.abstract img { width:300px !important; height:auto; display:block; text-align:center; margin-top:10px } .abstract { overflow-x:scroll } .abstract table { width:100%; display:block; border:hidden; border-collapse: collapse; margin-top:10px } .abstract td, th { border-top: 1px solid #ddd; padding: 4px 8px; } .abstract tbody tr:nth-child(even) td { background-color: #efefef; } .abstract a { overflow-wrap: break-word; word-wrap: break-word; }
A4831 - Searching the Best Statistical Model Predicting Asthma Exacerbation in Elderly
Author Block: H. Park1, S. Kim2; 1Seoul National University College of Medicine, Seoul, Korea, Republic of, 2Seoul National University Hospital, Health Care System, Gangnam Center, Seoul, Korea, Republic of.
Introduction: Data mining delivers insights, patterns, and descriptive and predictive models from the large amounts of data available today in many disorders. The purpose of this study is to search the best statistical model predicting asthma exacerbation in elderly Methods: We collected data of subjects participating a prospective, observational, and multi-centered cohort of elderly asthmatics aged 65 years or older which was described in our previous report. The primary outcome was the occurrence of an exacerbation during the first year of follow-up, defined as any hospitalization, urgent visit, or systemic steroid course for asthma. We tried to develop statistical models to predict exacerbations using decision tree, random forest, and logistic regression methods with 13 baseline characteristics: age, gender, control status, history of previous exacerbation, presence of rhinitis, smoking status, atopy, body mass index, % of FEV1 predicted, cognitive function, depression status, medication compliance, and proficiency in using inhaler devices. Results: Six hundred twenty eight elderly asthmatics were followed for one year and 137 of them (21.8%) experienced one or more asthma exacerbations. We found that a model developed by random forest was the best one showing better predictive performance compared to those developed by decision tree and logistic regression methods provided. Interestingly, depression status, medication compliance, and proficiency in using inhaler devices at baseline were identified in all three models and thus thought to be important variables in predicting future asthma exacerbations in elderly asthmatics. Conclusions: A model developed by random forest was the best model to predict asthma exacerbation in elderly. This model may be effectively used for mining of clinical variables of elderly asthmatics.