Automated Detection of Suicide Attempt Risk among Bereaved Individuals

自动检测丧亲者的自杀未遂风险

• 批准号: 9885577
• 负责人: Paul K Maciejewski
• 金额: $ 24.81万
• 依托单位: WEILL MEDICAL COLL OF CORNELL UNIV
• 依托单位国家: 美国
• 财政年份: 2020
• 资助国家: 美国
• 起止时间: 2020-01-01 至 2021-12-31
• 项目状态: 已结题
• 来源: https://reporter.nih.gov/project-details/9885577
• 关键词: Behavioral; Bereavement; Caring; Cessation of life; Clinical; Communities; Data; Data Collection; Data Set; Detection; Development; Diagnostic; Disease; Dreams; Feeling; Feeling suicidal; Future; Grief reaction; Healthcare; Home environment; Implicit Association Test; Individual; Life; Linguistics; Machine Learning; Measures; Memory; Mental Health; Modeling; Natural Language Processing; New York; Patient Self-Report; Patients; Persons; Reaction; Reporting; Research; Resources; Risk; Sampling; Savings; Severities; Subgroup; Suicide; Suicide attempt; Surveys; Techniques; Testing; Thinking; Time; Training; United Kingdom; Visit; base; end of life care; expressive writing; follow up assessment; follow-up; help-seeking behavior; informal support; loved ones; metropolitan; predictive modeling; predictive test; primary outcome; recruit; skills; suicidal behavior; suicidal individual; suicidal risk; suicide model; tool; web site; web-based tool; Behavioral; Bereavement; Caring; Cessation of life; Clinical; Communities; Data; Data Collection; Data Set; Detection; Development; Diagnostic; Disease; Dreams; Feeling; Feeling suicidal; Future; Grief reaction; Healthcare; Home environment; Implicit Association Test; Individual; Life; Linguistics; Machine Learning; Measures; Memory; Mental Health; Modeling; Natural Language Processing; New York; Patient Self-Report; Patients; Persons; Reaction; Reporting; Research; Resources; Risk; Sampling; Savings; Severities; Subgroup; Suicide; Suicide attempt; Surveys; Techniques; Testing; Thinking; Time; Training; United Kingdom; Visit; base; end of life care; expressive writing; follow up assessment; follow-up; help-seeking behavior; informal support; loved ones; metropolitan; predictive modeling; predictive test; primary outcome; recruit; skills; suicidal behavior; suicidal individual; suicidal risk; suicide model; tool; web site; web-based tool

Project Summary Bereaved individuals, especially those who meet criteria for Prolonged Grief Disorder (PGD) and those bereaved by suicide, have substantial risk of suicidal ideation, suicide attempts, and suicide. These vulnerable bereaved subgroups are also less likely to receive informal support or access mental health care, creating a pressing need to detect their suicide risk. Each year, well over a thousand unsolicited bereaved individuals visit our Cornell Center for Research on End-of-Life Care website and complete an online tool to determine if they meet criteria for PGD (over 6,000 completers to date). Compared to community-based bereaved samples, those who complete our online tool disproportionately meet criteria for PGD (e.g., ~30% vs. ~10%) and are suicide bereaved (~10% vs. ~2%). Thus, astoundingly, >35% of those who visit our Center website and complete our online diagnostic tool are at significantly elevated risk for suicidal ideation and/or attempts. To respond to the need to detect suicidal thoughts and behaviors (STBs) among our bereaved website visitors, we propose to develop a web-based tool for the detection of suicide attempt risk. We will leverage our Living Memory Home, an online memorial application residing on our Center website, by enhancing its features to optimize data collection, including data on potential implicit indicators of a bereaved person's suicide attempt risk (Aim #1). Data will be gathered on 100 Living Memory Home users daily for a week, followed by a 1-week, 1- and 6-month post-baseline follow-up assessment. The study will generate ~30 texts/subject in the Living Memory Home's Imagined Dialogues with the deceased, and reflections, dreams, stories, and touchstones in Narrative Notes (~3,000 texts in total from all users). Boot-strapping resampling, natural language processing and machine learning techniques will then be applied to develop machine learning models predicting suicide attempt risk based on bereaved subjects' baseline, 1-week, 1- and 6-month Columbia Suicide Severity Rating Scale scores (Aim #2). Our primary outcome will be the 1-week CSSRS score. We will, thus, develop and then pilot test an automated way to detect bereaved persons'; suicide attempt risk based on their interactions with the Living Memory Home. This is a first step toward development of a safe, accurate, potentially scalable online tool for the detection of suicide attempt risk among bereaved individuals.

项目摘要 丧亲的人，尤其是符合长期悲伤障碍标准（PGD）的人， 自杀，有自杀意念，自杀企图和自杀的巨大风险。这些脆弱 丧亲的亚组也不太可能获得非正式支持或获得精神卫生保健，创建一个 迫切需要检测其自杀风险。每年，超过一千个未经请求的丧亲者访问 我们的康奈尔大学研究中心在临终护理网站上，并完成在线工具，以确定他们是否是否 符合PGD的标准（迄今为止有6,000多个完成者）。与基于社区的丧亲样本相比 那些完成我们在线工具的人不成比例地符合PGD的标准（例如，〜30％对约10％） 自杀丧亲（〜10％vs.〜2％）。因此，令人惊讶的是，访问我们中心网站的人中，> 35％ 完成我们的在线诊断工具的自杀念头和/或尝试的风险显着升高。到 回应我们在丧亲的网站访问者中发现自杀思想和行为（STB）的需求，我们 建议开发一种基于网络的工具来检测自杀企图风险。我们将利用我们的生活 Memory Home是一个驻留在我们中心网站上的在线纪念应用程序，通过增强其功能 优化数据收集，包括有关丧亲者自杀未遂的潜在隐式指标的数据 风险（目标＃1）。数据将每天收集100个生活记忆用户一周，然后是1周 1个月和6个月的基线后随访评估。该研究将在生活中产生约30条文本/主题 记忆之家与死者的对话，以及在 叙述说明（所有用户总共约有3,000个文本）。启动式替换重新采样，自然语言处理 然后将应用机器学习技术来开发预测自杀的机器学习模型 尝试以丧亲的受试者的基准为1周，1周，1个月和6个月的哥伦比亚自杀严重程度等级的尝试风险 比例分数（AIM＃2）。我们的主要结果将是1周的CSSRS分数。因此，我们将发展，然后 飞行员测试一种自动化的方法来检测丧亲者'；自杀企图的风险基于他们与 生活记忆之家。这是开发安全，准确，可能可扩展的第一步 在线工具以检测死去的个体中的自杀企图风险。