A four-physician family practice pursuing PCMH recognition had a care manager, a nurse who had been with the practice for years, and a list of about 90 patients she followed. When we asked how those patients had been chosen, the honest answer was that physicians mentioned names and she added them. Some were genuinely high risk. Several were pleasant patients with well-controlled diabetes who liked the phone calls. Meanwhile a report we ran showed 41 patients with two or more emergency department visits in the prior year, and 28 of them were not on her list.
That is the problem risk stratification for care management is meant to solve. A practice has limited care management capacity, usually one person or a fraction of one, and a panel of several thousand. Deciding who gets that attention by memory produces a list of the patients who are visible, not the patients who are at risk. NCQA's PCMH standards, under the Care Management and Support concept, expect the practice to establish a systematic process and criteria for identifying patients who may benefit from care management, to consider multiple factors, and to be able to show the percentage of the panel identified. The standards do not prescribe a method. They require that you have one and can prove you used it.
This is the method we set up in small and mid-sized primary care practices: what data to pull, the criteria and how to weight them, the three tiers, how the monthly list gets to the care manager, and the documentation that satisfies an NCQA reviewer.
Key takeaways
- Risk stratification sorts the panel into tiers using defined criteria so care management goes to patients most likely to benefit, not those most often mentioned.
- NCQA requires a documented process using multiple factors and expects the practice to report what share of patients it identified; it does not require a purchased algorithm.
- A workable small-practice model uses utilization, clinical complexity, clinical control and social factors, each scored from the EHR and payer data you already have.
- The high tier should be small enough for your care manager to actually work, usually 2 to 5 percent of the active panel.
- Run it monthly, document the date and the counts, and keep the physician override in writing.
What NCQA is asking for
Under the PCMH standards, the care management concept includes a core criterion requiring the practice to consider a set of factors when identifying patients for care management: behavioral health conditions, high cost or high utilization, poorly controlled or complex conditions, social determinants of health, and patients referred by outside organizations, caregivers or the patients themselves. A second criterion asks the practice to monitor the percentage of the total patient population identified through its process. Related criteria cover the care plan, the patient's involvement in it, and how the plan is shared.
The evidence NCQA reviewers look for is the written process (the criteria and how they are applied), a report or screenshot showing the identification run with the resulting counts, and the percentage of the panel identified. What we see practices get wrong is either having no written criteria (the care manager's list from memory) or buying a risk score from a vendor and being unable to explain what it measures. A reviewer who asks "why is this patient in the high tier" needs an answer that comes from your criteria, not from a black box.
The data you already have
Small practices assume risk stratification needs a population health platform. It helps, but the core model runs from four sources most practices already hold. The EHR problem list and medication list give clinical complexity. The EHR results give clinical control: last A1c, last blood pressure, last depression screening score. The practice management system gives your own utilization: visits in the past twelve months, no-shows, and whether the patient has been seen at all. Payer data, from the care gap and utilization reports most Medicare Advantage and commercial plans send quarterly, gives what happens outside your walls: emergency department visits, inpatient admissions and readmissions, and specialist visit counts. Add the social factors your intake captures, such as the answers to a standardized social needs screen, transportation problems flagged at the front desk or a caregiver noted in the chart.
If the payer reports are not arriving, ask the plans' provider representatives; most will set up a monthly or quarterly file for any practice that requests it. Admission, discharge and transfer (ADT) feeds from your regional health information exchange are the other source worth setting up, because they tell you about the hospital visit the week it happens rather than the quarter after.
A three-tier model with point scoring
The model below is one we have used in family and internal medicine practices with panels of 3,000 to 8,000. Each patient receives points across four domains. The sum determines the tier. The specific thresholds are adjustable to your panel, and adjusting them and documenting why is itself good evidence of a systematic process.
| Domain | Criterion | Points |
|---|---|---|
| Utilization | Each emergency department visit in the past 12 months (maximum 3 counted) | 2 each |
| Utilization | Each inpatient admission in the past 12 months | 4 each |
| Utilization | Readmission within 30 days of a discharge | 4 |
| Clinical complexity | Each chronic condition on the problem list from a defined set (diabetes, heart failure, COPD, CKD stage 3 or above, coronary disease, serious mental illness, substance use disorder) | 2 each |
| Clinical complexity | Ten or more active medications | 3 |
| Clinical control | A1c above 9, or blood pressure above 160 over 100 at last two visits, or PHQ-9 of 15 or higher | 3 each |
| Social factors | Positive social needs screen for housing, food or transportation; no caregiver for a patient with cognitive impairment; language barrier without interpreter access | 2 each |
| Referral | Physician, patient, family or outside agency requests care management | Reviewed individually |
Tier 3, the high tier, is 12 points or more. Tier 2, moderate, is 6 to 11. Tier 1 is everyone else. In a panel of 5,000 active patients we typically see 2 to 4 percent land in tier 3, roughly 100 to 200 patients, and 10 to 15 percent in tier 2. If your tier 3 comes out at 8 percent, raise the threshold; a list the care manager cannot work is not a list. If it comes out under 1 percent, your utilization data is probably incomplete.
A fictional example: a 74-year-old woman with heart failure, COPD and diabetes (6 points), two emergency visits and one admission in the past year (8 points), twelve medications (3 points) and an A1c of 9.4 (3 points) scores 20. She is tier 3 regardless of whether any physician mentioned her name. A 61-year-old man with well-controlled hypertension and diabetes (4 points) and no other flags scores 4 and stays in tier 1, even if he calls often.
From the list to the care manager
Run the scoring on the first business day of each month against the active panel, defined as patients seen in the past 24 months or attributed to you by a payer. The output is three things: the tier 3 list, the tier 2 list, and a summary of counts and percentages by tier with the run date. The tier 3 list goes to the care manager, who compares it to the current caseload: new patients to enroll, patients who dropped a tier and may be ready to graduate, and patients who remain. The tier 2 list drives lighter-touch work: outreach for care gaps, pre-visit planning flags and a check at the next visit.
Physicians keep an override, in both directions. A physician can add a patient the score missed, with a one-line reason recorded in the chart, and can remove one the score caught for a documented reason (a patient in hospice, or one who has declined). NCQA does not object to clinical judgment; it objects to clinical judgment being the only method. Record the overrides on the monthly summary. In the family practice we described, the first run put 38 of the care manager's 90 patients in tier 3, added 64 she had never contacted, and moved 31 to tier 1. That reshuffle was uncomfortable and correct.
Documentation that satisfies a reviewer
Keep a one-page written process: the data sources, the criteria and points, the tier thresholds, the run schedule, who runs it, how the list reaches the care manager, and how overrides are recorded. Date it and version it when thresholds change. Keep each monthly summary with counts by tier and the percentage of the panel identified. Keep a de-identified example of the tier 3 list. When NCQA asks for evidence that the practice uses multiple factors and monitors the percentage identified, this folder answers both in a few pages. It also forms part of the annual reporting evidence, which is why we set it up during recognition and then leave it running; the work of PCMH annual review is much lighter for practices that never stopped.
Practices earlier in the process, still deciding whether to pursue recognition, usually find that building this model is worth doing regardless, because the tier 3 list is also the list of patients eligible for chronic care management billing and the list most likely to generate the admissions that quality contracts penalize. Our PCMH recognition and transformation work starts here for that reason.
Questions we hear
Can we just use the risk score our Medicare Advantage plan sends us?
You can use it as one input, and many practices add it as a domain worth a few points. Relying on it alone has two problems: it covers only that plan's members, so the rest of your panel is unscored, and you usually cannot explain what drives it. NCQA wants a process you can describe. A payer score inside your own model is fine; a payer score instead of your own model is thin.
How often should we re-run it?
Monthly is the standard we recommend for practices with a care manager, because admissions and emergency visits change the picture quickly. Quarterly is acceptable for a very small practice with limited data feeds, and the written process should say which one you do and why.
What do we do with tier 2, if the care manager only has time for tier 3?
Tier 2 is worked through the visit, not through outreach. Flag them in the schedule so the pre-visit planning huddle sees them, close care gaps when they are in the room, and make sure they have a follow-up booked before they leave. That takes almost no additional staff time and moves some of them out of tier 2 before they reach tier 3.
What to do this week
- Write down how care management patients are chosen today, honestly, and compare that list to a report of patients with two or more emergency visits in the past year.
- Confirm which payers send utilization and care gap reports, and request a monthly file from any that do not.
- Build the point model in a spreadsheet or your EHR reporting tool using the table above, and run it once against the active panel.
- Adjust the tier 3 threshold until the list is one your care manager can realistically carry, and record the threshold and the reason.
- Draft the one-page written process and schedule the first monthly run for the first business day of next month.
