Abstract by Zoe Gibbs
Using Asymmetric Cost Matrices to Optimize Wellness Intervention
Recently, insurance companies have started to implement intervention programs that may help reduce the medical costs of extremely high cost members. For these programs to be effective, the insurance company must identify and select potential high cost members to be assigned to an intervention before they incur high costs. We explore the use of machine learning, specifically the extreme gradient boosting algorithm, in calculating risk scores for members based on demographic, medical, and financial histories. To select members for intervention, we develop asymmetric cost matrices that model potential savings for assigning interventions to members with high risk scores. The cost matrices can be reduced to a function of the expected savings per dollar of intervention, which is easily used to optimize the threshold at which members with risk scores greater than the threshold are assigned an intervention. These techniques may help insurers select the optimal members for intervention programs and reduce overall costs.