Back to Module 5: Population Health and Equity

Lesson 2

Data for Population Health

About 4 min

Claims, clinical records, and social data each tell part of the story. This lesson covers what each source can and cannot do.

Population health runs on three questions asked continuously: who is on our list, what is happening to them, and who needs something next. Answering them means stitching together data sources never designed to work together.

The sources, honestly compared

SourceStrengthWeakness
ClaimsComplete across all settings; structured; runs benchmarks, risk scores, attributionWeeks-to-months lag; clinically thin (a billed test, not its result)
EHR / clinicalRich (labs, vitals, notes) and currentSees only care inside its own walls
ADT alertsNear real time: patient hit a hospital, anywhere in networkNarrow signal, needs a team ready to act
Social dataExplains what claims and charts cannotSparse capture; area-level indices are proxies, not people

Claims are broad and shallow; EHR data is deep and narrow. You need both, plus the connective tissue of health information exchange, which federal interoperability rules keep pushing toward standardized, API-accessible formats.

Worth remembering: ADT feeds are the single most valuable operational data source in population health. They trigger the transitions-of-care work that prevents readmissions while there is still time to do it.

Social data: the missing determinants

Housing instability, food insecurity, transportation, and isolation drive utilization but appear in neither claims nor charts. Screening programs and standardized codes exist; capture remains thin. Area-level deprivation indices attached to a patient’s address are a workable proxy for stratification and equity reporting, with the caveat that a neighborhood average is not a person. The gap matters twice: unrecorded social risk skews risk adjustment, and unseen need is unmet need.

Three disciplines that decide success

  1. Data quality is a program, not a project. Eligibility files disagree with attribution lists, duplicates accumulate, feeds silently break; someone must own reconciliation.
  2. Latency must match the use. ADT within hours, gap lists within days, spending analytics on the claims cycle. A monthly report cannot drive a daily workflow.
  3. The last mile decides everything. Information that does not land inside the workflow of someone empowered to act (the care manager’s queue, the huddle sheet, the EHR flag) is analytics as decoration.

Key takeaways

  • Claims are complete but slow and thin; clinical data is rich but walled; you need both.
  • Social drivers are mostly invisible in both; proxies help but have limits.
  • Success is measured at the last mile: the right name in front of the right person, in time to act.

Check your understanding

Which data feed is generally the most operationally valuable in population health, and why?

Share