Computer Adaptive Tests (CATs)
Computer Adaptive Tests (CATs)
Computer Adaptive Tests (CATs) are a type of measure in which the questions a person answers are tailored to that person. Each response is used to further refine a person’s score.
Item Banks
All CATs begin from an item bank.
- Item bank = collection of items (questions) all measuring the same thing (e.g., physical function)
- Items are ordered by level of difficulty/severity from low (e.g., I have difficulty getting in and out of bed) to high (I am able to run 5 miles) using Item Response Theory
Tailoring Items to a Respondent
- The first item administered in a CAT is usually one in the middle of the range of function or symptom severity.
- After a person provides a response, an estimated score is calculated.
- The CAT algorithm then selects the best item in the item bank for refining the estimated score.
- After a person provides a response, the estimated score is recalculated.
- The CAT continues to administer items until:
- A specified level of measurement precision is reached (e.g., <3.0 on a T-score metric) OR
- A specified maximum number of items was administered (generally 12 for HealthMeasures)
Why use CATs?
- Brief - Usually 4-12 items; can be completed in under a minute
- Precise - Low error
- Tailored content - CATs don’t ask questions that aren’t relevant to the respondent
- Cover a wide range of function or symptoms
Interested in using PROMIS Computer Adaptive Tests (CATs)?
Learn more about CATs
- Read about Item Response Theory
- Watch this 5-minute introductory video on What is a CAT?
- For a visual explanation of CAT, please see part 6 of Karon Cook’s video series “Applications of IRT".
PROMIS CAT Stopping Rules
Adults
- Minimum number of items administered = 4
- Stop when one of these occurs:
- 12 items are administered OR
- The standard error is below a threshold (0.3 on the theta metric or 3.0 on the T-score metric)
Pediatrics and Parent Proxy
- Minimum number of items administered = 5
- Stop when one of these occurs:
- 12 items are administered OR
- The standard error is below a threshold (0.4 on the theta metric or 4.0 on the T-score metric)