domain-healthcare:healthcare-analytics

Healthcare Analytics

When to use

• Building risk stratification models or care gap identification pipelines
• Implementing clinical decision support alerts and managing alert fatigue
• Calculating eCQM, HEDIS, or MIPS quality measures
• De-identifying PHI for analytics or research workloads
• Setting up an OMOP CDM data lake or federated research environment
• Incorporating SDOH data into care management workflows

Core principles

1. Risk models need clinical validation — HCC and LACE scores are starting points; tune thresholds against your actual population before driving care workflows
2. Override rate is your alert quality metric — any alert dismissed more than 90% of the time is noise; retire it or redesign it
3. Safe Harbor is a checklist, not a guarantee — removing 18 identifiers is necessary but not sufficient; verify residual re-identification risk before publishing data
4. OMOP CDM enables federation — standardize to OMOP once and run the same analytics across institutions without sharing patient-level data
5. SDOH belongs in the risk model — food insecurity and housing instability predict readmissions as reliably as clinical comorbidities

Reference Files

• references/population-health-cds.md — risk stratification models (HCC, LACE, Charlson), care gap identification, disease registries, CDS alert types, alert fatigue management, order sets
• references/quality-measures.md — eCQM with CQL and QRDA, HEDIS measure calculation, MIPS quality reporting and payment adjustments
• references/deidentification-research-data.md — HIPAA Safe Harbor and Expert Determination, limited datasets and DUAs, OMOP CDM structure, data lake zone architecture, outcomes analysis methods
• references/sdoh-data.md — SDOH categories, PRAPARE and AHC-HRSN screening tools, ICD-10 Z-codes, FHIR Observation storage, community resource referral integration