Establishing clinically-relevant terms and severity thresholds for Patient-Reported Outcomes Measurement Information System^® (PROMIS^®) measures of physical function, cognitive function, and sleep disturbance in people with cancer using standard setting

Purpose: Patient-Reported Outcomes Measurement Information System^® (PROMIS^®) physical function, cognitive function, and sleep disturbance measures are increasingly used in cancer care. However, there is limited guidance for interpreting the clinical meaning of scores. This study aimed to apply bookmarking, a standard setting methodology, to identify PROMIS score thresholds in the context of cancer care. Methods: Using item parameters, we constructed vignettes of five items covering the range of possible scores. Focus groups were held with cancer care providers and people with cancer. Terminology for categorizing levels of severity was explored. Participants rank ordered vignettes by severity and then placed bookmarks between vignettes representing different levels of severity. Group discussion was held until consensus on bookmark placement was reached. Results: Clinicians selected “within normal limits,” “mild,” “moderate,” and “severe” to describe levels of severity. Both patients and clinicians were able to apply these labels, but there was not unanimous support for any set of descriptors. Clinicians and patients agreed on all severity thresholds for sleep disturbance. For cognitive and physical function, clinicians and patients agreed on the threshold between “within normal limits” and “mild.” However, patients required greater dysfunction than clinicians before applying “moderate” and “severe” labels. Conclusions: Bookmarking can be applied to develop provisional score interpretation for PROMIS measures. Patients and clinicians were frequently consistent in their bookmark placement. When there was variance, patients required more dysfunction before assigning more severity. Additional research with other cancer samples is needed to evaluate the replicability and generalizability of our findings.