Forecasting Future Asthma Hospital Encounters of Patients With Asthma in an Academic Health Care System: Predictive Model Development and Secondary Analysis Study (Preprint)

• DOI: 10.2196/preprints.22796
• Publication Date: 2021-04-16T13:29:05Z
• Authors: Yao Tong; Amanda I Messinger; Adam B Wilcox; Sean D Mooney; Giana H Davidson; Pradeep Suri; Gang Luo
• Keywords: Preprint

<sec> <title>BACKGROUND</title> Asthma affects a large proportion of the population and leads to many hospital encounters involving both hospitalizations emergency department visits every year. To lower number such encounters, health care systems plans deploy predictive models prospectively identify patients at high risk offer them management services for preventive care. However, previous do not have sufficient accuracy serving this purpose well. Embracing modeling strategy examining candidate features, we built new machine learning model forecast future asthma with Intermountain Healthcare, nonacademic system. This is more accurate than previously published models. it unclear how well our generalizes academic systems, whose patient composition differs from that Healthcare. </sec> <title>OBJECTIVE</title> study aims evaluate generalizability University Washington Medicine (UWM), an <title>METHODS</title> All adult who visited UWM facilities between 2011 2018 served as cohort. We considered 234 features. Through secondary analysis 82,888 data instances 2018, in subsequent 12 months. <title>RESULTS</title> Our yielded area under receiver operating characteristic curve (AUC) 0.902. When placing cutoff point making binary classification top 10% (1464/14,644) largest forecasted risk, 90.6% (13,268/14,644), sensitivity 70.2% (153/218), specificity 90.91% (13,115/14,426). <title>CONCLUSIONS</title> showed excellent UWM, leading AUC higher all AUCs reported literature forecasting encounters. After further optimization, could be used facilitate efficient effective allocation resources improve outcomes. <title>INTERNATIONAL REGISTERED REPORT</title> RR2-10.2196/resprot.5039