Data-Driven Payment-Model Recommendation System for Cost-Effective Healthcare Delivery

Sukumar, Sreenivas ;   Jackson, Indigo ;   Frank, Aline ;   Wheeler, Jessica ;   Tourassi, Georgia

The ability to associate an optimal insurance plan to a patient has the potential to ensure quality care at affordable prices without being a burden for both the payers and the patients. Towards that goal, we begin by understanding payment models and beneficiary attributes and behaviors. We have developed an insurance payment-model recommender system based on data-driven discoveries that can guide the patient-centric delivery of healthcare services. Based on a de-identified sample of real healthcare transactions from insurance claim?processing operations, we have designed metrics of patient behavior and utilization to predict future interactions with the healthcare system. Our feature set consists of the following categories of attributes: (i) demographics - age, race, sex, geography (ii) temporal patterns - recency, frequency, monetary, periodicity of visits (iii) health dispositions - number of diagnosed chronic conditions, number of procedures, number of diagnoses, number of providers, number of services, and (iv) diagnosis related clinical pathways - a set of frequent procedures and drugs. The feature extraction module feeds similarity estimation modules that mine for beneficiaries with similar attributes and behaviors. Based on rules extracted from discussions with health insurance and healthcare delivery experts we use the feature and similarity metrics to recommend optimal payment models (such as fee-for-service, managed care, payment for co-ordination, payment for per