Psychometric validation of Quality of Dying and Death in Brazilian family members

Abstract

Abstract Context: Death is a complex, subjective phenomenon that requires the understanding of experiences to provide care and qualification of the end-of-life process. Objective: To analyze the psychometric properties of the portuguese version (Brazil) of the Quality of Dying and Death (QODD) scale on family members of patients who died in adult intensive care units. Methods: A methodological study was conducted with 326 family members of patients that died in three ICUs of public hospitals in the state of SP, Brazil. For such, the QODD 3.2a (25 items and six domains) was administered in the period from December 2020 to March 2022. The analysis was performed using the classic theory of tests as well as exploratory factor analysis. The goodness of fit of the model was tested using confirmatory factor analysis. Concurrent validity was evaluated using Spearman’s correlation coefficients between the scores of the overall scale and domains. Cronbach’s alpha coefficient and the intraclass correlation coefficient (ICC) were used for the evaluation of internal consistency and temporal stability, respectively. Results: Horn’s parallel analysis indicated two factors that were not confirmed in the exploratory factor analysis. A single factor retained 18 of the initial 25 items and the analysis of the goodness of fit to the unidimensional model resulted in the following: CFI = 0.7545, TLI = 0.690, chi-squared = 767.33, df = 135, RMSEA = 0.121 with 90%CI and p = 5.04409e-98. The inter-item correlations indicated a predominance of weak correlations among the items of the instrument. The items with the largest number of moderate correlations were questions 13b, 9b and 10b and a strong correlation was found between questions 15b and 16b. Cronbach’s alpha coefficient was 0.8 and the ICC was 0.9. Conclusion: The Quality of Dying and Death – Version 3.2a (intensive therapy) in Brazilian Portuguese has a unidimensional structure and acceptable reliability, but did not obtain a good fit to the proposed factorial model.