Language-based detection of depression with machine learning: systematic review and meta-analysis

Early detection of depression is critical for timely intervention. Natural language processing (NLP) and machine learning (ML) approaches have increasingly been used to automatically detect depression from text data, yet evidence regarding their performance remains limited. We systematically reviewed and meta-analyzed studies applying NLP and ML to identify depression from spoken or written language. Six electronic databases and additional sources were searched, yielding 892 articles, of which 123 met the inclusion criteria. One representative result per dataset was selected for quantitative synthesis. Pooled accuracy, based on 43 studies including 40,983 text samples, was 0.80. Pooled precision (28 studies) was 0.78, recall (33 studies) was 0.76, AUC (14 studies) was 0.79, and balanced accuracy (16 studies) was 0.71. Subgroup analyses showed significant differences by language, text source, feature type, and classifier. Accuracy was highest in studies using structured clinical interviews, non-English languages, and linguistic or embedding-based features. However, in one-at-a-time meta-regressions, only the text source remained a significant predictor, explaining 13.6% of the between-study variance. Publication bias was minimal. Automated depression detection from text shows promising performance with substantial heterogeneity. Findings highlight both limitations and potential of text-based depression detection and underscore the need for methodological standardization and validation before clinical use.