AI Use Case for Overcoming Access Hurdles

Impact

Increased brand share by 4% over two years through monthly trigger-based physician targeting

Reduced payer rejections and claim reversals by 20% by identifying anonymized patients likely eligible for reimbursement

Within six months of operationalizing, identified (anonymously) 5000 emerging PCSK9 patients translating to $60M incremental annual revenue potential

Background

A newly launched PCSK9 inhibitor cardiovascular brand was experiencing slow uptake and high rejections across payer plans. However, some patients with access were benefiting from the therapy.

Objectives

Determine “look-alike” emerging PCSK9 patients from open claims universe by matching key attributes against patients who obtained reimbursement. Setup monthly triggers at physician level for the identified patients who could access and benefit from the therapy

Solution

A big data cloud platform was setup for processing gigabytes of patient and claims data from APLD (Anonymous Patient Level Data) and EMR (Electronic medical records) sources. An ensemble of machine learning algorithms (Random Forest, SVM) was deployed for feature engineering and predictive modelling on a large number of variables across the continuum of care for patients.

The model assigned accurate probability scores for look-alike patients to seek the new therapy and overcome access hurdles. The identified patients were then mapped to HCPs for monthly trigger-based targeting.