Statistical Methods for Predicting Suicide Attempt

Columbia University Health Sciences, New York NY

Investigators

Linked publications & trials

Abstract

[unreadable] DESCRIPTION (provided by applicant): This application for a Mentored Quantitative Research Career Development Award has been submitted with the goal of supporting the development of the applicant's career as a quantitative researcher in psychiatry. The training and research plan is intended to build on the candidate's prior work in the statistical analysis of clinical, biological and gene expression data on mood disorders. The application delineates plans for training and mentoring in the areas of biostatistics, bioinformatics, clinical diagnoses and neurobiology of depression, and genomic analysis necessary to enable the applicant to pursue an independent career of scientific inquiry in psychiatric statistics. These goals will be met by a combination of supervision and didactic course work in the Division of Neuroscience of the Department of Psychiatry of Columbia University. The aim of the Research Plan is to develop and evaluate efficient and flexible statistical methodology for the prediction of suicide attempts and completed suicides in patients with mood disorders. For the purposes of this study, suicide attempt will be defined as a self-destructive act with at least some intent to end one's life. The focus will be on developing and evaluating predictive models based on a large number of retrospective and longitudinal measurements on each subject, as opposed to identification of individual risk factors. Aim 1 proposes to build/evaluate classification models for suicide attempters/nonattempters based on complete data. Aim 2 will build/evaluate survival analysis models when some of the subjects have censored data. Aim 3 will concentrate on classification of post-mortem brain samples from suicide victims and normal controls as a method of identifying putative risk factors for suicidal behavior. This will include large-scale gene expression analysis as well as the analysis of receptor bindings and other biochemical information in suicide victims and normal controls. Relevance: Suicide causes over 30,000 deaths annually in the USA, and the number of suicide attempts is thought to be at least several times that. The aim of this study is to develop and compare efficient and flexible statistical methodology for the prediction of future suicide attempts based on demographic, clinical and biological data tor the use of health professionals so that patients at risk for suicide can be identified and treated in advance. [unreadable] [unreadable] [unreadable]