RetentionHealth
Cohort-level analytics and behavioral signal system for GLP-1 telehealth programs that detects early patient drop-off risk without handling PHI.
RetentionHealth
Execution System
How cohort-level outcomes are tracked, behavioral signals are captured and aggregated, and risk scoring is computed without handling sensitive patient data.
The Problem
RetentionHealth is built on a PHI-avoidant architecture that eliminates HIPAA overhead while maintaining analytical power. The system uses token-based identity abstraction—participants access the platform via anonymous tokens with no personal identifiers stored. The backend runs on Cloudflare Workers with multi-worker architecture, REST APIs for participant/clinic/admin flows, and cron-based aggregation pipelines. Data is stored in three separate Cloudflare D1 databases for strict multi-tenant isolation. The event-driven data model captures raw behavioral events (engagement, sentiment, concerns) which are aggregated nightly into cohort-level signals and risk scores.
The System
The system architecture prioritizes deterministic scoring over machine learning for auditability and explainability. Cohort-level aggregation replaces individual tracking, enabling compliance and scalability. The workflow flows from clinic diagnostic submission → admin approval → cohort creation → participant token generation → behavioral event logging → nightly aggregation → risk scoring → clinic dashboard monitoring. Batch processing handles aggregation rather than real-time computation, balancing performance with simplicity. The enum-only data model prevents unsafe input and maintains data integrity across the multi-tenant platform.
How It Works
INPUT → LOGIC → EXECUTION → OUTPUT
INPUT: Clinic submits diagnostic, creates cohorts, uploads participant data → LOGIC: System generates anonymous tokens, captures behavioral events, applies signal weighting and risk aggregation logic → EXECUTION: Nightly cron processes raw events into cohort signals, computes stabilization metrics and risk scores → OUTPUT: Clinic dashboard displays cohort health, early warning signals, and ROI metrics for intervention and reporting.
What It Governs
How cohort-level outcomes are tracked, behavioral signals are captured and aggregated, and risk scoring is computed without handling sensitive patient data.
Result
Clinics detect and respond to early drop-off risk with actionable cohort-level insights while maintaining HIPAA compliance.
Past Builds
Explore more projects from our studio.
