQ) and the Veterans RAND 12-item Health Survey (VR-12), respectively. Investigator-developed items were used to assess symptoms, functional impact, and health care utilization in the prior year. Associations between patient characteristics and prespecified outcomes-very poorly controlled asthma in the past 2 weeks, symptoms ≥2 days/wk, and ≥2 patient-reported asthma-related emergency department (ED) or urgent care (UC) visits in the prior year-were evaluated using multivariable analyses. RESULTS:Asthma control was distributed across AIRQ categories as follows: 34.6% well controlled, 32.9% not well controlled, and 32.5% very poorly controlled. Patient-reported burden was notable: 27.6% reported symptoms at least a few times per week, 52.7% reported systemic corticosteroid use ever, and 41.4% reported ≥1 ED/UC visit in the prior year, including 19.7% with ≥2 visits. The mean [standard deviation] VR-12 scores indicated impaired HRQOL (physical component summary: 43.3 [9.7]; mental component summary: 44.7 [11.8]). Hispanic/Latino and Black/African American patients had higher risks of very poorly controlled asthma and ≥2 ED/UC visits in the prior year compared with non-Hispanic White patients. CONCLUSIONS:Among surveyed adults with utilization-defined mild asthma and a recent asthma-related healthcare encounter, substantial patient-reported burden was observed across multiple domains.

Background Uncontrolled asthma symptoms and prior-year exacerbations increase the risk for future exacerbations in school-aged children. The Pediatric Asthma Impairment and Risk Questionnaire (Peds-AIRQ) is an 8-item, equally weighted, yes/no asthma control tool validated for children aged 5-11 years that assesses current symptoms and exacerbation history. Objective To determine the relationship of asthma control, as assessed by the Peds-AIRQ, and subsequent 3-month exacerbation occurrence. Methods At enrollment, children with asthma (aged 5-11 years) and their parents/caregivers completed a baseline Peds-AIRQ. Parent/caregiver-reported exacerbations were captured monthly for 3 months post-enrollment via electronic survey. Logistic regression models used baseline Peds-AIRQ score, asthma control category (well-controlled [WC], not well-controlled [NWC], very poorly controlled [VPC]), age, sex, race, ethnicity, and body mass index as covariates, and ≥1 subsequent 3-month exacerbation as the dependent variable. Results Of 399 children enrolled in the Peds-AIRQ validation study, 313 completed ≥1 follow-up survey over 3 months: mean (SD) age 7.9 (1.9) years; 63.3% boys; 71.6% White; 25.9% Hispanic/Latino. Baseline Peds-AIRQ categorized 39.3% of children as having WC asthma, 43.1% NWC, and 17.6% VPC. During the 3-month follow-up, 69 (22%) children experienced 114 exacerbations. Each 1-point increase in baseline Peds-AIRQ score was associated with 32% increased odds of ≥1 subsequent 3-month exacerbation (OR [95% CI]: 1.32 [1.16-1.51], P < .001). Children with VPC asthma had greater odds of exacerbation than those with WC (OR [95% CI]: 3.85 [1.79-8.24], P < .001) or NWC asthma (OR [95% CI]: 2.56 [1.26-5.21], P < .01). Conclusion At point-of-care, Peds-AIRQ can help identify children at risk for exacerbation over 3 months.

BACKGROUND:Early-onset atopic dermatitis (AD) is a known precursor to respiratory atopic diseases, but identifying which children will develop persistent moderate-to-severe asthma and allergic rhinitis at school age remains difficult. OBJECTIVE:We sought to develop and validate machine learning models that predict individualized risk for moderate-to-severe persistent asthma and allergic rhinitis in children diagnosed with AD before age 3. METHODS:We conducted a retrospective birth cohort study using longitudinal electronic health record data from Kaiser Permanente Southern California. Two prediction models were developed for each outcome (asthma and rhinitis) among children aged 5-11: a comprehensive electronic health records model using detailed, structured clinical variables; and a simplified clinical model that was based on fewer, routinely available clinical features. Model performance was evaluated by area under the curve (AUC), sensitivity, positive predictive value (PPV), and calibration across risk strata. RESULTS:Among 10,688 eligible children, asthma models demonstrated strong discrimination (AUC = 0.893 comprehensive; AUC = 0.892 simplified). At 95% specificity, the comprehensive model achieved 40.4% sensitivity and 39.3% PPV; the simplified model reached 36.2% sensitivity and 33.8% PPV. Rhinitis models showed moderate performance (AUC = 0.793 and AUC = 0.773); at 90% specificity, the comprehensive model achieved 35.5% sensitivity and 72.7% PPV, while the simplified model reached 34.0% sensitivity and 69.2% PPV. Calibration was acceptable, with strong agreement in the highest-risk groups. CONCLUSION:Machine-learning models using early-life clinical data can accurately stratify risk for moderate-to-severe persistent asthma and allergic rhinitis by school age, supporting proactive, individualized care.

Background: Deep learning models for clinical risk prediction are commonly developed using large numbers of input features and long patient histories, increasing computational cost and implementation complexity. However, there is limited evidence on how reducing feature dimensionality and sequence length affects predictive performance and clinically relevant model behavior in longitudinal electronic health record (EHR)-based prediction. Methods: Using transformer-based models and longitudinal EHR data from adults with mild asthma within an integrated healthcare system, we systematically evaluated feature-level and sequence-level dimensionality reduction. Reduced-feature models were constructed using attribution-based (Integrated Gradients), univariate performance-based, and clinically guided feature selection strategies. Maximum sequence length was varied using percentile-based truncation of longitudinal histories. Model performance was assessed using discrimination, calibration, risk stratification, threshold-based event capture, and computational efficiency across six-fold cross-validation defined by geographically distinct Medical Service Areas. Findings: Models using substantially fewer input features achieved discrimination comparable to the high-dimensional reference model. Moderate sequence-length truncation reduced training time by more than 50% while preserving discrimination. Although reduced-dimensional models showed attenuated predicted risk at the highest risk levels, they identified similar high-risk groups and captured comparable proportions of observed asthma exacerbation events at clinically relevant decision thresholds. Interpretation: Transformer-based clinical risk prediction can be achieved using simpler and more computationally efficient input representations than commonly assumed. While dimensionality reduction may affect calibration at the upper tail of predicted risk, discrimination and threshold-based clinical performance remain largely preserved. Treating input dimensionality and sequence length as explicit, tunable design parameters may facilitate the development of scalable and resource-efficient clinical prediction models.

Acute asthma exacerbation (AAE) is among the most serious outcomes of asthma, and accurate prediction of its risk remains a key challenge. Existing electronic health record-based prediction models have typically adopted either cohort or case-control sampling designs, yet their comparative performance has not been systematically evaluated. To address this gap, we developed transformer-based deep learning models to predict AAE among adults with mild asthma, directly comparing cohort and case-control designs using identical predictors and architecture across two large integrated healthcare systems, Kaiser Permanente Southern California (KPSC) and Kaiser Permanente Northwest (KPNW). Models were trained on retrospective data from KPSC and externally validated in KPNW. Mean area under the receiver operating characteristic curve (AUC) was 0.85 for case-control models and 0.70 (KPSC)/0.71 (KPNW) for cohort models. Both designs generalized well across systems, indicating robust feature learning and population transferability. Although calibration appeared well aligned within each analytical framework, absolute predicted probabilities diverged between designs, reflecting how event-enriched sampling inflates apparent risk and affects interpretability. These findings demonstrate that study design strongly influences model behavior and should be aligned with the intended use when developing predictive models for clinical deployment.

Objective: Transformer-based models for clinical prediction using longitudinal electronic health record (EHR) data are often developed with large feature sets and long patient histories under the assumption that more data improves performance. However, high-dimensional inputs and long sequences increase computational burden, potentially limiting scalability. We evaluated how feature dimensionality and sequence length affect predictive performance, calibration, risk stratification, and computational efficiency in EHR prediction. Methods: Using longitudinal EHR data from adults with mild asthma in an integrated healthcare system, we evaluated input representation design for predicting acute asthma exacerbation. Feature dimensionality was reduced using Integrated Gradients attribution scores, univariate performance-based selection, and clinically guided selection strategies. Sequence length was varied using percentile-based truncation of patient histories. Performance was assessed using discrimination, calibration, high-risk classification, threshold-based event capture, and computational efficiency across regions. Results: Models using fewer features achieved discrimination comparable to the 80-feature reference model, with AUROC values ranging from 0.843 to 0.864 versus 0.870 for the full model. Moderate sequence-length truncation reduced training time by more than 70% with minimal loss in discrimination. Although reduced-dimensional models showed attenuation of predicted risk at the upper tail, they identified similar high-risk populations and captured comparable proportions of asthma exacerbation events at clinically relevant thresholds. Conclusion: Transformer-based prediction models maintained strong performance across reduced feature sets. While dimensionality reduction modestly affected calibration at the highest risk levels, moderate sequence-length reduction substantially reduced computational burden with limited change in overall discrimination. These findings highlight trade-offs between input complexity, predictive performance, and computational efficiency.

BACKGROUND:The Asthma Impairment and Risk Questionnaire (AIRQ) predicts 12-month exacerbation occurrence for patients aged 12 years and older. OBJECTIVE:To assess the short- and long-term exacerbation prediction ability of the AIRQ in patients with mild-to-moderate and severe asthma. METHODS:This post hoc analysis from the AIRQ longitudinal study classified patients with asthma aged 12 years and older as having mild-to-moderate or severe disease based on prescribed pharmacotherapy. Participant-reported severe asthma exacerbations were assessed monthly for 12 months. For both severity groups and relative to baseline AIRQ control category, exacerbation occurrence was assessed via logistic regression and Kaplan-Meier time-to-first event analyses for the overall 12-month period, months 0 to 3 (short term), and months 4 to 12 (long term) post-enrollment. RESULTS:Of 1070 patients who completed 1 or more follow-up assessment, 374 (35.0%) had mild-to-moderate asthma and 696 (65.0%) had severe asthma. Over the 12-month follow-up, 134 patients (35.8%) with mild-to-moderate disease vs 355 patients (51.0%) with severe disease experienced 1 or more exacerbation (P < .001). In months 0 to 3 and months 4 to 12, the proportion of patients experiencing 1 or more exacerbation was lower in those with mild-to-moderate than severe asthma (76 [21.0%] vs 201 [29.8%], P = .002; 93 [26.1%] vs 283 [41.4%], P < .001, respectively). For both severity groups, in the 12-month follow-up, months 0 to 3, and months 4 to 12, baseline AIRQ control category predicted exacerbation occurrence and time-to-first exacerbation (P < .001 for all). CONCLUSION:The AIRQ predicts short- and long-term exacerbation occurrence in patients with mild-to-moderate and severe asthma. Understanding how current asthma control relates to exacerbation risk could facilitate point-of-care shared decision-making on management optimization.

Objective: While prior studies showed lower socioeconomic status was associated with worse acute asthma exacerbations (AAE) and asthma-related emergency department/urgent care (ED/UC) visits, few have quantified contributions of risk factors. We aim to identify and quantify risk factors driving these disparities across levels of neighborhood deprivation. Methods: We conducted a cohort study of 79,562 adults with mild asthma in Kaiser Permanente Southern California, United States, during 2013-2018, with follow-up through 2019, comparing participants in Neighborhood Deprivation Index (NDI) quintiles five (Q5, most deprived) and one (Q1, least deprived). Outcomes included AAE and asthma-related ED/UC visits. We employed Oaxaca-Blinder decomposition to quantify contributions of 14 risk factors to these disparities. Results: Individuals in NDI Q5 had a 1.5% higher rate of AAE and a 3% higher rate of ED/UC visits than those in Q1. The 14 risk factors explained 61.2% of the AAE disparity and 52.3% of the ED/UC disparity. Race and ethnicity contributed most, 33.1% of the AAE disparity and 22.8% of the ED/UC disparity, followed by body mass index, prior asthma exacerbations, and exercise levels. Conclusions: These results underscore the need for targeted interventions addressing socioeconomic and behavioral factors to reduce asthma-related health disparities, particularly in disadvantaged populations.

BACKGROUND:Uncontrolled asthma in childhood is associated with exacerbations and impaired health-related quality of life. Commonly used control tools for children aged 5 to 11 years assess asthma symptoms but not exacerbations, potentially leading to overestimation of control and suboptimal management. OBJECTIVE:This cross-sectional observational study aimed to validate the Pediatric Asthma Impairment and Risk Questionnaire (Peds-AIRQ), a novel control tool assessing symptom impairment and exacerbation risk. METHODS:A total of 399 children aged 5 to 11 years with physician-diagnosed asthma were recruited from one primary care and seven specialty care sites across the United States. Parents and caregivers, with input from their child, answered 18 yes/no questions about asthma symptoms and exacerbations. Children were categorized as having well-controlled (WC), not well-controlled (NWC), or very poorly controlled (VPC) asthma according to a validation standard using Global Initiative for Asthma symptom control questions plus prior-year exacerbations. Items with the greatest ability to discriminate among control categories and cut points were determined through logistic regression analyses. RESULTS:Models yielded a Peds-AIRQ composed of five impairment-based and three risk-based questions. The Peds-AIRQ yielded areas under receiver-operating characteristic curves of 0.85 to differentiate WC versus NWC/VPC asthma and 0.83 to differentiate WC/NWC versus VPC asthma. Score cut points of 0 to 1, 2 to 4, and 5 to 8 yes responses were determined to best represent WC, NWC, and VPC asthma, respectively. CONCLUSIONS:The Peds-AIRQ is a validated numerical assessment tool for children aged 5 to 11 years that can be used at point-of-care to evaluate asthma control based on current symptoms and exacerbation history.

BACKGROUND:Knowledge of risk factors in early childhood predisposing to moderate-severe persistent asthma (MS-Asthma) in later childhood is needed. OBJECTIVE:To identify the risk factors for MS-Asthma at age 5 to 11 years in children with early-onset atopic dermatitis. METHODS:Electronic health records identified a birth cohort to 11 years of 10,688 children with AD onset between birth and age 36 months. International Classification of Diseases, Ninth Revision/Tenth Revision-coded visits and laboratory data to 36 months were used to detect potential child and maternal risk factors for MS-Asthma. MS-Asthma was defined as Global Initiative for Asthma-step-care level of 3 or higher for a minimum of 4 years from ages 5 to 11 years. Robust Poisson regression determined risk ratios for MS-Asthma. RESULTS:Compared with children who did not develop MS-Asthma (N = 10,168), those developing MS-Asthma (N = 520 [4.9%]) from age 5 to 11 years were significantly (P ≤.01) more likely to be male, of non-Hispanic Black ethnicity, preterm, not-exclusively breast-fed for 1 month, and up to age 3 years have more perinatal respiratory disorders and respiratory infections, food allergy, allergic rhinitis, asthma, allergic sensitizations, and elevated blood eosinophil levels. Mothers of children with MS-Asthma had significantly more comorbidities and antibiotics dispensed, and less preexisting diabetes (P < .001). Significant adjusted risk factors observed before 36 months associated with increased MS-Asthma at age 5 to 11 years included food allergy; 6 or more atopic dermatitis medication dispensings; nasal corticosteroid dispensing; and number of dispensings of inhaled short-acting beta-agonists, montelukast, and inhaled corticosteroids. Significant protective risk factors were 1-month exclusive breast-feeding and preexisting maternal diabetes. CONCLUSIONS:Risk models for MS-Asthma in later childhood were developed on the basis of early childhood and maternal factors using administrative data.

BACKGROUND:Individuals with mild asthma account for 30% to 40% of asthma exacerbations requiring emergency consultation, and nearly 30% had not-well controlled or poorly controlled asthma symptoms over the previous 4 weeks. OBJECTIVE:We sought to estimate the prevalence of various asthma symptoms and assess their association with future acute asthma exacerbations (AAEs) in patients with mild asthma. METHODS:This was a retrospective cohort study. Using administrative data, we identified 198,873 adults aged 18 to 85 years, who met criteria for mild asthma between 2013 and 2018. The presence of cough, wheezing, dyspnea, and chest tightness in the 12 months before the index visit (t0) was extracted from clinical notes and patient/provider communications through natural language processing. We used Poisson regression models with robust SEs to assess the associations between symptoms and AAEs in the 12 months after t0, controlling for potential confounders. RESULTS:The prevalence of cough, wheezing, dyspnea, and chest tightness was 67.0%, 47.7%, 41.3%, and 13.2%, respectively. Moreover, 6.5% of patients experienced an AAE in the 12 months after t0. All four symptoms were associated with increased AAE risk in the unadjusted analysis. After adjusting for other patient characteristics, only wheezing (adjusted relative risk, 1.13; 99% CI, 1.07-1.20) and dyspnea (1.17; 1.12-1.23) were associated with an increased risk of future AAEs. The risk increased with the frequency of the documented symptoms. CONCLUSION:Patients with mild asthma who exhibit symptoms of dyspnea and wheezing (especially on multiple occasions) are at an increased risk for future AAEs and may benefit from therapeutic intervention and/or trigger-avoidance education.

Background:Asthma-related symptoms are significant predictors of asthma exacerbation. Most of these symptoms are documented in clinical notes in a free-text format, and effective methods for capturing asthma-related symptoms from unstructured data are lacking. Objective:The study aims to develop a natural language processing (NLP) algorithm for identifying symptoms associated with asthma from clinical notes within a large integrated health care system. Methods:We analyzed unstructured clinical notes within 2 years before a visit with asthma diagnosis in 2013-2018 and 2021-2022 to identify 4 common asthma-related symptoms. Related terms and phrases were initially compiled from publicly available resources and then refined through clinician input and chart review. A rule-based NLP algorithm was iteratively developed and refined via multiple rounds of chart review followed by adjudication. Subsequently, transformer-based deep learning algorithms were trained using the same manually annotated datasets. A hybrid NLP algorithm was then generated by combining rule-based and transformer-based algorithms. The hybrid NLP algorithm was finally applied to the implementation notes. Results:A total of 11,374,552 eligible clinical notes with 128,211,793 sentences were analyzed. After applying the hybrid algorithm to implementation notes, at least 1 asthma-related symptom was identified in 1,663,450 out of 127,763,086 (1.3%) sentences and 858,350 out of 11,364,952 (7.55%) notes, respectively. Cough was the most frequently identified at both the sentence (1,363,713/127,763,086, 1.07%) and note (660,685/11,364,952, 5.81%) levels, while chest tightness was the least frequent at both the sentence (141,733/127,763,086, 0.11%) and note (64,251/11,364,952, 0.57%) levels. The frequency of multiple symptoms ranged from 0.03% (36,057/127,763,086) to 0.38% (484,050/127,763,086) at the sentence level and 0.10% (10,954/11,364,952) to 1.85% (209,805/11,364,952) at the note level. Validation against 1600 manually annotated clinical notes yielded a positive predictive value ranging from 96.53% (wheezing) to 97.42% (chest tightness) at the sentence level and 96.76% (wheezing) to 97.42% (chest tightness) at the note level. Sensitivity ranged from 93.9% (dyspnea) to 95.95% (cough) at the sentence level and 96% (chest tightness) to 99.07% (cough) at the note level. All 4 symptoms had F1-scores greater than 0.95 at both the sentence and note levels, regardless of NLP algorithms. Conclusions:The developed NLP algorithms could effectively capture asthma-related symptoms from unstructured clinical notes. These algorithms could be used to facilitate early asthma detection and predict exacerbation risk.