Case Study: Scoring and Segmenting Key Opinion Leader Physicians With Innovative and Customizable Machine Learning Techniques

The Situation

A pharmaceutical company with an existing cancer therapy recently received approval to label its treatment for first-line use. With the potential to reach a new market, the client aimed to identify priority healthcare providers (HCPs) likely to adopt its therapy as a first-line treatment.

An EVERSANA client sought to determine the most influential key opinion leaders based on factors such as patient and prescription volume, early adoption of comparable products and influence metrics related to scholarly activities (e.g., involvement in clinical trials, publications and participation in scientific meetings).

The Solution

The key challenges included identifying precise metrics to pinpoint key opinion leaders in this specific therapy area, consolidating information from various sources and using this data to generate meaningful, targeted segments of healthcare providers. This process required extensive data integration, technical and clinical expertise, and the use of customizable machine learning (ML) techniques. Customizable ML refers to the development and application of ML models that can be tailored to meet the specific requirements and preferences of a particular use case or user, allowing them to influence model behavior based on their unique requirements or domain knowledge.

EVERSANA collaborated with the client to generate the patient cohort and understand their treatment journeys to identify relevant HCPs. Data on HCPs was gathered from diverse sources, including claims, clinical trials, open payments and publications. Scores were calculated for each HCP based on these data points. Finally, EVERSANA employed ML-based clustering to create segments of HCPs with varying priorities for targeted marketing.

The Results

EVERSANA generated personalized influence scores for over 30,000 HCPs in the target therapy domain using a wide variety of data. We employed unsupervised learning techniques to generate f ive HCP segments, with additional priority placed on those HCPs most likely to adopt the product for first-line treatment, including Elite (top 2%) and High (top 7%) priority HCP targets that maximized value for the client.