Making the Invisible Visible
AI-Driven Personalization for a Patient Support Program Serving 85,000 Chronic Disease Patients
Client
Top-5 Global Pharmaceutical Company
Engagement Type
Patient Support / Generative AI
Geography
United States
Program Scale
85,000 active patients
The Challenge
The immunology franchise operated one of the industry's largest patient support programs for a high-cost biologic therapy in psoriasis and psoriatic arthritis — a program with 85,000 active enrollees, a team of 420 patient support specialists, and an annual operating cost in excess of $120 million. The program delivered financial assistance, injection training, adherence support, and care coordination services to patients across the full therapy lifecycle.
Despite the program's scale and investment, the franchise's commercial analytics team had identified a persistent and commercially significant performance gap: patient persistence rates at the 12-month mark were 57%, compared to 74% in the pivotal clinical trials and 69% in the best-performing competitive support programs in adjacent disease areas. The 12-point gap between actual and achievable persistence represented, in commercial terms, a substantial volume of lost revenue — and, more importantly to the program's clinical leadership, 12 points of patients who were discontinuing a therapy that was working clinically, for reasons that were not primarily clinical.
A patient satisfaction survey analysis conducted 18 months prior to the Nightingale engagement had produced a finding that the program's leadership found uncomfortable: 62% of patients who had discontinued therapy within 12 months reported that their last contact with the patient support program had been at month 3 or earlier. The program was operating on a reactive contact model — reaching out when patients called in or when pharmacies generated refill alerts — and was losing contact with a substantial portion of its enrollee population well before the 12-month persistence threshold.
The Nightingale Approach
Patient Segmentation and Journey Mapping
The foundational analytical work was a comprehensive patient journey analysis using three years of the program's own interaction data, pharmacy fill records, patient-reported satisfaction survey data, and clinical claims from the franchise's real-world evidence database. The analysis mapped the behavioral patterns of patients who persisted through 12 months, patients who discontinued between months 3 and 6, and patients who discontinued after month 6 — identifying the behavioral signals in the program interaction data that differentiated these groups prospectively.
AI Content Personalization Engine
A Generative AI content personalization system was built to generate individualized patient communications calibrated to each patient's segment, current journey stage, most recent interaction history, and predicted risk status. The system operated within a compliance framework that constrained all generated content to approved claims library language, required MLR pre-screening of all message templates and personalization parameters, and maintained complete audit trail documentation.
Predictive Persistence Risk Engine
A predictive model was built to identify patients at elevated risk of discontinuation 60 to 90 days before the typical discontinuation event — early enough for proactive outreach to be meaningful. The model's features included pharmacy refill gaps, changes in patient support program contact frequency, specific interaction content patterns associated with pre-discontinuation dissatisfaction, and formulary change events that created coverage uncertainty.
The Results
The AI-driven personalization system was deployed across the full 85,000-member enrollee base at month 8 of the engagement, following a four-month pilot with 12,000 members that validated the personalization engine's performance and the compliance framework governing its operation.
At the 12-month post-deployment evaluation, the program's persistence rate at 12 months had increased from 57% to 68% — an 11-percentage-point improvement that represented approximately 9,400 additional patients persisting to the 12-month mark compared to the pre-implementation baseline. Patient satisfaction scores, measured through the program's quarterly survey, increased meaningfully — with the largest improvements on the dimension of "my support program understands what I need," rising from 41% agreement pre-implementation to 68% post-implementation.
The predictive persistence risk engine identified 7,200 at-risk patients in its first six months of operation. Of those patients reached through the enhanced proactive outreach protocol the risk prediction triggered, 58% remained on therapy at 90-day follow-up — compared to 31% for patients at similar predicted risk levels in the pre-implementation period who received standard reactive outreach only.
"Eleven percentage points of persistence. That's 9,400 patients who stayed on a therapy that was helping them. The personalization engine made every one of those patients feel like the program was talking specifically to them — because it was. That's what changes behavior."— Head of Patient Services, Global Immunology Franchise