This is the first study systematically to interrogate a commonly used clinical index for ulcerative colitis, using robust psychometric and statistical methodologies. The current 9-item SCCAI was found to have limited range, unscalable items and poor reliability, which is inefficient and likely to lead to redundancy between questionnaire items. When scale reductions were applied through a series of systematic revisions, it was found that a 4-item scale that included daytime frequency, nocturnal frequency, urgency and rectal bleeding provided the optimum trait information and scalability. Both the 7- and 4-item versions of the scale possessed weak reliability, but the 4-item scale was marginally more reliable. These findings imply that a reduced-item SCCAI would reduce response burden, whilst retaining maximal clinical information. Given that these questions are asked on a daily basis by the TrueColours platform, this reduction may enhance completion rates and sustain adherence to the programme.
IRT has been successfully used to assess the efficiency of clinical measurement tools and subsequent item reduction, in a range of clinical contexts. It has been used to propose reductions in the burden of psychiatric evaluation tools, including the 19-item feelings scale for depression 21, the 16-item Anxiety Sensitivity Index 22 and 65-item Social Responsiveness Scale in autism spectrum disorder 23. This methodology is increasingly adopted to modify indices measuring patient reported outcomes in chronic diseases, including the 11-item Roland-Morris disability scale for chronic pain 24, 14-item Valued Life Activities scale rheumatoid arthritis patients 25 and 31-item Qualeffo-31 for quality of life in osteoporosis 26. All these studies have shown high correlation between the shortened index and the original, with regard to sensitivity, specificity and precision, despite the reduced number of items. One shortened scale demonstrated superior psychometric properties compared to the original 23.
A key strength of this study is the use of ePROs from the TrueColours’ database. This allowed a diverse, real-world sample of patients with differing levels of disease activity to be collected. This enabled IRT analysis to scrutinize the merit of each item in differentiating patients with inactive vs. severely active disease. Furthermore, rigorous quantitative analysis of the SCCAI of this large cohort permitted a sophisticated evaluation of an existing index constructed prior to the contemporary criteria for index development 27,28.
Limitations to this study include that test-retest reliability was not measured, because each patients’ first response on the programme was the only one considered. Additionally, the binarisation of item responses has not been validated, even if it has clinical credibility. Furthermore, weak reliability was maintained at individual measures of the latent trait (disease activity) and overall using a 4-item scale, suggesting that the scale could benefit from the testing of revised items during validation. External validity of the reduced scale has yet to be tested, which is best achieved by comparing symptom responses measured by the modified SCCAI with more objective measures of disease activity, such as fecal calprotectin, endoscopic and histological disease activity.
Item response theory is a valid and robust psychometric methodology, which may be used to analyse patient-reported outcome questionnaires. We have shown that reduction of the SCCAI from the original 9-item to a 4-item scale provides optimum trait information. Changing the index would minimise the patient response burden in an era where ePROs are a pivotal component of improving outcomes.