Impact of speeding on the validity of patient-reported outcome measures. An analysis of psychometric properties stratified by response times

Background: Randomised trials and meta-evidence increasingly rely on patient-reported outcome measures (PROMs). The psychometric properties of patient-reported outcome measures PROMs, including validity, reliability, and responsiveness, are typically established using high-quality datasets, which may not reflect the data quality in clinical trials. Increasing survey burden and fatigue may lead to issues with short response times that can reflect insufficient engagement. This phenomenon, often referred to as speeding can lead to random, patterned, or otherwise invalid responses. Objective: This study aims to investigate how response times influence the psychometric properties of the Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) pain domain in patients with chronic postsurgical pain. Methods: The study is based on responses to the 5-item Western Ontario and McMaster Universities Osteoarthritis Index (WOMAC) pain domain (Likert-scale, version 3.1) from 2,031 patients who underwent total hip arthroplasty (THA), 2,172 patients who underwent total knee arthroplasty (TKA), and 870 patients who underwent unicompartmental knee arthroplasty (UKA) more than one year previously. Each of the three datasets, containing patients who have undergone THA, UKA and TKA, are individually stratified into response time deciles. For each of the deciles, we will evaluate if the data fit a congeneric measurement model, i.e. a model that assumes that the set of observed items all measure the same underlying latent factor. This evaluation of construct validity is done using Item Response Theory (IRT) and Confirmatory Factor Analysis (CFA). Perspective: The results will be submitted for publication in peer-reviewed journals. We will seek to make the report freely available, either by open-access publication or through publication on a preprint server, e.g. www.medrxiv.org.