Achieving highly accurate social media measurement without patient level integration

In the highly regulated pharmaceutical industry, evaluating digital campaign performance requires strict adherence to privacy and compliance standards. Social media companies are highly sensitive to these concerns and can pose the greatest challenges in measuring healthcare outcomes. Symphony Health, an ICON plc company, partnered with a leading social media company to develop a novel ad measurement solution for a large pharmaceutical company. The goal was to leverage Symphony’s rich anonymised patient level data while adhering to the higher privacy standard that the social media platform required. 

The Symphony Health data science team successfully modified its standard patient level integration methodology with unsupervised machine learning algorithms to achieve the higher privacy standard while maintaining 90% of the accuracy of its standard approach. This new method can readily be applied in any environment with similarly stringent privacy standards.

This study faced several unique obstacles due to platform limitations, scale, and compliance requirements: 

• No Access to Patient-Level Data: The social media company’s privacy policies prohibited sharing exposed user data, requiring an indirect measurement strategy.
• Massive Population Reach: The campaign reached millions of patients, creating challenges in identifying and isolating the campaign’s effects.
• Granular Segmentation Required: The pharma client wanted insight into specific target groups such as 18–30-year-olds, necessitating a refined measurement capability.
• Single-Channel Measurement: The analysis focused solely on the walled garden of the social media company, ruling out multi-channel attribution approaches.
• Unrestricted Campaign Reach: The pharma client aimed to run the campaign broadly, with minimal constraints on targeting or segmentation

Symphony Health developed a test vs. control methodology at the Designated Market Area (DMA) level, integrating campaign data with anonymised claims to evaluate the campaign’s impact while ensuring privacy and compliance. 

• Data Integration: The social media company provided DMA-level campaign data (e.g., impressions), and Symphony Health was able to integrate this data to national medical claims to track prescription behaviours over time.
• Test and Control Design: Using several machine learning (ML) techniques, DMAs were clustered based on patient demographics, diagnosis and prescription history, payment type, and HCP interaction history. Five DMA clusters were formed with one randomly selected and split into 70% test and 30% control groups. Over 1,000 iterations were conducted to match disease prevalence between groups, ensuring balanced comparison prior to campaign launch.
• Statistical Confidence: Symphony Health was then able to evaluate New to Brand (NTB), continued patients, and switchers, with a Z-test assessing statistical significance of conversion differences between groups

Symphony Health was able to show, with >99% statistical confidence:

• +7.8% Total Patient Lift in test group vs. control.
• +6.6% Increase in NTB Patients, showing strong acquisition.
• Additional insights were delivered by age, gender, and location, supporting further optimisation.

This case study underscores the flexibility and effectiveness of Symphony Health’s data-driven approach to delivering actionable campaign insights without patient level integration of exposure and claims data. This solution is adaptable to any large-scale, single-channel digital campaign, offering a repeatable framework for evaluating promotional effectiveness across social media and other highly privacy-sensitive platforms.