Feasibility of Assessing Clinical Remission in Asthma via a Customized Data Extraction Approach from a Real-World Clinical Electronic Medical Records Database in Japan: A Retrospective Real-World Study

The study proposed and demonstrated a customized pipeline for extracting asthma clinical remission components from real-world electronic medical records, with ACT scores extracted via structured query language and exacerbation-related hospitalization and emergency transport data extracted via a large language model. Relative to conventional remission assessment based on manual chart review or prospective registries, this work combines structured queries with language-model extraction to assemble the multi-source fields needed for remission determination from an existing electronic medical records database. The methods description specifies the extraction approaches and data source (the JAMDAS database), but this summary-level text does not report accuracy validation metrics for the language-model extraction.

Among 32,258 eligible patients initiating triple therapy, the proportion with recorded objective measures needed for remission assessment was low: ACT 1.8-3.0%, FEV1% 0.5-1.5% and FeNO 4.5-10.1% across 3 to 18 months. This quantifies how often remission-related measures are recorded in Japanese real-world practice, complementing remission data that have mostly come from clinical trials or specialty registries. Descriptive counts from a large 32,258-patient database, with proportion ranges spanning multiple time points from 3 to 18 months.

Among evaluable patients, the proportion meeting all clinical remission components rose from 41.1% (384/934) at 3 months to 69.3% (199/287) at 18 months. Provides longitudinal estimates of guideline-defined remission over time among Japanese real-world asthma patients on triple therapy. Denominators are evaluable patients with ACT data (934 at 3 months, 287 at 18 months); the evaluable sample narrows substantially over time and only 2.9-3.0% of the overall population entered this analysis.

The study also assessed treatment patterns, adherence and persistence, and extracted FEV1%/forced vital capacity and FeNO testing frequency as companion dimensions of remission-assessment feasibility. Places remission determination alongside treatment behavior and testing behavior within the same database framework, so feasibility assessment extends beyond remission proportions alone. This summary-level text lists these assessment dimensions but does not provide specific values for adherence, persistence or pulmonary function testing frequency.

The setting this study defines is patients with asthma in the Japanese electronic medical records database (JAMDAS) who initiated fluticasone furoate/umeclidinium/vilanterol triple therapy, with remission determined by the Japanese 2024 guideline definition (oral corticosteroid-free, no exacerbations, ACT >= 23) and assessment windows from 3 to 18 months after initiation at 3-month intervals. It can support feasibility exploration and longitudinal description of remission in such databases and provide a reference for researchers concerned with recording availability; for databases with different recording habits or other remission definitions, field availability and definition mapping need to be reassessed.

A careful reader may still watch: how the accuracy of language-model extraction of hospitalization and emergency transport data is validated; how the narrowing evaluable sample from 934 to 287 patients affects the stability of the 18-month remission estimate; the specific values for adherence, persistence and FEV1%/FVC and FeNO testing frequency; and whether similar patterns of remission-component availability appear in databases with different recording habits. This is summary-level information without figures or full result tables, so these details await fuller presentation in the original article.