Identifying patterns of neuropathological diagnoses in medico-legal autopsies – A latent class analysis of 1270 neuropathologically examined cases from Finland

Neuropathological diseases and injuries are common causes of death and frequently encountered in medico-legal autopsies. Previous studies have suggested that many neuropathological findings co-occur. This retrospective and register-based study aimed to identify latent subgroups of neuropathological diagnoses, including both disease and injury findings. The sample comprised 1270 neuropathologically examined medico-legal autopsy cases from the Forensic Medicine Unit of the Finnish Institute for Health and Welfare over the years 2016–2022. Neuropathological diagnoses were obtained from neuropathologists’ reports, and background characteristics were collected from medico-legal cause-of-death investigation documents. Latent class analysis (LCA) was applied to identify clustering of neuropathological diagnoses among the sample. The dataset comprised 60.6% males, and the median age at death was 64 (interquartile range 43–77) years. LCA models with 2–8 classes were explored; the 3-class model was considered best fit, based on Bayesian Information Criterion values and interpretability. Class 1 was characterized by high prevalences of neurodegenerative and degenerative cerebrovascular diseases, as well as highest age. Class 2 included most cases with acute haemorrhages, their potential causes, and secondary complications. Class 3 comprised a diverse range of neuropathological diagnoses, including cases with epilepsy-related consultation questions and diagnoses. Awareness of co-occurring neuropathological diagnoses would be expected to advocate diagnostic vigilance and facilitate contextual interpretation of co-existing findings in the medico-legal practice. Having demonstrated the feasibility and potential relevance of LCA in forensic neuropathology, this study encourages future applications of LCA in other medico-legal contexts. The latent class solution with three classes appeared to fit the dataset best. Class 1 was characterized by neurodegeneration. Class 2 was characterized by acute haemorrhagic conditions. Class 3 comprised a diverse range of neuropathological diagnoses.