Crosswalk between PROMIS computer adaptive tests and numerical rating scales in cancer patients: Anxiety, depression, pain interference, physical function.

Aims: The study aimed to establish crosswalk between PROMIS computer adaptive tests (CATs) and numerical rating scales (NRS) in emotional distress-anxiety, emotional distress-depression, pain interference, and physical function. Methods: This retrospective analysis assessed data collected from patients from E2C2 (4/2019-1/2023), a pragmatic symptom management trial. We linked 0-10 numerical rating scales (NRS) to corresponding PROMIS CAT scores using responses within 10 min of each other and randomly selected one visit per participant. A single-group equipercentile linking method was used, with polynomial log-linear smoothing to reduce sample errors. Assumptions were tested by examining NRS-PROMIS correlations and population invariance across subgroups by sex, race, and ethnicity. Linked scores were compared with observed PROMIS scores using intraclass correlations and differences in means and standard deviations (SDs). Results: The sample sizes for the four domains ranged from 32,434 to 33,081, with 34,029 unique patients with median age 64 (range 18-102), 56 % female, 92 % white, and 1 % Hispanic. About 30 % of patients provided the highest or lowest NRS scores. The NRS and PROMIS scores were strongly correlated (0.72-0.78). Population invariance across three subgroups and four domains was supported with the standardized Root Expected Mean Square Difference (REMSD) ranging from 0.8 % to 7 % SD, which were under 8 % SD threshold. The correlations between the linked and observed T scores were high ranging from 0.73 to 0.81. All four domains had small mean differences from the observed T scores. Conclusion: This study linked the NRS and PROMIS CAT in four domains and created a crosswalk table for conversion. This study facilitates the interpretability of NRS on the metric of PROMIS CAT T-scores.