Maxwell Levis , Monica Dimambro , Joshua Levy , Brian Shiner
{"title":"Using Natural Language Processing to develop risk-tier specific suicide prediction models for Veterans Affairs patients","authors":"Maxwell Levis , Monica Dimambro , Joshua Levy , Brian Shiner","doi":"10.1016/j.jpsychires.2024.09.031","DOIUrl":null,"url":null,"abstract":"<div><div>Suicide is a leading cause of death. Suicide rates are particularly elevated among Department of Veterans Affairs (VA) patients. While VA has made impactful suicide prevention advances, efforts primarily target high-risk patients with documented suicide risk. This high-risk population accounts for less than 10% of VA patient suicide deaths. We previously evaluated epidemiological patterns among VA patients that had lower classified suicide risk and derived moderate- and low-risk groupings. Expanding upon VA's leading suicide prediction model, this study uses national VA data to refine high-, moderate-, and low-risk specific suicide prediction methods. We selected all VA patients who died by suicide in 2017 or 2018 (n = 4584), matching each case with five controls who remained alive during treatment year and shared suicide risk percentiles. We extracted all sample unstructured electronic health record notes, analyzed them using natural language processing, and applied machine-learning classification algorithms to develop risk-tier-specific predictive models. We calculated area under the curve (AUC) and suicide risk concentration to evaluate predictive accuracy and analyzed derived words.</div></div><div><h3>Results</h3><div>Our high-risk model (AUC = 0.621 (95% CI: 0.55–0.68)), moderate-risk (AUC = 0.669 (95% CI: 0.64–0.71)), and low-risk (AUC = 0.673 (95% CI: 0.63–0.72)) models offered significant predictive accuracy over VA's leading suicide prediction algorithm. Derived words varied considerably, the high-risk model including chronic condition service words, moderate-risk model including outpatient care, and low-risk model including acute condition care.</div><div>Study suggests benefit of leveraging unstructured electronic health records and expands prediction resources for non-high-risk suicide decedents, an historically underserved population.</div></div>","PeriodicalId":16868,"journal":{"name":"Journal of psychiatric research","volume":"179 ","pages":"Pages 322-329"},"PeriodicalIF":3.7000,"publicationDate":"2024-09-24","publicationTypes":"Journal Article","fieldsOfStudy":null,"isOpenAccess":false,"openAccessPdf":"","citationCount":"0","resultStr":null,"platform":"Semanticscholar","paperid":null,"PeriodicalName":"Journal of psychiatric research","FirstCategoryId":"3","ListUrlMain":"https://www.sciencedirect.com/science/article/pii/S0022395624005521","RegionNum":2,"RegionCategory":"医学","ArticlePicture":[],"TitleCN":null,"AbstractTextCN":null,"PMCID":null,"EPubDate":"","PubModel":"","JCR":"Q1","JCRName":"PSYCHIATRY","Score":null,"Total":0}
引用次数: 0
Abstract
Suicide is a leading cause of death. Suicide rates are particularly elevated among Department of Veterans Affairs (VA) patients. While VA has made impactful suicide prevention advances, efforts primarily target high-risk patients with documented suicide risk. This high-risk population accounts for less than 10% of VA patient suicide deaths. We previously evaluated epidemiological patterns among VA patients that had lower classified suicide risk and derived moderate- and low-risk groupings. Expanding upon VA's leading suicide prediction model, this study uses national VA data to refine high-, moderate-, and low-risk specific suicide prediction methods. We selected all VA patients who died by suicide in 2017 or 2018 (n = 4584), matching each case with five controls who remained alive during treatment year and shared suicide risk percentiles. We extracted all sample unstructured electronic health record notes, analyzed them using natural language processing, and applied machine-learning classification algorithms to develop risk-tier-specific predictive models. We calculated area under the curve (AUC) and suicide risk concentration to evaluate predictive accuracy and analyzed derived words.
Results
Our high-risk model (AUC = 0.621 (95% CI: 0.55–0.68)), moderate-risk (AUC = 0.669 (95% CI: 0.64–0.71)), and low-risk (AUC = 0.673 (95% CI: 0.63–0.72)) models offered significant predictive accuracy over VA's leading suicide prediction algorithm. Derived words varied considerably, the high-risk model including chronic condition service words, moderate-risk model including outpatient care, and low-risk model including acute condition care.
Study suggests benefit of leveraging unstructured electronic health records and expands prediction resources for non-high-risk suicide decedents, an historically underserved population.
期刊介绍:
Founded in 1961 to report on the latest work in psychiatry and cognate disciplines, the Journal of Psychiatric Research is dedicated to innovative and timely studies of four important areas of research:
(1) clinical studies of all disciplines relating to psychiatric illness, as well as normal human behaviour, including biochemical, physiological, genetic, environmental, social, psychological and epidemiological factors;
(2) basic studies pertaining to psychiatry in such fields as neuropsychopharmacology, neuroendocrinology, electrophysiology, genetics, experimental psychology and epidemiology;
(3) the growing application of clinical laboratory techniques in psychiatry, including imagery and spectroscopy of the brain, molecular biology and computer sciences;