An Innovative Approach to Identify Asthma among Preschool Children through Automated Chart Review
Saturday, March 3, 2018: 3:00 PM
S330AB (Convention Center)
Young J. Juhn, MD MPH, , , ,
RATIONALE: A timely asthma diagnosis for preschool children is often difficult as lung function tests are not always feasible. To address this challenge, we assessed feasibility of natural language processing (NLP) algorithm-based asthma ascertainment for children prior to 4 years of age.

METHODS: Asthma status of the 1997-2007 Olmsted County Birth Cohort was defined by the NLP algorithm for Predetermined Asthma Criteria (NLP-PAC) during the first 4 years of life (Yes/No). We assessed concurrent validity by using Asthma Predictive Index (API) and predictive validity by assessing subsequent physician diagnosis of asthma during the study period (1997-2015) and subsequent lung function measures for a random sample of the Birth cohort (n=221). We also assessed construct validity by assessing the association of NLP-PAC positivity with the known comorbidities (ie, other atopic conditions and pneumonia).

RESULTS: Of the eligible 8,196 study subjects (51% male, 80% White), median age (IQR) was 12 (9-14) years. Children who meet NLP-PAC (n=1,679, 20%), compared to those who do not, were more likely to meet API (68% vs. 9%), have a physician diagnosis of asthma (62% vs. 10%), have other atopic conditions (49% vs. 37%), have pneumonia (32% vs. 17%) (p<0.001 in all), and have lower FEV1/FVC% (median [IQR], 93 [87, 99] vs. 98 [92, 101], p=.003).

CONCLUSIONS: Two thirds of children who meet the PAC prior to 4th birthday had a subsequent physician diagnosis of asthma and lower lung function during school age. NLP-PAC can be a useful population management tool for early identification of asthma and treatment among preschool children.