Symposia
Suicide and Self-Injury
Kate H. Bentley, Ph.D. (she/her/hers)
Clinical Psychologist
Massachusetts General Hospital
Boston, Massachusetts
Alexander Millner, Ph.D.
Research Associate
Harvard University
Cambridge, Massachusetts
Adam Bear, Ph.D.
Machine Learning Engineer
Harvard University
Cambridge, Massachusetts
Lia E. Follet, M.A.
Research Assistant
Harvard University
Brighton, Massachusetts
Rebecca Fortgang, Ph.D.
Postdoctoral Fellow
Harvard University
Cambridge, Massachusetts
Kelly Zuromski, Ph.D. (she/her/hers)
Research Associate
Harvard University
Cambridge, Massachusetts
Evan Kleiman, Ph.D. (he/him/his)
Assistant Professor
Rutgers University
Piscataway, New Jersey
Daniel Coppersmith, AM
Graduate Student
Harvard University
Cambridge, Massachusetts
Adam Haim, Ph.D.
Chief of the Treatment and Preventive Intervention Research Branch
National Institute of Mental Health
North Bethesda, Maryland
Suzanne A. Bird, M.D.
Director Acute Psychiatric Services Unit
Massachusetts General Hospital
Boston, Massachusetts
Matthew K. Nock, Ph.D. (he/him/his)
Research Scientist
Harvard University
Cambridge, Massachusetts
Real-time monitoring (namely, ecological momentary assessment [EMA]) is increasingly being used to better understand and predict suicidal thoughts and behaviors (STBs). There is a potential ethical obligation for researchers to intervene when receiving information about suicidal thoughts in real-time. A possible concern, however, is that intervening when receiving responses that indicate high risk for suicide during naturalistic EMA research on STBs may impact how participants respond to questions about suicidal thoughts, and thus affect the validity and integrity of collected data. We leveraged data from a large ongoing EMA study of adults and adolescents (N = 434) recruited during a hospital visit for STBs with the aim of understanding whether monitoring and intervening on high-risk responses affects subsequent participant responding. In this study, incoming real-time self-report data on suicidal intent are closely monitored and responses above an a priori “high-risk” threshold trigger a series of response-contingent automated pop-up messages and phone-based outreach from the study team. Regression discontinuity analyses were used to test whether the responses that followed response-contingent interventions were different from responses that did not follow response-contingent interventions. Overall, we did not find robust evidence to support the notion that intervening on high-risk responses influences study data. Although there was discontinuity in subsequent responses (to suicidal intent, suicide urge, and negative affect items) at the threshold used to trigger high-risk response-contingent interventions, it was not clear that such discontinuity was due specifically to the response-contingent interventions. The likelihood of responding to survey prompts also did not change from before to after receiving response-contingent interventions. Over one-fifth (22.13%) of all suicidal intent ratings that initially fell above the high-risk threshold were lowered by participants to below the threshold after viewing the automated pop-up messages but before submitting the survey, and adolescents were more likely than adults to change their initial suicidal intent ratings from above to below the high-risk threshold (p < .01). Future studies that are explicitly designed to assess the potential impact of intervening on high-risk responses in real-time research on STBs are needed, as this line of work will inform the optimization of effective, scalable strategies for intervening during moments of high suicide risk.
