Claims data can show what dementia care costs, but not how a person feels or how a caregiver is coping. GUIDE, CMS's dementia care model, fills that gap with two surveys. The PROMIS-10 asks patients about their physical and mental health. The Zarit Burden Interview asks caregivers how much the caregiving role is weighing on them. Together they feed two of the model's five performance measures, and from performance year 2026 both carry up to 3 percent of upside in the performance-based adjustment.
This guide explains what each survey captures, how it is scored, how CMS turns the results into measures, and what the measures can and can't tell you. It is written for GUIDE program leads, quality teams and care navigators.
Why Value-Based Programs Use Surveys
CMS's request for applications says GUIDE aims to enhance quality of care by improving quality of life for people with dementia and reducing burden and strain on their caregivers. It calls quality of life the primary patient-reported outcome measure for the model, and it names reducing caregiver burden as another primary objective.
Neither outcome appears in claims. Only the people involved can report them, so CMS asks participants to collect survey data at assessments and send it through the Patient Assessment and Alignment Form. The other three GUIDE measures, high-risk medication use, total per-capita cost and nursing home admissions, come from claims and need no reporting.
What the PROMIS-10 Captures
The PROMIS Global Health scale, known as the PROMIS-10, is a 10-item measure of health-related quality of life. According to the HealthMeasures scoring manual, it produces two scores, physical health and mental health, each calculated from four of the items.
Global Physical Health reflects physical function, fatigue and pain, according to a description of the instrument in a federally registered trial.
Global Mental Health reflects mood, emotional distress and satisfaction with social activities and roles.
Each item uses a five-point scale, and the pain question is rated 0 to 10 and recoded to five points. Raw scores are converted to T-scores, which are standardized so that 50 is the average for the U.S. general population, with a standard deviation of 10. Higher scores mean better health, so a T-score of 40 is one standard deviation below the average.
The PROMIS-10 is a general measure of health-related quality of life. It was not designed specifically for dementia, and that is worth keeping in mind when you interpret results.
How GUIDE Turns the PROMIS-10 Into a Measure
CMS's Performance Measurement Manual calls the measure Quality of Life for Patients Living with Dementia (QoL-PLWD). It works like this:
Two data points. The measure needs two PROMIS-10 responses from the same patient, collected 12 months apart, plus or minus 60 days. Participants collect the first at the initial comprehensive assessment and the next at the annual assessment. The PROMIS-10 is not required at reassessments.
Maintained or improved. The measure counts patients whose scores stayed the same or got better over that period. For a condition that usually progresses, holding steady counts as success.
Two sub-scales, scored separately. CMS evaluates Global Physical Health and Global Mental Health on their own, because changes in one don't consistently coincide with changes in the other.
Risk adjustment. CMS uses an observed-to-expected ratio. Logistic regression models predict the chance of maintaining or improving based on baseline PROMIS-10 score, dementia severity, whether the patient has a caregiver and residence type. An O/E ratio at or above 1.0 means a participant's patients are doing as well as or better than expected.
Minimum size. CMS scores it only if there are at least 20 eligible patients.
What It Is Worth
For performance year 2025, established program track participants were scored against a benchmark of 0.91 on each sub-scale, which is the 30th percentile of GUIDE participant performance. The payment contribution was -1 percent if neither benchmark was met, +1.5 percent if one was met and +3 percent if both were met. New program track participants earned +2 percent for submitting PROMIS-10 data for at least 90 percent of aligned patients.
From performance year 2026, all participants are scored on performance, with a range of -1 percent to +3 percent. CMS will base the benchmark on performance year 2025 and 2026 data.
The manual also shows how ratios were spread across participants. In cumulative GUIDE data through April 2026, physical health ratios ran from 0.71 at the low end to 1.35 at the high end, with a median of 1.03. Mental health ratios ran from 0.61 to 1.65, with a median of 0.98.
What the Zarit Burden Interview Captures
The Zarit Burden Interview (ZBI-22) is a 22-item questionnaire that measures the burden a caregiver perceives from providing care. Caregivers rate each item from 0 (never) to 4 (nearly always), and the total runs from 0 to 88, with higher scores indicating greater burden. Trial protocols describe the questions as covering the caregiver's health, psychological well-being, finances, social life and relationship with the person they care for.
Published studies commonly interpret total scores in bands. A study of dementia caregivers in the United Kingdom describes scores above 21 as mild to moderate burden, above 41 as moderate to severe, and above 61 as severe. A 2025 paper uses 0 to 21 for little or no burden, 21 to 40 for mild to moderate, 41 to 60 for moderate to severe, and 61 or above for severe burden.
GUIDE's request for applications used the scale to set model tiers in 2023, separating scores of 0 to 60 from 61 to 88. CMS's current manual describes tiers in terms of dementia complexity, caregiver status and residence type, so confirm the current tier criteria in CMS's payment methodology guidance.
How GUIDE Turns the ZBI-22 Into a Measure
The Caregiver-Reported Impact in Dementia measure, known as CID, uses ZBI-22 data from the caregivers identified in the assessment form. According to CMS's manual:
Lower is better. Lower scores suggest that care teams are helping caregivers sustain their role.
The caregiver is named. CMS defines the caregiver as a relative or unpaid nonrelative who helps with daily activities and is listed by name in the assessment form.
It is required now. CID contributed 0 percent to the payment adjustment in performance year 2025, but participants still had to submit ZBI-22 data for caregivers.
It counts from PY 2026. The potential adjustment is -1 percent to +3 percent. When the manual was published in May 2026, CMS had not yet released the benchmark or the continuous alignment criterion for this measure.
It repeats when things change. If the caregiver or the level of caregiver burden changes, a reassessment is required, and the participant must administer the ZBI-22 again and submit the results.
Minimum size. CMS scores CID only if there are at least 20 eligible patients.
What These Surveys Can and Can't Tell You
They capture perception. Both surveys record how people say they feel. That is the point, since quality of life and burden are subjective, but the answers can vary with mood, the day or the setting.
A total score hides detail. A ZBI-22 total can't tell you whether burden comes from finances, sleep, guilt or a specific behavior. Talk with the caregiver about the answers, not only the number.
The PROMIS-10 isn't dementia-specific. It measures general health-related quality of life. CMS adjusts for dementia severity and other factors in the risk model, but the instrument itself wasn't designed around dementia.
Timing matters. The QoL measure needs two responses about a year apart. A missed or badly timed assessment can keep a patient out of the measure.
Some patients may find self-report hard. The manual points participants to the beneficiary alignment fact sheet on GUIDE Connect for reporting procedures. Check it for guidance on patients who have difficulty completing the survey themselves.
These are measures of a program, not of individuals. CMS compares aggregated ratios. A single patient's score is a prompt for care planning, not a verdict.
Practical Steps for Care Teams
Build both surveys into the assessment workflow. Decide who administers each one, and where results are recorded.
Schedule the second PROMIS-10 early. Set a reminder for about 12 months after the first, inside the 60-day window.
Administer the ZBI-22 with care. The questions can be personal, so choose a private moment and follow up on what the caregiver says.
Use results in care planning. A high burden score should lead to a conversation about training, support groups, one-on-one calls and respite, not only a data entry.
Watch for missing data. Track completion by patient and by caregiver, and fix gaps before the submission deadline.
Plan for changes. When a caregiver changes, or burden rises, remember that a reassessment and a new ZBI-22 are required.
Where Elli Cares Fits
Elli Cares doesn't administer, score or submit the PROMIS-10 or the ZBI-22. It gives care teams a way to hear from families between assessments. Daily wellbeing check-ins, family observations and the shared care circle can help a care navigator notice when a caregiver may be struggling or a routine has changed, so the conversation can start before the next scheduled survey. Elli doesn't guarantee better scores, and it doesn't replace CMS reporting. You can read more on our GUIDE page.
Frequently Asked Questions
What is the PROMIS-10?
A 10-item patient-reported measure of health-related quality of life. It produces a Global Physical Health score and a Global Mental Health score, each converted to a T-score where 50 is the U.S. general population average.
What is the Zarit Burden Interview?
A 22-item questionnaire that measures the burden caregivers perceive. Items are scored from 0 to 4, and total scores run from 0 to 88, with higher scores meaning greater burden.
How does GUIDE use the PROMIS-10?
It feeds the Quality of Life for Patients Living with Dementia measure, an observed-to-expected ratio of patients whose scores are maintained or improved over about 12 months, scored separately for physical and mental health.
How does GUIDE use the Zarit Burden Interview?
It feeds the Caregiver-Reported Impact in Dementia measure. Lower scores are better. It counts toward the payment adjustment from performance year 2026, with a range of -1 percent to +3 percent.
Are these surveys required in GUIDE?
Yes. Participants report both through the Patient Assessment and Alignment Form. The ZBI-22 is required for identified caregivers, and the PROMIS-10 is collected at the initial and annual assessments.
