Dataset Information

Improving ascertainment of suicidal ideation and suicide attempt with natural language processing.

ABSTRACT: Methods relying on diagnostic codes to identify suicidal ideation and suicide attempt in Electronic Health Records (EHRs) at scale are suboptimal because suicide-related outcomes are heavily under-coded. We propose to improve the ascertainment of suicidal outcomes using natural language processing (NLP). We developed information retrieval methodologies to search over 200 million notes from the Vanderbilt EHR. Suicide query terms were extracted using word2vec. A weakly supervised approach was designed to label cases of suicidal outcomes. The NLP validation of the top 200 retrieved patients showed high performance for suicidal ideation (area under the receiver operator curve [AUROC]: 98.6, 95% confidence interval [CI] 97.1-99.5) and suicide attempt (AUROC: 97.3, 95% CI 95.2-98.7). Case extraction produced the best performance when combining NLP and diagnostic codes and when accounting for negated suicide expressions in notes. Overall, we demonstrated that scalable and accurate NLP methods can be developed to identify suicidal behavior in EHRs to enhance prevention efforts, predictive models, and precision medicine.

SUBMITTER: Bejan CA

PROVIDER: S-EPMC9452591 | biostudies-literature | 2022 Sep

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Improving ascertainment of suicidal ideation and suicide attempt with natural language processing.

Bejan Cosmin A CA Ripperger Michael M Wilimitis Drew D Ahmed Ryan R Kang JooEun J Robinson Katelyn K Morley Theodore J TJ Ruderfer Douglas M DM Walsh Colin G CG

Scientific reports 20220907 1

Methods relying on diagnostic codes to identify suicidal ideation and suicide attempt in Electronic Health Records (EHRs) at scale are suboptimal because suicide-related outcomes are heavily under-coded. We propose to improve the ascertainment of suicidal outcomes using natural language processing (NLP). We developed information retrieval methodologies to search over 200 million notes from the Vanderbilt EHR. Suicide query terms were extracted using word2vec. A weakly supervised approach was des ...[more]

PMID: 36071081

Similar Datasets

Project description:BackgroundMedical training poses significant challenge to medical student wellbeing. With the alarming trend of trainee burnout, mental illness, and suicide, previous studies have reported potential risk factors associated with suicidal behaviours among medical students. The objective of this study is to provide a systematic overview of risk factors for suicidal ideation (SI) and suicide attempt (SA) among medical students and summarize the overall risk associated with each risk factor using a meta-analytic approach.MethodsSystemic search of six electronic databases including MEDLINE, Embase, Education Source, Scopus, PsycInfo, and CINAHL was performed from database inception to March 19, 2021. Studies reporting original quantitative or epidemiological data on risk factors associated with SI and SA among undergraduate medical students were included. When two or more studies reported outcome on the same risk factor, a random-effects inverse variance meta-analysis was performed to estimate the overall effect size.ResultsOf 4,053 articles identified, 25 studies were included. Twenty-two studies reported outcomes on SI risk factors only, and three studies on both SI and SA risk factors. Meta-analysis was performed on 25 SI risk factors and 4 SA risk factors. Poor mental health outcomes including depression (OR 6.87; 95% CI [4.80-9.82] for SI; OR 9.34 [4.18-20.90] for SA), burnout (OR 6.29 [2.05-19.30] for SI), comorbid mental illness (OR 5.08 [2.81-9.18] for SI), and stress (OR 3.72 [1.39-9.94] for SI) presented the strongest risk for SI and SA among medical students. Conversely, smoking cigarette (OR 1.92 [0.94-3.92]), family history of mental illness (OR 1.79 [0.86-3.74]) and suicidal behaviour (OR 1.38 [0.80-2.39]) were not significant risk factors for SI, while stress (OR 3.25 [0.59-17.90]), female (OR 3.20 [0.95-10.81]), and alcohol use (OR 1.41 [0.64-3.09]) were not significant risk factors for SA among medical students.ConclusionsMedical students face a number of personal, environmental, and academic challenges that may put them at risk for SI and SA. Additional research on individual risk factors is needed to construct effective suicide prevention programs in medical school.

Project description:ObjectiveThe authors sought to identify predictors of imminent suicide attempt (within 30 days) among U.S. Army soldiers following their first documented suicidal ideation.MethodsUsing administrative data from the Army Study to Assess Risk and Resilience in Servicemembers, the authors identified 11,178 active-duty Regular Army enlisted soldiers (2006-2009) with medically documented suicidal ideation and no prior medically documented suicide attempts. The authors examined risk factors for suicide attempt within 30 days of first suicidal ideation using logistic regression analyses, including sociodemographic and service-related characteristics, psychiatric diagnoses, physical health care visits, injuries, and history of family violence or crime perpetration or victimization.ResultsAmong soldiers with first documented suicidal ideation, 830 (7.4%) attempted suicide, 46.3% of whom (N=387) attempted suicide within 30 days (rate, 35.4 per 1,000 soldiers). Following a series of multivariate analyses, the final model identified females (odds ratio=1.3, 95% CI=1.0, 1.8), combat medics (odds ratio=1.6, 95% CI=1.1, 2.2), individuals with an anxiety disorder diagnosis prior to suicidal ideation (odds ratio=1.3, 95% CI=1.0, 1.6), and those who received a sleep disorder diagnosis on the same day as the recorded suicidal ideation (odds ratio=2.3, 95% CI=1.1, 4.6) as being more likely to attempt suicide within 30 days. Black soldiers (odds ratio=0.6, 95% CI=0.4, 0.9) and those who received an anxiety disorder diagnosis on the same day as suicidal ideation (odds ratio=0.7, 95% CI=0.5, 0.9) were less likely.ConclusionsSuicide attempt risk is highest in the first 30 days following ideation diagnosis and is more likely among women, combat medics, and soldiers with an anxiety disorder diagnosis before suicidal ideation and a same-day sleep disorder diagnosis. Black soldiers and those with a same-day anxiety disorder diagnosis were at decreased risk. These factors may help identify soldiers at imminent risk of suicide attempt.

Project description:BackgroundAssessing patients' suicide risk is challenging, especially among those who deny suicidal ideation. Primary care providers have poor agreement in screening suicide risk. Patients' speech may provide more objective, language-based clues about their underlying suicidal ideation. Text analysis to detect suicide risk in depression is lacking in the literature.ObjectiveThis study aimed to determine whether suicidal ideation can be detected via language features in clinical interviews for depression using natural language processing (NLP) and machine learning (ML).MethodsThis cross-sectional study recruited 305 participants between October 2020 and May 2022 (mean age 53.0, SD 11.77 years; female: n=176, 57%), of which 197 had lifetime depression and 108 were healthy. This study was part of ongoing research on characterizing depression with a case-control design. In this study, 236 participants were nonsuicidal, while 56 and 13 had low and high suicide risks, respectively. The structured interview guide for the Hamilton Depression Rating Scale (HAMD) was adopted to assess suicide risk and depression severity. Suicide risk was clinician rated based on a suicide-related question (H11). The interviews were transcribed and the words in participants' verbal responses were translated into psychologically meaningful categories using Linguistic Inquiry and Word Count (LIWC).ResultsOrdinal logistic regression revealed significant suicide-related language features in participants' responses to the HAMD questions. Increased use of anger words when talking about work and activities posed the highest suicide risk (odds ratio [OR] 2.91, 95% CI 1.22-8.55; P=.02). Random forest models demonstrated that text analysis of the direct responses to H11 was effective in identifying individuals with high suicide risk (AUC 0.76-0.89; P<.001) and detecting suicide risk in general, including both low and high suicide risk (AUC 0.83-0.92; P<.001). More importantly, suicide risk can be detected with satisfactory performance even without patients' disclosure of suicidal ideation. Based on the response to the question on hypochondriasis, ML models were trained to identify individuals with high suicide risk (AUC 0.76; P<.001).ConclusionsThis study examined the perspective of using NLP and ML to analyze the texts from clinical interviews for suicidality detection, which has the potential to provide more accurate and specific markers for suicidal ideation detection. The findings may pave the way for developing high-performance assessment of suicide risk for automated detection, including online chatbot-based interviews for universal screening.

Dataset Information

Improving ascertainment of suicidal ideation and suicide attempt with natural language processing.

Publications

Improving ascertainment of suicidal ideation and suicide attempt with natural language processing.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets