Should you build or buy Population Health Management?
Population health management software aggregates claims and clinical data across a health system's patient panel to stratify risk, close care gaps, run HEDIS-style quality measures and support value-based contracts. Health systems, ACOs and health plans in risk-based arrangements buy or build it to manage outcomes across a patient population rather than one visit at a time.
Buy the platform and own the data underneath it. Providence and Geisinger built enterprise data platforms and custom clinical programs in-house; Intermountain's comparable CKD program runs on a licensed vendor data lake (MDClone) rather than a self-build. None of it replaces a HEDIS-certified measure engine or a care-management workflow, and getting measure logic wrong across contract years is expensive. If you already run a lakehouse with measure-governance staff, build the analytics and keep the vendor for certified reporting and workflow.
Build it, buy it, or bridge?
When building makes sense
Geisinger built its own data warehouse spanning clinical and claims data, social determinants and patient-reported outcomes, with an in-house analytics team supporting population health directly, and Providence runs a comparable production data platform on Snowflake and Databricks with about 2,000 self-service users. Both are real examples of a health system owning the analytics and risk-stratification layer rather than depending entirely on a vendor, and the case is strongest when you already carry a mature data platform and the staff to run it. Intermountain shows the more common pattern even among sophisticated systems: its chronic kidney disease program, which identified more than 80,000 patients and drove an 86 percent reduction in unplanned hospitalizations and roughly $6 million in savings in its first year, runs on MDClone's longitudinal data lake, a licensed vendor platform, not an in-house build. What no health system's build replaces is an NCQA-certified HEDIS measure engine or the chart-chase workflow that resolves ambiguous cases against payer claims. Build the cohorting and targeting logic where you have real data maturity, and keep certified measure calculation on a vendor regardless.
When buying makes sense
Buy the platform for certified quality measurement and care-team workflow, and own the data layer underneath it whenever you already have the staff to run one. Arcadia, Innovaccer, Epic Healthy Planet and Health Catalyst all carry the NCQA-certified measure specifications, chart-chase logic and care-management workflow that a health system would otherwise have to build and keep current every contract year as specifications change. The strongest independent build on record, at Geisinger, is an enterprise data platform layered beside a vendor relationship, not a full replacement for one; Providence runs a comparable platform on Snowflake and Databricks, while Intermountain's targeted clinical program runs on a licensed vendor data lake (MDClone) rather than an in-house build. If you do not already run a mature data platform with measure-governance staff, buying the whole platform, data layer included, is the faster and safer path; if you do, buy the certified reporting and workflow pieces and build the analytics you can actually maintain correctly year over year.
The desk read
Build-versus-buy analysis for Population Health Management is being written. In the meantime, the framework that drives every B4 call is on the B4 Index page.
Frequently asked
Have any health systems actually replaced their population health vendor with an internal build?
Not entirely. Geisinger built its own in-house analytics capability, and Providence runs a comparable production platform on Snowflake and Databricks, but neither is documented as having replaced a certified HEDIS measure engine or care-team workflow application. Intermountain's chronic kidney disease program is often cited alongside these, but it runs on MDClone's licensed data lake rather than an in-house build. The builds that do exist sit alongside vendor tools for certified reporting, not instead of them.
Why does the certified measure engine matter so much?
Quality measure specifications like HEDIS change and get re-certified regularly, and getting the calculation wrong affects shared-savings payments across an entire contract year. A vendor that maintains NCQA certification carries that governance burden continuously; replicating it internally means taking on measure-specification tracking as a permanent function, not a one-time build.
What's actually worth building versus buying in this category?
Cohorting, risk stratification and outreach targeting are worth building if you already run a mature data platform, since that is where competitive advantage in a value-based contract shows up. Certified measure calculation, chart-chase workflow and the care-team application are better left to a vendor whose core business is keeping those current.
How much does a population health platform typically cost?
Base platform pricing has been reported around a few hundred thousand dollars a year per 100,000 members before add-on modules for care management or health-plan reporting, though exact pricing varies by vendor and scope. Systems that already run their own analytics layer can often avoid the add-on modules and license only the certified reporting core.