OBJECTIVES : To explore the patients’ characteristics associated with hospitalization through emergency department (ED) visit with asthma exacerbation by using association rule mining (ARM)

METHODS : A retrospective cross-sectional study was performed using ED visit cases from National Emergency Department Information System database 2016. In 2016, 9,127,979 ED visit cases collected from 410 EDs throughout the nation. We included ED visit cases with discharge diagnosis of asthma (ICD-10 code of J45/J46) as the main diagnosis. ARM was performed with Apriori algorithm to explore association rules between patients’ characteristics and hospitalization through ED visit. Hospitalization includes admissions to general ward and intensive care unit. The association rules (A=>B) meant that if a person has a feature A, a person also has a feature B (hospitalization). We used support, confidence, and lift to select interesting rules. Support was defined as the proportion of an itemset in the data. Confidence was used as a measure of reliability. Lift was used to estimate whether the occurrences of A and B were independent or not. The higher confidence, the more likely to have an interesting relation between A and B. The higher lift (lift >1), the more likely that there was a dependency between A and B. We eliminate redundant rules that have similar meanings.

RESULTS : We included 37,350 ED visit cases for the analysis. We extracted association rules with a support ≥0.5%, a confidence ≥80% and a lift >1. After eliminating redundant rules, the rule with the highest confidence and lift was the rule ({over-80-year-old, female, comorbid with pneumonia}=>hospitalization) with confidence: 80.9% and lift: 2.94.

CONCLUSIONS : Among patients visited the ED with asthma exacerbation, elderly, female, and comorbid with pneumonia were importantly associated with hospitalization. Cautious management is needed for patients with asthma having these characteristics.