Natural Language Processing: Opportunities and Challenges for Patients, Providers, and Hospital Systems

Abstract

In medicine, language, such as “history” of present illness and “chief complaints,” is used to understand patients' experience. Language is a rich source of data by which physicians make inferences, build therapeutic alliances, document formulations in the electronic health record, and care for patients. In psychiatry, language is the main source of data as there are no objective laboratory tests for diagnosis of psychiatric illness. It is also a therapeutic tool in psychotherapy. There now exist computer-based automated approaches to language analysis that can serve as tools for diagnosis, documentation, and treatment. Here, we review the history of natural language processing and provide examples of its use in psychiatry, such as psychosis prediction and detection of intoxication, and in extraction of information from psychiatric portions of the electronic health record. We also review future directions for studying discourse (and turn-taking), and potential applications of automated analysis to concurrent modes of communication, including prosody, pauses, face expression, and gesture. [ Psychiatr Ann. 2019;49(5):202–208.]

Uncovering the linguistic characteristics of psychotherapy: a computational approach to measure therapist language timing, responsiveness, and consistency

“…It may be that patient improvement is caused by unmeasured covariates, or that therapist language is responsive to patient improvement or decompensation. Other approaches exist for feature implementation and should be evaluated, especially in the context of accuracy and appropriateness across demographic and clinical patient characteristics 43,[67][68][69][70][71] . Phase 3 -clinical relevance.…”

Cognitive Emotional Embedded Representations of Text to Predict Suicidal Ideation and Psychiatric Symptoms

“…In medicine, NLP (text mining) and machine learning techniques have been mainly used for analysis of Electronic Health Records, and so assist clinicians in their work [37,38]. Nevertheless, to directly study patients' language is a promising field, especially in psychiatry, as language is the expression of thought, and, consequently, a window into the mind and emotions [39].…”

Finding the Best Match – A Case Study on the (Text-)Feature and Model Choice in Digital Mental Health Interventions

“…Such improvements are particularly needed in mental health care, where patients often go undiagnosed [39], and require long-time monitoring and care [40]. While many data types (e.g., log-in or questionnaire data) are available [18], text data presents itself as a propitious option in a eld that has always primarily relied on language for diagnosis and treatment [22,26,41]. Several research branches emerged to leverage text data's vast occurrence in the context of mental health [34].…”

“…Most of these attempts focus on user journey data, including log-in data and other indicators of online behavior [19][20][21]. At the same time, human language is the primary tool in psychiatry and psychology [22][23]. Accordingly, DMHIs often include asynchronous text-driven communication with participants, generally involving (1) open-text intervention exercises and (2) direct communication with e-coaches [24].…”