Evaluation of the factor structure, prevalence, and validity of disturbed grief in DSM-5 and ICD-11

Background: Persistent complex bereavement disorder (PCBD) is a disorder of grief included in DSM-5 Section 3. Prolonged Grief Disorder (PGD) is a disorder of grief that will enter the forthcoming ICD-11. This study evaluated the factor structure, prevalence, and validity of disturbed grief as per DSM-5 and ICD-11. Methods: With data from a community sample (N =512), we used confirmatory factor analysis (CFA) to evaluate the fit of different factor models for PCBD and PGD, determined diagnostic rates for probable PCBD and PGD, and used sensitivity/specificity analyses to evaluate the performance of individual items as indicators of PCBD and PGD. We calculated associations of PCBD-caseness and PGD-caseness with concurrently assessed symptoms of posttraumatic stress disorder (PTSD) and depression and, in a subset of 280 participants, with these same symptoms assessed one year later, to examine concurrent and predictive validity of PCBD and PGD. Results: For PCBD-symptoms, a three-factor model with distinct factors of separation distress, reactive distress, and social/identity disruption fit the data well; for PGD-symptoms a two-factor model with distinct separation distress symptoms and additional symptom (e.g., guilt, anger, blame) yielded acceptable model fit. Overall, items evidenced strong sensitivity and negative predictive power, and relatively poor specificity and positive predictive power. The prevalence of probable DSM-5 PCBD (6.4%) was significantly lower than the prevalence of ICD-11 PGD (18.0%). Both PCBD and PGD were significantly associated with concurrent overall grief, depression, and PTSD; PCBD but not PGD was associated with symptoms one year beyond baseline. Limitations: Limitations include our reliance on self-reported data and symptoms of PCBD and PGD being derived from two questionnaires. Conclusions: Findings provide preliminary evidence for the validity of both the PCBD and PGD constructs, albeit that prevalence rates of both constructs and predictive validity differ—which needs further scrutiny.