Risk Stratification in Care Management

Risk stratification is a structured clinical and administrative process that sorts patient populations into discrete risk tiers based on their likelihood of adverse health outcomes, high-cost utilization, or preventable deterioration. Applied within care management models and frameworks, it determines which individuals receive intensive intervention and which require only monitoring or preventive contact. The accuracy of stratification directly affects resource allocation, quality benchmarks, and regulatory compliance across Medicare, Medicaid, and commercial payer programs.

Definition and scope

Risk stratification in care management is formally defined as the segmentation of a patient population into mutually exclusive risk categories — typically three to five tiers — using quantifiable clinical, behavioral, and social data. The Centers for Medicare & Medicaid Services (CMS) references risk stratification as a foundational component of the Chronic Care Management (CCM) framework and the Medicare Shared Savings Program (MSSP), both of which require participating accountable care organizations to demonstrate panel-level risk assessment capability (CMS MSSP Regulations, 42 CFR Part 425).

The scope of stratification extends beyond clinical diagnosis codes. The Agency for Healthcare Research and Quality (AHRQ) identifies four primary data domains used in rigorous stratification models:

• Clinical complexity — diagnosis burden, comorbidity counts, and active medication lists
• Utilization history — emergency department visits, inpatient admissions, and 30-day readmission rates
• Functional status — activities of daily living (ADL) scores, cognitive assessments, and disability indices
• Social determinants of health — housing instability, food insecurity, transportation barriers, and health literacy levels (AHRQ Health Literacy Universal Precautions Toolkit)

These domains intersect directly with social determinants of health in care management, where stratification tools increasingly incorporate SDOH screening instruments such as the Protocol for Responding to and Assessing Patients' Assets, Risks, and Experiences (PRAPARE).

How it works

Stratification operates as a multi-stage analytical process, not a single scoring event. The National Committee for Quality Assurance (NCQA) — which accredits health plan care management programs — describes stratification as a continuous cycle tied to case identification, enrollment, and outcome tracking (NCQA Health Plan Accreditation Standards).

Staged process structure:

• Data aggregation — Clinical data from electronic health records, claims streams, pharmacy records, and patient-reported outcomes are compiled into a unified patient profile.
• Scoring and weighting — Algorithms apply weighted scores to each domain. Predictive models such as the Hierarchical Condition Categories (HCC) model, used extensively in Medicare Advantage risk adjustment, assign relative risk scores based on diagnosis clusters (CMS HCC Model, 42 CFR Part 422).
• Tier assignment — Patients are placed into risk tiers. A three-tier model commonly designates: Tier 1 (high complexity, active case management), Tier 2 (moderate risk, disease management or care coordination), and Tier 3 (low risk, preventive outreach). Five-tier models subdivide the high-risk segment to distinguish catastrophic from high-need patients.
• Intervention matching — Each tier maps to a defined care management intensity, ranging from dedicated care manager assignment (Tier 1) to automated wellness reminders (Tier 3 or 5).
• Reassessment — Stratification scores are refreshed at defined intervals — typically 90 days for high-risk patients and annually for low-risk populations — to capture clinical change.

The utilization management in healthcare process depends on stratification outputs to prioritize prior authorization review and concurrent review caseloads.

Common scenarios

Chronic disease populations: Patients with two or more chronic conditions — for example, Type 2 diabetes combined with chronic kidney disease — typically score in the high or complex tier under most stratification models. Chronic disease care management programs use stratification to differentiate stable, well-controlled patients from those with HbA1c values above 9% or eGFR declining at a clinically significant rate.

Post-acute transitions: Patients discharged from inpatient settings undergo stratification to determine transitional care management intensity. cms.gov/Outreach-and-Education/Medicare-Learning-Network-MLN/MLNProducts/Downloads/Transitional-Care-Management-Services-Fact-Sheet-ICN908628.pdf)).

Behavioral health integration: Stratification in behavioral health contexts must account for suicide risk screening scores (such as the Columbia Suicide Severity Rating Scale, C-SSRS), substance use disorder severity, and psychiatric hospitalization history. The Substance Abuse and Mental Health Services Administration (SAMHSA) recommends integrated stratification approaches that weight behavioral health indicators alongside medical ones (SAMHSA Behavioral Health Integration Resources).

Pediatric and geriatric variants: Pediatric care management stratification prioritizes NICU history, developmental delay indices, and family social risk scores. Geriatric care management models weight fall risk assessments (Morse Fall Scale or STEADI toolkit from the CDC), cognitive decline measures, and polypharmacy counts — typically flagging patients taking 10 or more concurrent medications as elevated risk (CDC STEADI Initiative).

Decision boundaries

Stratification produces clinically meaningful decisions only when tier boundaries are defined with precision. Two contrasting model types illustrate the tradeoffs:

Rule-based models apply explicit logical cutoffs — for example, "any patient with 2 or more inpatient admissions in the prior 12 months is high-risk." These models are auditable and align clearly with regulatory documentation requirements under NCQA and URAC accreditation standards (URAC Case Management Accreditation). Their limitation is rigidity; a patient with 1 admission plus 6 ED visits may be underclassified.

Predictive analytic models use regression or machine learning algorithms to generate continuous risk scores, then establish tier cutoffs at specific score thresholds — often at the 75th and 90th population percentiles. These models capture interaction effects between variables but require prospective validation, ongoing bias monitoring, and clear documentation to satisfy CMS program integrity requirements.

Care managers reviewing stratification outputs must understand that tier assignment is a probabilistic signal, not a clinical diagnosis. False-negative classifications — placing a deteriorating patient in a lower tier — carry patient safety implications addressed in complex care management protocols. The care management quality metrics framework used by NCQA's HEDIS measures evaluates stratification accuracy indirectly through outcome indicators such as hospital readmission rates and follow-up after hospitalization rates.

Stratification decisions also carry reimbursement implications. The care management reimbursement and billing landscape ties CCM, TCM, and Principal Care Management (PCM) billing codes directly to documented complexity and risk tier, making tier-assignment documentation a compliance requirement, not merely a clinical preference.

References

• CMS MSSP Regulations, 42 CFR Part 425
• AHRQ Health Literacy Universal Precautions Toolkit
• NCQA Health Plan Accreditation Standards
• CMS HCC Model, 42 CFR Part 422
• SAMHSA Behavioral Health Integration Resources
• CDC STEADI Initiative
• URAC Case Management Accreditation