When it comes to deal with risk adjustment and population health management, payers and providers usually use Ordinary Least Square (OLS) linear regression. Even though many machine learning algorithms have high prediction accuracy, sometimes they lack in interpretation. Approaches like penalized linear regression with some machine learning fundamental features can be a good choice considering the interpretablity as well. Penalized linear regression outperformed OLS with full and reduced (selected by lasso) set of predictors, based on R-squared, RMSE and MAPE. The data used was obtained from IMS LifeLink Health Plan Claims Database. It consists of fully adjudicated and de-indentified medical and pharmaceutical claims from health insurance plans. The predictors were previous diseases and symptoms as indicated by recorded medical diagnoses and pharmacy claims data prior to 2013. The target outcome for all the predictive models were total healthcare costs in 2013.
Input variables : Diseases and symptoms recorded from medical diagnoses and claims data
Output Variables : Total healthcare cost
Visit Model : github.com
Additional links : journals.plos.org
| Model Category | : | Public |
| Date Published | : | November, 2018 |
| Healthcare Domain | : |
Payer
Provider |
| Code | : | github.com |
| Utilization Management |
| ED Admission |