Learning Health System Implementation: Building a Hub-and-Spoke Model for Hypertension Management Through the QI Hub

In July 2022, the Ohio Department of Medicaid (ODM) launched a statewide Regional Quality Improvement (QI) Hub program using a hub-and-spoke model to embed Learning Health System (LHS) capabilities in primary care. The Ohio State University (OSU) Hub piloted the approach across ten affiliated sites, aiming to raise overall hypertension control by ≥ 10 percentage points and promote more consistent levels of hypertension control across patient populations by June 2027. This report describes the first 18 months post-implementation.

A mixed-methods, formative evaluation followed the Exploration-Preparation-Implementation-Sustainment (EPIS) framework. The hub provided centralized data extraction (bi-weekly Epic pulls), an R Shiny dashboard, and tailored coaching on Plan-Do-Study-Act cycles (PDSA) while participating in statewide learning collaboratives. Blood pressure (BP) control rates were tracked with statistical process control (SPC) charts. Site-level logistic regressions tested interaction effects between demographics (non-Hispanic Black vs. White) and implementation period.

Among 22,563 hypertensive adults (73,264 encounters), the baseline centerline was 72.9% controlled BP. Two SPC shifts, April 2024 to 77.3% and July 2024 to 79.1%, produced a + 6.2 percentage-point (+ 8.5%) relative improvement, equating to ~ 4542 additional controlled BP encounters beyond baseline expectations. Site-level absolute changes ranged from -3.9% to + 14.6%. Across sites, six narrowed and four widened the Black-White BP control gap. Site-level sensitivity analysis showed no significant change in the gap. Qualitative data highlighted the importance of near real-time feedback and flexible coaching; staffing turnover constrained progress in several under-resourced clinics.

Implementing a hub-and-spoke LHS model within a large academic health system was feasible and associated with sustained gains in hypertension control over 18 months. Centralized analytics, adaptive QI support, and statewide peer learning were key enablers, whereas workforce instability remained a barrier. Early results support the model's potential to build LHS capacity in primary care, with ongoing work needed to strengthen uniform impact and long-term sustainability.