Statistical issues in the identification of risk factors for suicidal behavior: the application of survival analysis

Abstract

Studies of suicide risk factors generally examine suicidal behavior as a dichotomous outcome. Survival analytic techniques are discussed in which the time until a suicide attempt from a specific point, such as prior attempt or onset of illness, is also examined. These procedures can incorporate information on those lost to followup or "censored." One survival analytic technique, Cox's proportional hazards model, is a particularly informative statistical technique for the study of suicidal risk factors because several covariates can be incorporated. Illustrative analyses estimate the significance of different risk factors, and demonstrate that there is a 32% increase in the relative risk of a suicide attempt associated with each prior attempt.