The Patient-Reported Outcomes Measurement Information System (PROMIS®) includes measures of physical, mental, and social health for adults and children.

Is there a way to convert PROMIS T-scores to percentiles?

8 months 3 weeks ago #726

Ideally, we would advise users who want to use percentiles when reporting PRO scores to actually calculate percentiles by rank ordering the T-scores in a large sample that is specific to the user’s population of interest. While that would ensure that the percentile is specific and valid for the user’s purpose, we realize that this is not practical for most users. The T-scores can indeed be converted to an estimate of percentiles by using the standard normal curve figure, where a T-score of 50 is the 50th percentile, 60 represents the 84th percentile, while 40 represents the 16th percentile. Now, it’s important to keep in mind the direction of the health concept that is being measured. So for example, a 60 on depression represents the 84th percentile in depression, a relatively high level of depression. This means that about 84% of the estimated general population have fewer depression symptoms than the person who obtained a score of 60. Said another way, a T-score of 60 on depression also means that only 16 percent have more depressive symptoms than the person reaching this score.

Because the PROMIS domains are not exactly normally distributed – few health concepts are -- this normal curve conversion to percentiles is an approximation. Using an approximation may be fine in some situations (e.g., comparing large group means) but may require more consideration in other situations (e.g., using a single score at one point in time for patient care). Further studies with very large, representative samples would be required to get exact percentages in the general population and to see how much they differ from the normal curve. We have some information on percentiles in the general population for selected PROMIS domains at . See “Sub-Norms for Interpreting Scores“

One reason people are interested in transforming T-scores to percentiles is to aid in score interpretation. PROMIS has developed multiple tools to aid in score interpretation including:
- Identifying the reference population with a mean of 50 (see )
- Identifying score cut-points for mild, moderate, and severe (see )
- Using multiple methods for estimating meaningful change (see ) and listing some published results (see )
- Mapping T-scores to G-code severity modifiers used by the Centers for Medicare and Medicaid (see )
- Mapping T-scores to other widely used PRO measures’ scores (see )
- Creating figures showing the most likely response to short form items for each T-score (see )

Please Log in or Create an account to join the conversation.

8 months 3 weeks ago #725

Is there a formula for converting T-scores to percentiles?

Please Log in or Create an account to join the conversation.

Moderators: HealthMeasures