AI Use Case for Predicting Patient Compliance

Impact

ML solution predicted patients’ risk of non-compliance with treatment regimen, on a weekly basis

Pre-emptive intervention by Patient Services led to 18% increase in compliance across all of client’s brands. This translated to $74M incremental annual revenue

The Compliance Optimizer solution continued to be successfully deployed for the 3rd year post launch

Background

Compliance measures how closely a patient follows the prescribed frequency and dosage of a therapy. Decline in compliance had significant adverse impact on revenue for the client, a large rare-disease Biotech. However, past interventions occurred when patients had already discontinued therapy

Objectives

Build a dynamic Patient Journey solution with real-time tracking of KPIs

Enable the Patient Support Program (PSP) to proactively address refill or shipment delays and reimbursement hurdles which were causing patients to drop off therapy

Solution

All plausible drivers for patients’ non-compliance were identified based on real-time tracking by PSP. Multiple AI/ML classification techniques (Logistic Regression, Random Forest, Gradient Boosting, SVM) were comparatively evaluated using cross-validation accuracy metrics. The final model predicted each patient’s non-compliance risk on a weekly basis with an accuracy greater than 80%.

The weekly list of high-risk patients from a non-compliance perspective were targeted for pre-emptive patient-specific intervention by the Patient Services team leading to a turnaround in compliance rates across brands.