Patient Modeled Audiences
Stefanie DeSantis, Director Growth Marketing
Patient Modeled Audiences (PMA) is a patent-pending audience targeting solution that is HIPAA-certified and uses a unique approach to reach specific patient populations in a privacy-safe way. In fact, privacy is the cornerstone of the solution. In a novel approach, DeepIntent took an emerging technology called differential privacy and uses it to safely analyze and build patient lookalike models from clinical data without any risk of identifying individual patients.
As a requirement, the system relies heavily on AI that autonomously operates within a data clean room where patient health data is inaccessible to any human. This means no one at DeepIntent or the client ever has access to actual patient data. The AI scans through anonymized patient data sets located behind a “privacy firewall” that is deliberately constructed to prohibit data from leaving the clean room. The process builds models that extract patterns about certain conditions. For example, you can ask the model to build an audience segment for type 2 diabetes patients and the AI will generate thousands of demographic personas such as “men over 45 years old, that live in the Northeast, and watch late night television.” Once finished, the audience models are rigorously verified to be HIPAA compliant and then exported to DeepIntent’s demand-side platform (DSP) where they are used guide the company’s ability to serve ads to likely patients.
Modeled audiences are completely unique to each client’s needs, and advertisers have full visibility into the quality of the targeting segment. As measured by third parties, PMA provides over 25% higher audience quality for patient campaigns than leading third-party health data providers activated within a generalized DSP like The Trade Desk.