AB1324 MEASUREMENT PROPERTIES OF THE 10-ITEM CONNOR-DAVIDSON RESILIENCE SCALE AMONG PATIENTS WITH TOTAL KNEE REPLACEMENT BASED ON ITEM RESPONSE THEORY

Abstract

Background The number of older adults undergoing total knee arthroplasty (TKR) for knee osteoarthritis (OA) has been increasing. While psychological resilience may have a role in outcomes after TKR[1], no psychological resilience scale has been validated among patients with knee OA. Objectives To examine the measurement properties of the 10-item Connor-Davidson Resilience Scale (CD-RISC10) among older adults with knee OA. Methods Using data from 700 consecutive consented patients with knee OA enlisted for TKR, we examined four measurement properties of CD-RISC10. For structural validity, we fitted graded response model (grm) and examined standardized root-mean square residual (SRMR) for unidimensionality, Chi-square p-value for item fit, residual correlation for local independence and item-characteristic curves for monotonicity. For internal consistency, we examined Cronbach’s alpha. For construct validity, we hypothesized that CD-RISC10 as a measure of psychological resilience would be moderately correlated (Pearson’s R 0.30–0.50) with self-efficacy and mental well-being but weakly correlated (Pearson’s R 66 years), and language (English vs Chinese) and reported the McFadden’s R2. We used mirt and lordif packages on Rx64 v3.5.1 to examine grm and DIF respectively. We reported the findings based on COnsensus based Standards for the selection of health Measurement Instruments (COSMIN)[2]. Results Of the 700 patients (age 65.5±6.9 years old), 70.6% were females, 83.9% Chinese, 55% responded questionnaire in Chinese. Mean CD-RISC10 score was 27.5±5.2, with minimal floor (0.1%) and ceiling effects (3.9%). CD-RISC10 demonstrated unidimensionality (SRMR=0.053), local independence (residual correlation=-0.025), monotonicity (adequate looking item characteristic curves) and internal consistency (Cronbach’s alpha=0.91). Items 1 and 4 had potential misfit (Chi-square p>0.01). All a-priori hypotheses were met (Table 1). No item had significant DIF with age, gender and language, except item 4 had borderline non-uniform DIF across language (McFadden’s R2=0.03). Conclusion We demonstrated structure validity, internal consistency, construct validity and measurement invariance of CD-RISC10 in the measurement of psychological resilience among older adults with knee OA.