Natural language processing applied to psychiatric clinical notes: Scoping review (Preprint)

BACKGROUND Psychiatric clinical narratives in electronic health records (EHRs) provide rich longitudinal information that can support clinical decision-making. Using historical medical data can enable earlier identification of mental illness, better characterization of disease trajectories, and more personalized treatment planning. Natural language processing (NLP) transforms unstructured psychiatric records into analyzable representations for research and care. OBJECTIVE This scoping review aims to systematically summarize NLP methodologies for psychiatric EHR narratives, compare major modeling paradigms and application areas, and highlight emerging large language models (LLMs) trends, key challenges, and future research directions. METHODS Following the Preferred Reporting Items for Systematic Reviews and Meta-Analyses Extension for Scoping Reviews (PRISMA-ScR) guidelines, a literature search was conducted for articles on NLP methods based on psychiatric clinical records published from January 2021 to December 2025 in Ovid MEDLINE In-Process & Other Nonindexed Citations, Ovid MEDLINE, Ovid EMBASE, PubMed, Scopus, Web of Science, Science Direct, and the ACM Digital Library. This scoping review analyzed NLP methods applied to psychiatric clinical records, focusing on major trends, identifying suitable features for traditional machine learning-based models, applications of pre-trained language models (PLMs), and key challenges. Approaches were categorized as rule-based, traditional machine learning, deep learning, hybrid, and LLMs–based methods across information extraction and text classification tasks. RESULTS A total of 383 publications were retrieved. After screening titles, abstracts and full-text, 101 were selected. Rule-based methods (n=36) and hybrid approaches (n=34) remained the most widely used techniques, largely favored for their interpretability in handling nuanced, subjective clinical narratives. These were followed by deep learning (n=15) and traditional machine learning (n=10) methods. These methods were applied to key domains including diagnosis, treatment optimization, and risk stratification. Notably, LLM-based methods (n=6) represent a rapidly emerging trend, indicating a sigHybrid NLP approaches remain dominant, combining domain rules with deep learning for extraction and classification. Deep learning and PLMs continue to advance, with domain adaptation supporting medical terminology and clinical semantics. LLMs may further automate complex workflows via zero-shot capabilities and reasoning, alongside growing interest in temporal modeling and multimodal integration. Key future needs include improved generalizability, privacy protection, and careful attention to ethical implications in clinical deployment.nificant paradigm shift toward generative and reasoning-based applications. CONCLUSIONS Hybrid NLP approaches remain dominant, combining domain rules with deep learning for extraction and classification. Deep learning and PLMs continue to advance, with domain adaptation supporting medical terminology and clinical semantics. LLMs may further automate complex workflows via zero-shot capabilities and reasoning, alongside growing interest in temporal modeling and multimodal integration. Key future needs include improved generalizability, privacy protection, and careful attention to ethical implications in clinical deployment.