healthsystemCIO
Corewell Health’s Jarve Says Population Health Data Challenges Demand Internal Builds
Why It Matters
Understanding the trade‑offs between vendor solutions and internal builds helps health systems make strategic decisions about data infrastructure, especially as they aim to improve population health outcomes and reduce clinician burnout. As AI tools become more accessible, the episode highlights how they can streamline analytics but also underscores the importance of clean, well‑structured data and integrated workflows.
Key Takeaways
- •Corewell built internal patient‑centered longitudinal data model.
- •Complex clinical data required custom solution; vendors couldn’t scale.
- •Epic lacks native ambient AI; Corewell uses external LLMs.
- •Governance bridges IT and clinicians to align pop‑health goals.
- •AI tools reduce documentation time but need discrete data extraction.
Pulse Analysis
Corewell Health faced a fragmented data landscape across its 23 hospitals, legacy systems, and varied documentation practices. Recognizing that traditional EHRs, designed for fee‑for‑service encounters, could not aggregate longitudinal clinical metrics, the organization engineered a patient‑centered longitudinal data model covering its 3.2 million active lives. By internalizing the data model, Corewell avoided costly vendor customizations and gained control over unique data elements such as ejection fractions and social determinants, enabling real‑time population health insights that standard Epic registries alone could not provide.
The team’s strategic choice to remain an Epic‑first shop did not mean relying solely on out‑of‑the‑box functionality. While Epic’s Chronicles and Caboodle feed core registries, Corewell extended these pipelines with custom analytics and a proprietary population health data model (PhDM). This architecture allows insights to flow back to point‑of‑care workflows, reducing cognitive load for clinicians. The discussion highlighted the distinction between “complicated” problems—suitable for off‑the‑shelf claim groupers—and “complex” clinical data challenges that demand bespoke solutions, reinforcing the importance of internal capability when vendor ROI is unclear.
Artificial intelligence, particularly large language models, is now accelerating Corewell’s analytics. An ambient AI bridge auto‑generates visit notes, cutting physician documentation time, but the extracted data must be transformed into discrete, trustworthy fields for the PhDM. The organization is experimenting with LLMs to auto‑write SQL queries against its data model, promising faster population health assessments. However, concerns about HIPAA compliance, data cleanliness, and Epic’s lag in native AI features remain. Corewell’s governance model—linking IT, operations, and clinical leadership—ensures that any external AI or vendor solution is evaluated for a one‑to‑three‑year horizon, balancing stickiness against the friction of change.
Episode Description
Most health systems trust vendors to solve their population health data problems. Corewell’s Associate CMIO explains why that approach collapsed and what his team built instead to bridge the gap between clinical complexity and enterprise analytics.
Source: Corewell Health’s Jarve Says Population Health Data Challenges Demand Internal Builds on healthsystemcio.com - healthsystemCIO.com is the sole online-only publication dedicated to exclusively and comprehensively serving the information needs of healthcare CIOs.
Comments
Want to join the conversation?
Loading comments...