Evidence of non-response bias in patient reported outcome measurement information system surveys

Background

Patient-reported outcome measurement instrument surveys (PROMIS) have been increasingly used to assess clinical outcomes following a variety of medical procedures. Despite their value in evaluating patient functional status, response rates to these surveys remain low. Understanding differences in demographics between responders and non-responders to PROMIS may be beneficial to improving generalizability and response rates.

Objective

The primary aim of this study was to compare demographic characteristics between patients who respond to PROMIS surveys and those who do not, hypothesizing no significant differences between these groups.

Methods

Adult new patient visits from 2018 to 2022 in Department of Physical Medicine & Rehabilitation outpatient clinics at a single academic medical center were analyzed. Patients completed PROMIS surveys (physical function, pain interference, and depression) on iPads at each clinic visit. Demographic variables such as gender, race, BMI, smoking status, employment status, marital status, and Area Deprivation Index (ADI) were collected. Univariate and multivariate analyses were conducted to assess for variables associated with an increased likelihood of responding to PROMIS surveys.

Results

A total of 29,830 patients were included in this study. Of the total patient cohort, 8331 (27.9 %) responded to the PROMIS surveys. Significant demographic differences were found between responders and non-responders. Patients in the least deprived ADI quartile were more likely to respond compared to those in the most deprived quartile (33.5 % vs 23.7 %, p < 0.001). Employed patients, white patients, non-smokers and married individuals were more likely to respond. On multivariate analysis, unemployment (OR 0.71, p = 0.006), increased BMI (OR 0.93, p = 0.014), and higher ADI (OR 0.94, p = 0.003) were significantly associated with lower response rates.

Conclusions

PROMIS response rates are influenced by patient demographics, with lower response rates observed in unemployed, non-white, and socioeconomically deprived populations. These findings highlight the need for targeted interventions to increase response rates and ensure equitable data collection in PROMIS surveys to enhancing the generalizability of clinical decisions made using PROMIS data.