Identification and Prediction of High Risk Complications of Asthmatic Conditions
Keywords: Asthma, complications, predictive modeling, evidence-based guidelines
In this presentation we demonstrate the methodology and advantages of applying predictive modeling techniques to identify and predict high risk asthmatic complications, and to forecast future total medical costs, total acute costs, emgency department visits and hospital admissions for the next one or two years. Trends/timelines regarding asthma complications, as well as savings opportunities for care management will also be explored.
This case study is conducted using three years of insurance claims data for medical and pharmacy claims. Members with asthma-related claims/diagnosis/treatment in the first year were selected from two million plus commercial members and over half a million Medicaid members were studied. There are 23,874 commercial asthmatic members and 24,293 Medicaid members included in the study.
The patterns and trends for Asthma conditions, complications and costs are studied for the first, second, third, and a combined two-year period (second and third year). The following high-cost complications of asthma are evaluated: pneumonia, GERD, acute bronchitis and bronchiolitis, acute and chronic respiratory failure, pulmonary congestion and hypostasis, acute edema of the lung, pulmonary collapse/atelectasis, pleural effusion, etc.
Multiple predictive models are constructed using various predictive modeling techniques, such as linear regression, logistic regression, decision trees and support vector machines, etc. These models predict future emergency department visits, hospitalizations, total costs, asthma costs, and pulmonary costs in the second and third year respectively, and then in the combined two year period. The impact of evidence-based guideline compliance and gaps is also studied, in particular its relationship with future complications and costs. The resutls show that Near half of Asthma conditions are co-morbidities and/or transitional. Predictve models can fair accurately prediction high risk complications and costs. It can lead to more appropriate care interventions and greater return-on-investment.