ScAN: Suicide Attempt and Ideation Events Dataset

Proceedings of the 2022 Conference of the North American Chapter of the Association for Computational Linguistics: Human Language Technologies, pages 1029 - 1040 July 10-15, 2022 ©2022 Association for Computational Linguistics ScAN: Suicide Attempt and Ideation Events Dataset Bhanu Pratap Singh Rawat1,*, Samuel Kovaly1,#, Wilfred R. Pigeon2,3,†, Hong Yu1,3,4,‡ 1CICS, UMass-Amherst, 2University of Rochester, 3U.S. Department of Veterans Affairs, 4Center of Biomedical and Health Research in Data Sciences *brawat@umass.edu, #skovaly@umass.edu, †wilfred_pigeon@urmc.rochester.edu, ‡hong_yu@uml.edu Abstract Suicide is an important public health concern and one of the leading causes of death world- wide. Suicidal behaviors, including suicide attempts (SA) and suicide ideations (SI), are leading risk factors for death by suicide. In- formation related to patients’ previous and cur- rent SA and SI are frequently documented in the electronic health record (EHR) notes. Ac- curate detection of such documentation may help improve surveillance and predictions of patients’ suicidal behaviors and alert medical professionals for suicide prevention efforts. In this study, we first built Suicide Attempt and Ideation Events (ScAN) dataset, a subset of the publicly available MIMIC III dataset spanning over 12k+ EHR notes with 19k+ annotated SA and SI events information. The annotations also contain attributes such as method of sui- cide attempt. We also provide a strong baseline model ScANER (Suicide Attempt and Ideation Events Retreiver), a multi-task RoBERTa- based model with a retrieval module to extract all the relevant suicidal behavioral evidences from EHR notes of an hospital-stay and, and a prediction module to identify the type of suici- dal behavior (SA and SI) concluded during the patient’s stay at the hospital. ScANER achieved a macro-weighted F1-score of 0.83 for identify- ing suicidal behavioral evidences and a macro F1-score of 0.78 and 0.60 for classification of SA and SI for the patient’s hospital-stay, re- spectively. ScAN and ScANER are publicly available1. 1 Introduction For decades, suicide has been one of the lead- ing causes of death (CBHSQ, 2020). The suicide rate in the United States increased from 10.5 per 100, 000 in 1999 to 14.2 in 2018, a 35% increase (Hedegaard et al., 2020). Globally, 740, 000 people 1The annotations, code and the models are availble at https://github.com/bsinghpratap/ScAN. Context: [*Name*] is a 22 year old male with a history of angry and impulsive behavior who is transferred from an outside hospital s/p Tylenol overdose. [*Name*] reports that he and his girlfriend broke up last Wednesday, and that he subsequently went on an alcohol and cocaine binge lasting from Thursday to Saturday. He has used alcohol and cocaine regularly in the past, but he denies having had a binge of this quantity or duration before. On Saturday night, [**Name**] told his father that he had tried to hang himself at a nearby park, but the rope had broken. Annotations: Two instances of suicide attempts are annotated in the paragraph. 1. [*Name*] ... overdose: Annotated for suicide attempt and is assigned 'unsure' category as there is no definite documentation that it is a suicide attempt. 2. On Sat ... broken: This is …

Content truncated to 500 words