Developing an item bank to measure psychological distress in cancerpatients using Rasch models

Abstract

8104 Background Psychological distress (PsyD) is common in cancer patients (CP). Self-report screening instruments are disappointing in detecting CP needing specialist intervention for PsyD, and CP with increased PsyD below psychiatric disorder, needing support within cancer care. Items assessing PsyD may fit on a single scale, so we may reduce the number of items CP must answer and tailor questions to CP own level of PsyD. We aim to apply Rasch analyses to instruments measuring PsyD, to develop a bank of items, and to evaluate how these items fit a unidimensional scale across the range of severity of PsyD (in logits). Methods A Rasch common-item equating analysis was applied to our database. The Beck Depression Inventory, General Health Questionnaire, Mental Health Inventory, State-Trait Anxiety Inventory, FACT Emotional Well-being scale, Patient Health Questionnaire, Hospital Anxiety and Depression Scale, and EORTC QLQ-C30 Emotional Functioning scale were analysed separately. Items that misfit the single scale (infit mean square >1.4) were removed. Then items from each instrument were added with common items as anchor points. Results A sample of 4923 CP provided an initial bank of 83 items. When 18 items were removed due to misfit, 65 items had infit and outfit mean square <1.4. No factor structure was evident in the residuals on principal components analysis, confirming a good unidimensional structure. Internal reliability was high (person separation index 0.85). Item locations ranged from -5 to +3 logits. Therefore many items assessed high levels of PsyD. Person abilities ranged from -5 to +1. Many CP reported low levels of PsyD. Limited overlap between item and person locations indicates a need to develop further items. Conclusions In a large sample of CP we have applied a Rasch model to items from widely used previously validated self-report PsyD questionnaires. Many items fit a single continuum of PsyD. Additional items may improve the item bank, leading in future to a fixed questionnaire and computer adaptive testing to measure PsyD in CP on a single continuum, using fewer items, tailored to the patient. No significant financial relationships to disclose.