AND AIMS: Suicide prevention is a top VHA priority. Suicide prevention in every health system is hampered by difficulties with predicting the risk of suicidal behavior, due to low base rates leading to very low positive predictive values. Very recently, machine learning regression tree methods have succeeded in better identifying a group at particularly high risk of suicide post-discharge from military hospitals. This advance is greatly needed, since the first months to year post-discharge has been repeatedly shown to be one of the very highest-risk periods for suicide that is known. Nevertheless, suicide and suicidal behavior risk prediction post-discharge (and at any other time) is still extremely challenging. For instance, this study's Principal Investigator has found that, in the relatively recent past, a large majority (73%) of VHA patients with depression denied suicidal ideation even when asked within 7 days of their suicide death. A clear need exists to develop measures of suicidal behavior risk that are not heavily dependent on patient self-report. Recently, our Co-Investigators conducted nonlinear dynamic analysis of movement data from non-Veteran inpatients and identified a signal that was correlated more strongly to suicidal ideation than any other characteristic tested. RESEARCH DESIGN: A prospective cohort study of 115-300 Veterans will be conducted to determine if the previously-identified specific actigraphy-based measurements highly associated with suicidal ideation in non- Veterans will predict suicidal ideation, suicidal behavior, and/or rehospitalizatin in Veterans. METHODS: An analysis of 115-300 Veterans admitted to the Bedford, Massachusetts VAMC acute psychiatry unit will be conducted. The primary analysis will focus upon 75-200 Veterans with current suicidal ideation or recent suicidal behavior (SI/SB) who do not have a primary psychotic disorder, Alzheimer's, or Parkinson's disease, and who are not undergoing alcohol detoxification. A separate analysis will be conducted of 40-100 patients undergoing alcohol detoxification, half with SI/SB and half without SI/SB. Participants will wear a small, unobtrusive, wristwatch-like actigraph on their nondominant wrist, and complete self-rated and clinician- rated assessments of suicidal ideation, as well as self-rated assessments of the severity of other psychiatric symptoms. A Resiliency Index (RI) will be calculated using nonlinear dynamic analysis of the amplitude of movements over time frames from 6 minutes - 2 hours. These time frames are the periods for which a clear structure to the movement data is evident, with patients with suicidal ideation showing less variation in amplitude than patients without suicidal ideation. If medications given for alcohol detoxification are determined to not interfere with the RI, then a secondary analysis will examine the entire sample of 115-300 Veterans.
One Aim will focus upon determining whether the original Resiliency Index or alternative movement data indices, such as one based on the change in the movement data over the hospitalization, predicts the presence and severity of suicidal ideation among Veteran inpatients.
This aim will also examine the sensitivity and specificity of the RI for detecting the presence of any suicidal ideation, and of substantial suicidal ideation. (In non- Veterans, the RI showed a sensitivity of 72% and a specificity of 100% for detecting any suicidal ideation, and 86% and 88%, respectively, for detecting substantial ideation).
The second Aim will determine whether the RI predicts subsequent suicidal behavior or rehospitalization over the next 1 month, 4 months, or 1 year after discharge, alone or combined with data about symptom severity, past history, and the present hospitalization. IMPACT: This study will contribute substantially to the VHA's high priority efforts to reduce suicide and suicidal behavior among Veterans. The approach studied here potentially likely particular value for suicidal behavior risk assessment in that it is not dependent on patient self-report of symptoms. This study is strongly supported by the VHA Suicide Prevention Program as a novel and potentially highly beneficial approach to suicidal behavior risk assessment, alone or combined with other readily available information.
Suicide risk assessment is highly imperfect and uncertain. This proposal seeks to improve suicide risk assessment for one of the most vulnerable, highest-risk populations known: patients recently discharged from psychiatric hospitalization. Our group has identified a novel predictor of suicidal ideation, a key component of suicide and suicidal behavior risk that patients sometimes fail to recognize or report. This predictor involves sophisticated analysis of patient movement data, and thus does not depend on patient self-report. We will expand and elaborate on our study in non-Veterans that showed a 'Resiliency Index' based on patient movement data correlated more strongly with suicidal ideation than any other characteristic. We will also examine the index's prediction of post-discharge suicidal behavior. Supported by VHA's Suicide Prevention Program, this study is a key step towards an integrated, multi-dimensional suicide risk assessment to protect our Veterans, especially those recently discharged from hospitalization.
