# The latent structure of ICD-11 Prolonged Grief: Replicated factor mixture models in two national cohorts

## TL;DR

This study explores the structure of prolonged grief in the general population using data from two national cohorts and finds that symptoms are primarily dimensional with severity-based groupings.

## Contribution

The study introduces a three-factor, four-class factor-mixture model for prolonged grief that is reproducible across two independent populations.

## Key findings

- A three-factor measurement structure was supported in both cohorts, with instability in items related to anger, avoidance, blame, and difficulty accepting.
- A three-factor, four-class factor-mixture model provided the most interpretable and reproducible summary across samples.
- Symptom profiles progressed monotonically with severity, with blame and anger as high-severity indicators and avoidance having low discriminative value.

## Abstract

Prolonged Grief Disorder (PGD) is recognised in ICD-11 and DSM-5-TR, yet its latent structure in the general population remains uncertain. We examined whether population-level prolonged grief, assessed with the International Prolonged Grief Disorder Scale (IPGDS), is best characterised by a single severity continuum or a hybrid structure combining dimensions with latent classes, using cross-sectional data from two independent general-population cohorts: a UK community panel (N = 1,777) and a nationally recruited Irish sample (N = 950). Analyses followed a stepwise sequence of exploratory and confirmatory factor analyses, latent profile analysis, and factor mixture modelling. In both cohorts, a three-factor measurement structure was supported, with Items 5–8 (anger, avoidance, blame, difficulty accepting) showing consistent cross-sample instability. A three-factor, four-class factor-mixture model provided the most interpretable and reproducible summary across samples, preserving the factor structure while allowing latent classes to accommodate atypical item behaviour. The same four severity-ordered classes and prevalence ranking were recovered in both cohorts, with high classification precision (entropy = 1.00; average posterior probabilities ≥.97). Symptom profiles progressed monotonically with severity; blame and anger functioned as high-severity indicators, whereas avoidance contributed little to distinguishing the highest-severity class. At the population level, PGD symptoms were primarily dimensional, with latent classes best interpreted as pragmatic severity groupings rather than discrete categories. For surveillance and screening, dimensional assessment focused on separation distress may be complemented by brief checks for persistent blame and anger, with avoidance of lower discriminative value in community samples. As these findings are based on cross-sectional data, future longitudinal research is needed to test the temporal stability of class membership and its prognostic value for functional outcomes and service use.