lev-os Managing Population Health Analytics
Structures population health analysis with risk stratification, care gap identification, and intervention targeting for value-based care programs. This skill covers the analytic workflow from population definition through outcome measurement, using clinical, claims, and social determinants data.
Why This Skill Exists
Value-based care models (ACOs, bundled payments, capitation, MSSP) require organizations to manage the health of defined patient populations rather than individual encounters. Without structured analytics, organizations cannot identify which patients are high-risk, which preventive services are overdue, where care coordination is failing, or whether interventions are improving outcomes. Population health analytics transforms raw EHR and claims data into actionable intelligence for care managers, quality teams, and executive leadership. Poor analytics leads to misallocated care management resources, missed quality measure targets, and financial losses under risk-bearing contracts.
The financial stakes are significant: Medicare Shared Savings Program (MSSP) ACOs manage over 11 million beneficiaries, Medicare Advantage enrollment exceeds 30 million, and commercial value-based contracts are expanding. Organizations without population health analytics capability cannot identify rising-risk patients before they become high-cost, cannot close care gaps efficiently, and cannot demonstrate value to payer partners.
Checkpoint A --- Intake & Scoping
Answer every question before proceeding. Mark unknowns with [VERIFY].
- Population definition --- How is the population defined? (Attributed lives under an ACO contract, empaneled patients in a primary care panel, all patients with a specific chronic condition, health plan enrolled members)
- Contract and program context --- Which value-based program applies? (MSSP, Medicare Advantage, commercial ACO, Medicaid managed care, employer direct contract)
- Data sources --- What data is available? (EHR clinical data, claims/encounter data, pharmacy dispensing, ADT alerts, HIE data, SDOH screening data, patient-reported outcomes)
- Quality measures --- Which quality measures are tied to the program? (CMS Web Interface measures, HEDIS measures, MIPS quality measures, custom contract measures)
- Risk stratification model --- What risk model is in use or planned? (HCC, LACE, Charlson Comorbidity Index, Johns Hopkins ACG, proprietary vendor model)
- Care management capacity --- How many care managers/coordinators are available? What is their current caseload?
- Reporting requirements --- What reports are required, how often, and for whom? (Monthly quality dashboards, quarterly contract reports, annual HEDIS submission)
Required Documents
- Value-based contract terms with quality measure specifications and financial benchmarks
- Patient attribution or enrollment file with effective dates
- Quality measure technical specifications (CMS eCQM specifications, HEDIS technical specifications)
- Risk stratification model documentation and scoring methodology
- Care management program description (eligibility criteria, intervention protocols, staffing model)
- SDOH screening tool and data collection workflow documentation
Step 1 --- Build the Population Registry
Create and maintain the master patient list:
- Attribution reconciliation --- Match attributed/enrolled patients to the organization's EHR patient records. Use deterministic matching (MRN, insurance member ID) supplemented by probabilistic matching (name + DOB + address) for unmatched records
- Panel assignment --- Assign patients to primary care providers based on attribution logic (most recent visit, contractual assignment, patient selection). Flag patients without a PCP assignment for outreach
- Status tracking --- Maintain patient status: active (engaged in care), attributed but not seen (gap-in-care), recently hospitalized, recently discharged from ED, deceased, disenrolled
- Demographic stratification --- Segment the population by: age group, sex, race/ethnicity (for health equity analysis), primary language, insurance type, zip code (for SDOH proxy analysis)
- Data integration --- Link EHR clinical data, claims data, pharmacy data, and ADT notifications to each patient record. Resolve conflicts between data sources (e.g., claims say patient had A1c test; EHR has no result --- indicates the test occurred outside the network)
Step 2 --- Execute Risk Stratification
Assign risk scores to stratify the population:
HCC risk scoring --- For Medicare populations, calculate Hierarchical Condition Category (HCC) risk scores using the CMS-HCC model (V28 for payment year 2025+). Input: patient demographics + ICD-10-CM diagnosis codes from eligible encounters
Clinical risk models --- Apply condition-specific risk tools: LACE score for 30-day readmission risk, ASCVD pooled cohort equations for cardiovascular risk, PHQ-9/GAD-7 for behavioral health severity
Utilization-based stratification --- Layer utilization data: ED visits in past 12 months, inpatient admissions, specialist visits, prescription fills, total cost of care. High utilizers with modifiable conditions are priority intervention targets
SDOH risk factors --- Incorporate social determinants: food insecurity, housing instability, transportation barriers, social isolation (from SDOH screening using AHC HRSN or PRAPARE tools coded with LOINC and SNOMED CT)
Risk tier assignment --- Classify patients into actionable tiers:
- Tier 1 (Rising Risk) --- 2+ chronic conditions, moderate utilization, identifiable care gaps. Intervention: proactive care coordination
- Tier 2 (High Risk) --- Complex multimorbidity, recent hospitalization, high utilization. Intervention: intensive care management
- Tier 3 (Complex/Critical) --- Highest-risk patients with behavioral health comorbidity, SDOH barriers, and frequent acute utilization. Intervention: multidisciplinary team-based care
Behavioral health integration --- Layer behavioral health risk factors into the stratification model: PHQ-9/GAD-7 scores, substance use screening results (AUDIT-C, DAST-10), behavioral health ED visits, and psychiatric hospitalization history. Behavioral health comorbidity is the strongest modifier of total cost of care for chronic disease populations
Step 3 --- Identify Care Gaps
Systematically detect missing preventive and chronic care services:
Quality measure gap analysis --- For each quality measure, identify patients in the denominator who are not yet in the numerator. Example: HEDIS Comprehensive Diabetes Care --- patients with diabetes who have not had HbA1c testing (LOINC 4548-4) in the measurement year
Preventive care gaps --- Screen for overdue services: mammography (age 50-74, every 2 years), colorectal cancer screening (age 45-75), influenza vaccine (annual), pneumococcal vaccine (age 65+), annual wellness visit (Medicare)
Medication adherence gaps --- Using pharmacy claims data, calculate Proportion of Days Covered (PDC) for chronic medications: statins (target PDC >= 80%), antihypertensives (>= 80%), oral diabetes medications (>= 80%). PDC < 80% is a care gap
Transition-of-care gaps --- Following hospital discharge, identify patients who have not completed: post-discharge PCP visit within 7 days (TRC measure), medication reconciliation, follow-up labs if ordered at discharge
Gap prioritization --- Rank gaps by: (a) patient risk tier (address Tier 2-3 first), (b) measure weight in the contract (higher financial impact first), (c) clinical urgency (overdue cancer screening vs. overdue wellness visit)
SDOH-driven care gaps --- Identify care gaps attributable to social determinants: transportation barriers preventing appointment attendance, food insecurity undermining diabetes management, housing instability affecting medication storage and adherence, and health literacy barriers preventing self-management education effectiveness
Step 4 --- Generate Actionable Outputs
Transform analytic findings into operational tools:
- Care manager worklists --- Prioritized patient lists by risk tier with: patient name, PCP, risk score, open care gaps, last encounter date, preferred contact method. Refresh daily or weekly based on care management workflow
- Provider scorecards --- Per-PCP dashboards showing: panel size, risk distribution, measure performance rates, care gap closure rates, utilization trends. Benchmark against peer providers within the organization
- Outreach lists --- Patient contact lists for specific campaigns: diabetes A1c overdue, mammography screening, annual wellness visit scheduling. Include patient communication preferences and language
- Executive dashboards --- Aggregate population metrics: total attributed lives, risk score distribution, overall quality measure performance, total cost of care trend, hospitalization rate, ED utilization rate. Compare against contract benchmarks and prior periods
- Equity analysis --- Stratify all metrics by race, ethnicity, language, and geography to identify health disparities. Flag measures where performance varies by > 5 percentage points across subgroups
Step 5 --- Measure Intervention Outcomes
Close the loop by measuring whether interventions improve outcomes:
Pre/post comparison --- For patients who received care management intervention, compare utilization and quality measures before and after enrollment. Use appropriate statistical methods (paired t-test for continuous measures, McNemar's test for binary outcomes)
Control group analysis --- When possible, compare intervened patients against similar non-intervened patients using propensity score matching or difference-in-differences design
Return on investment --- Calculate the cost of the care management program (staffing, technology) against the avoided cost (reduced hospitalizations, ED visits, unnecessary specialist referrals). Present per-member-per-month (PMPM) impact
Quality measure trajectory --- Track monthly quality measure rates to detect whether care gap closure is on pace to meet end-of-year targets. Project year-end performance and flag measures at risk of missing targets by mid-year
Contract performance projection --- Use analytics to project year-end contract performance: projected quality measure rates vs. targets, projected total cost of care vs. benchmark, estimated shared savings or losses, and resource allocation recommendations to close performance gaps before year-end
Checkpoint B --- Analytics Review
Before releasing population health reports:
Population registry matches the official attribution/enrollment file within 1%
Risk stratification model is applied consistently with documented scoring methodology
Care gap identification uses current measure specifications and value sets
Outreach lists exclude patients who are deceased, disenrolled, or have opted out of contact
Provider scorecards use appropriate risk adjustment when comparing across panels
Equity analysis is included for all reported measures
Data sources are current (claims data lag is documented and accounted for)
Reports are reviewed by clinical and quality leadership before distribution
Behavioral health risk factors are integrated into stratification models
SDOH-driven care gaps are identified and addressed with targeted interventions
Contract performance projections are generated at least quarterly with actionable recommendations
Quality Audit
Population definition matches the contractual attribution methodology
HCC risk scores use the correct CMS-HCC model version for the payment year
Quality measures use the exact CMS/HEDIS/MIPS specification including exclusions
Care gap identification accounts for services rendered outside the organization (via claims data)
Risk stratification tiers have defined intervention protocols matched to available capacity
Medication adherence calculations use PDC (not MPR) per CMS Star Rating methodology
SDOH data uses standardized coding (LOINC for screening instruments, SNOMED CT for findings)
Intervention outcome analysis uses appropriate comparison methodology
All reports include data freshness dates and known limitations
Behavioral health screening data (PHQ-9, GAD-7, AUDIT-C) is incorporated into risk stratification
SDOH care gap analysis links social barriers to specific clinical care gaps
Contract performance projections use validated methodology with documented assumptions
Analytics outputs are validated against final contract settlement data for accuracy assessment
Guidelines
Attribution is not panel management. Attributed patients who have never visited the organization require different outreach strategies than established patients with care gaps
Claims data has a 60-90 day lag. Never make real-time clinical decisions based on claims alone; use EHR data for current clinical status and claims for historical utilization context
Risk scores are predictive, not prescriptive. A high-risk score identifies patients who may benefit from intervention, but clinical judgment determines the appropriate intervention type
Quality measure "numerator hits" from claims may not be visible in EHR data and vice versa. Use all available data sources for the most complete picture
Care gap outreach without culturally appropriate communication is ineffective. Tailor outreach by language, communication preference, and health literacy level
Do not report race/ethnicity-stratified data without sufficient sample sizes per subgroup (minimum n=30). Small denominators produce unreliable rates that can mislead equity improvement efforts
SDOH screening data is sensitive and requires appropriate consent and data governance. Not all SDOH findings should be shared across all care team members
Refresh risk stratification at least quarterly. Annual-only risk scoring misses patients whose risk profile changes mid-year due to new diagnoses, hospitalizations, or life events
Social determinants data is the most underutilized dimension in population health analytics. Organizations that integrate SDOH into risk stratification and care gap identification consistently outperform those relying solely on clinical and claims data
Analytics without action is waste. Every analytic output should have a defined clinical workflow response: who sees the data, what decisions it informs, and how outcomes are tracked back to the intervention