자살행위 예방을 위한 임상의사결정지원시스템의 효과 및 시행

This study is designed to test the effects of providing emergency department (ED) clinicians with enhanced information about their patients' risk of a suicide attempt in the 6-months after leaving the ED. The study will be conducted in five EDs. The primary study will be conducted in four EDs within Mass General Brigham (MGB): Massachusetts General Hospital (MGH), MGB-Salem, Brigham and Women's Hospital (BWH), and Brigham and Women's Faulkner Hospital (BWFH). The investigators propose to enroll 4,000 participants over the study period and to provide the treating clinician with a risk score for a randomly selected half of those 4,000 participants. The primary outcome will be rate of suicide attempts over the next 6-months. It is anticipated that provision of this risk score will change clinician decision-making, leading clinicians to be more likely to hospitalize high-risk patients and less likely to hospitalize low-risk patients. A smaller (N=250) pilot implementation study has been added in a new non-MGB ED at Cambridge Health Alliance (CHA).

The primary study will enroll 4,000 patients presenting to the ED at MGH MGB-Salem, BWH, and BWFH. Study inclusion criteria include presentation to the ED, adult status (≥18 years-old), and ability to provide an email address.

Exclusion criteria include the presence of any factor that impairs an individual's ability to comprehend and effectively participate in the study, including an inability to speak or write English fluently, the presence of gross cognitive impairment due to florid psychosis, intellectual disability, dementia, acute intoxication, or the presence of extremely agitated or violent behavior. These broad inclusion criteria maximize the clinical applicability of results while the exclusion criteria ensure the ethical principle of respect for persons.

According to prior studies in the MGB system, investigators anticipate the MGB sample will be about 73% White/European American, 9% African-American, 4% Asian American, and the remainder of other or mixed ethnicity. The implementation study will enroll 250 patients presenting to the ED at CHA and will have the same inclusion/exclusion criteria as the primary study. Based on recent numbers reported in the CHA 2021 Annual Report, it is anticipated that the CHA sample will be about 37% White/European American, 16% African- American, 9% Asian American, and the remainder of other or mixed ethnicity.

Study Research Assistants (RAs) will use well-established protocols of reviewing patients presenting to the ED, where investigators have been conducting IRB approved research on the prediction of suicidal behavior for >15 years. RAs will first consult potentially eligible patients' assigned clinician to obtain their permission to approach the patient to briefly describe the study and ask if they are interested in learning more about it. Only those patients determined to meet the inclusion/exclusion criteria will be approached. To further ensure communication between the research and clinical team, investigators will give a presentation about the study to the ED clinical teams before starting recruitment and at regular intervals throughout the award period. RAs will obtain written informed consent from participants (using REDCap, described below) while in the ED, and administer a baseline self-report assessment battery (on an iPad). Contact information (including the full name, phone number, and if available, email address for both the participant and a collateral contact for emergency situations) will be collected via secure REDCap data collection surveys at baseline, and with participants' permission, select health data (e.g., readmission for a suicide attempt over the next 6 months) will be extracted from the EHR (and stored in a de-identified database). Participants will receive an email-based survey to assess the receipt of treatment and the presence of a suicide attempt at 1-month and 6-months after their ED visit. Participants will receive financial incentives for each survey completed.

In the primary study, after completing a brief baseline assessment, participants will be randomly assigned (1:1) to have their clinician receive a predicted probability that this patient will make a suicide attempt after leaving the ED. There is no intervention with the patient themselves. Investigators hypothesize that clinicians receiving this evidence-based prediction score will be more likely to hospitalize high-risk patients and less likely to hospitalize low-risk patients and that this will lead to a decrease in the rate of suicide attempts in the experimental condition relative to the control condition. The primary endpoint in this study is suicide attempt during the six months post-randomization. Suicide attempts will be measured using two methods: (a) self-report surveys and (b) documentation in the EHR during the follow-up period. Self-report questions about suicide attempts will be those from the Self-Injurious Thoughts and Behaviors Interview (SITBI), a widely used measure of suicidal thoughts and behaviors with strong reliability and validity and one of the measures recommended in the PhenX toolkit. Investigators also will examine all new medical records during the 6-months following each participant's index ED visit to determine if the participant revisited the hospital due to suicide ideation or suicide attempt during the follow-up period. This approach has been used in prior studies to measure such outcomes. In prior work in the APS we have found high agreement between self-reports on the SITBI and the documentation of suicide attempts in electronic medical records (K=.84). Secondary endpoints are suicide attempt within 1-month post-randomization. Additional exploratory hypotheses are: (1) among patients estimated by the model to be high-risk, clinicians of patients in the experimental condition will be more likely to recommend hospitalization than clinicians of patients in the control condition; (2) among patients estimated to be at intermediate risk, clinicians of patients in the experimental condition are (a) less likely to recommend hospitalization and (b) more likely to recommend partial hospitalization compared to clinicians of patients in the control condition, and (3) among patients estimated to be at low risk, clinicians of patients in the experimental condition are more likely to discharge to home than clinicians of patients in the control condition. Investigators will additionally investigate the mediating role of clinician choice on the association between the intervention and patient outcomes using mediation analysis.