BR2H
Behavioral Risk to Health
Validation & Evidence Library · Volume IV, 2024
The dossier benefits consultants, brokers, and risk committees forward to actuarial review — Milliman-audited savings, the 91.3% AUC methodology, 47 peer-reviewed publications, and the governance framework that holds it together.
Headline figures · Audited
Each figure below traces to a named methodology, study period, and reviewing body — no marketing rounding.
Methodology · Chaptered Brief
A citable technical brief for actuarial review. Each chapter stands alone; together they describe a defensible model with a documented governance surface.
The model ingests continuous CGM, HRV, and step-cadence from consented wearables; FHIR-encoded EHR observations; and adjudicated medical, pharmacy, and dental claims. Every signal is timestamped, source-tagged, and routed through a HIPAA-compliant single-tenant lake with HITRUST CSF v11 attestation.
Model coefficients were trained on 1.7M members and locked before prospective validation across a held-out cohort of 2.4M. The reference AUC of 91.3% for 30-day readmission is reported on the prospective arm only, with confidence intervals published in the State of Workforce Risk 2024 appendix.
Reference: Chen M., Okafor L., et al., JMIR Medical Informatics, 2024.
BR2H-Score outputs are passed through isotonic regression against observed 12-month high-cost events. Calibration refreshes every 90 days using the prior quarter's adjudicated outcomes, with drift monitored via population stability index and PSI-flagged sites returned to the modeling team for retraining review.
A 38-person clinical advisory board signs off on every model release. The platform is SOC 2 Type II, HITRUST CSF v11, ISO 27001, and HIPAA-compliant, audited annually by Coalfire. Production deployments carry a 99.982% uptime SLA, verified by independent third-party monitoring.
Comparison · RFP-ready
A defensible matrix benefits consultants can paste into an RFP response. Dimensions mirror the criteria most risk committees score on.
| Dimension | BR2H-Score™ | Traditional claims-only risk-adjustment | Consumer wellness & step apps |
|---|---|---|---|
| Predictive horizon | 12-month behavioral-risk score | Retrospective risk-adjustment | Daily engagement metrics only |
| Signal breadth | 180+ passive streams (CGM, HRV, cadence, EHR, claims) | Adjudicated claims only | Steps, sleep, self-report |
| Validation evidence | 91.3% AUC · 2.4M lives · Milliman-audited | Actuarial tables, periodic refresh | Engagement studies, no clinical endpoints |
| Peer-reviewed base | 47 publications, Stanford PRC co-developed | Limited; rarely published | Marketing white papers |
| Compliance posture | HITRUST CSF v11 · SOC 2 II · ISO 27001 · HIPAA | HIPAA-aligned, varied attestation | Consumer-grade, rarely HITRUST |
| Audited economic outcome | $2,840 PEPY reduction · 17 deployments · Milliman | Cohort modeling, not deployment-audited | Engagement KPIs, no claims delta |
Compiled from public methodology disclosures as of Q4 2024. Send corrections to [email protected].
In seventeen years of auditing population-health economics, I have not seen a predictive model whose deployment evidence matched its validation claims so cleanly. Milliman's independent review of the BR2H cohort returned a PEPY delta that is, frankly, unusual for a 12-month behavioral-risk instrument.
Publication Index · Selected, 2021–2024
Four representative rows from the 47-paper catalog. Full index available to qualified research collaborators on request.