A Psychometric Evaluation of the Staff-Reported EOLD-CAD Measure among Nursing Home Residents with Cognitive Impairment

Abstract

Objectives: The End-of-Life Dementia-Comfort Assessment in Dying (EOLD-CAD) scale is one of the few outcome instruments designed to capture symptom burden and well-being among nursing home residents with dementia; however, psychometric evaluations of the EOLD-CAD are limited. Although the instrument is often used to assess outcomes prospectively, it was originally developed and tested as a postmortem assessment. The purpose of this study is to evaluate the instrument properties of the EOLD-CAD using staff reports from a large sample of nursing home residents with cognitive impairment prior to death. Methods: Using data from the multi-state UPLIFT clinical trial, this study evaluated the psychometric properties of the EOLD-CAD from 168 nursing home staff members reporting outcomes for 611 living residents with moderate to severe cognitive impairment. Staff also reported on resident quality-of-life using two different single item measures. We conducted confirmatory factor analysis (CFA) and assessed construct validity, inter-item reliability, and observer report bias. Results: CFA produced a four-factor solution. All factor loadings were > 0.40, ranging from 0.61-0.95 for Physical Distress, 0.71-0.91 for Dying Symptoms, 0.61-0.78 for Emotional Distress, and 0.89-0.94 for Well-Being. Model indices suggest a good fit to the data with root mean square error of approximation (RMSEA) = 0.053 (95% CI = (0.044, 0.062)), comparative fit index (CFI) = 0.971, and standardized root mean square residual (SRMR) = 0.093, with the SRMR slightly above the conventional threshold of > 0.08. Based on intraclass correlation coefficients (ICC), patterns of observer reports were identified among staff who provided data for multiple residents. ICCs were notably high (> 0.60) for Well-Being items. The EOLD-CAD elicited a Cronbach's alpha of 0.73, and the instrument was negatively correlated with items measuring resident quality of life. Conclusions: We found that when the EOLD-CAD was completed by nursing home staff familiar with the respective residents, observer-based patterns were detectable. Such patterns were adjusted for in our CFA, from we found that the EOLD-CAD exhibited multidimensionality with a four-factor structure capturing: Physical Distress, Emotional Distress, Dying Symptoms, and Well-Being. In addition to the CFA, the EOLD-CAD demonstrated generally valid and reliable psychometric properties in our population of long-stay nursing home residents with moderate to severe cognitive impairment.