The Blueprint

Risk-Stratifying Patients - Two Example Approaches

by Amanda Berra

Why Risk-Stratifying is Important

Ability to risk-stratify patients is one of the major differences between beginning and advanced PCMHs. While beginning PCMHs have their hands full just establishing a patient/disease registry, identifying patients with target conditions, and flagging gaps in care, the more advanced PCMHs are able to sort patients into high, moderate, and low-risk categories. (Note -- for all practical purposes, "risk" here means "risk of hospital admission").

In an unstratified approach, all patients in the target conditions are either where they should be, according to EBM, or not. So for operational purposes, all patients with any gaps in care become equal priorities for resourcing and treatment. In contrast, sorting patients into risk bands helps providers prioritize, achieving maximum bang for buck in deploying clinical/operational resources to assist the patients who need help most urgently.

Operational uses of risk stratification among PCMHs:

  • Designing workflows in which patients are directed to different clinician types depending on that patient's risk -- ensuring all staff are operating at top of license
  • Systematically allocating more health coach/support staff resources to different panels -- supporting sustainability of panel per provider without reducing panel size
  • Making "Move patients at highest risk bands into lower risk bands over time" the operational goal of clinical protocols--lining up PCMH activities with the priorities of other stakeholders
  • Risk-Stratifying and Contracting/Incentives

To amplify the last point about stakeholder interest, I think it's fair to say that many payers assume that PCMHs will be disproportionately resourcing and helping to stabilize patients at greatest risk of hospital admissions -- and they consider this a major value proposition for supporting the PCMH model. (See for example CareFirst's graphic showing the pyramid of enrollee illness burdens, risk-stratified from Band 1 "catastrophic conditions" to Band 5 "healthy", page 38, here)

That means:

  • Some medical home contracts will -- at least in future -- have incentives built in that dovetail with risk stratification. Not just capitation, but any form of cross-continuum goals (such as reducing hospitalizations), or shared-savings contracts. In fact, some conventional FFS P4P contracts for integrated systems already have goals like these -- not to mention existing Medicare incentives around avoiding preventable readmission (for hospitals). Risk stratification is critical to building systems that perform well against measures such as these.
  • For any of you that do not today have medical home contracts in place, but may be working toward commercial contracting support for the future, being able to acknowledge the importance of risk-stratification, and demonstrate risk-stratification capabilities, should help to make the case.

Now for two examples of the "how" ...

Example 1: Lehigh Valley Health Network: Risk Factor Calculation

Lehigh Valley Health Network (LVHN), located in eastern PA, a participant in the Pennsylvania Chronic Care Initiative (CCI) recently shared this benchmark-driven example:

  • Low risk = 0-2 points
  • Moderate risk = 3-5 points
  • High risk = 6-8 points

Source: Stoll, M. "Getting There from Here: In Pursuit of High Value Care at Lehigh Valley Health Network," Medical Home News, April, 2011

Example 2: Vanderbilt University Medical Center "Degree of Control" Risk Stratification

 Vanderbilt University Medical Center's new primary care team pilot risk-stratifies patients using an operational definition of how controlled a given patient's diabetes is:

Sample Group 3: Biannual health coach team contact
Sample Group 2: Monthly health coach team contact
Sample Group 1: Biweekly health coach team contact
Source: Health Care Advisory Board, "Transforming Primary Care: Building a Sustainable Network for Comprehensive Care Delivery" 2010

More Examples, Please

The examples shown here are specific for diabetes, so they actually fall short of what at least some payers have laid out as their vision for the PCMH; ideally, the risk profile should actually be "all-condition," especially for patients who have multiple co-morbidities. Do you have an examples of a risk-stratification approach for condition or conditions beyond diabetes? If you do, please e-mail me ( and I will update this post, so all the Project participants can access ideas for working more conditions into their models.

More information and resources

Comments? Questions? Email me --