Linguistic Markers in At-Risk Mental States Using Natural Language Processing: A Systematic Review.
Level 3 - non-randomized controlled study
Systematic review of non-randomized observational and diagnostic/prognostic studies
PubMed 42072900 · doi:10.3390/healthcare14080999
What was done
A systematic review conducted according to PRISMA 2020 guidelines evaluating linguistic markers analyzed via natural language processing (NLP) in individuals with at-risk mental states (ARMS). PubMed, PsycInfo, and Scopus were searched from inception to October 2025. Fifteen studies comprising 1,313 participants were included out of 90 initial search results.
What was found
Alterations in semantic coherence, syntactic complexity, referential cohesion, and speech/content poverty distinguished ARMS individuals from healthy controls. NLP-analyzed markers predicted the onset of psychosis with accuracy levels ranging from 79% to 100%.
Why it matters
Automated linguistic analysis offers an objective, non-invasive computational approach to complement traditional clinical assessments in early psychosis risk stratification.
Limits
The review notes significant methodological heterogeneity and marked variability in sample sizes across studies. Performance metrics such as sensitivity, specificity, and positive predictive value are not detailed in the abstract, nor is the extent of out-of-sample validation reported.