Project Summary/Abstract Suicide is a devastating public health problem. Over 40,000 people die by suicide each year in the United States and it is the fourth leading contributor to years of life lost. The suicide rate has not changed over time, prediction of suicide has not improved over time, and the efficacy of interventions has not changed over time. In order to improve the understanding, prediction, and prevention of suicide, there is an urgent need for precise conceptualizing, operationalizing, and describing of suicidal phenomena. Suicidal thoughts are an antecedent of suicidal behavior and a central part of the pathway to suicide. Many features of suicidal thoughts, such as their duration, are largely unknown. Preliminary descriptive work has used smartphones to observe how suicidal thoughts unfold in daily life and consistently found that suicidal thoughts seem to ebb and flow within- people over time. In light of the accumulating evidence of the variability of suicidal thinking, theorists are beginning to argue that suicide should be viewed through the lens of dynamical systems. Included in these theories is the notion that there are multiple discrete states of suicide risk that people move through over time. Despite the powerful theoretical and clinical implications of suicidal states, no empirical work has tested and validated suicidal states. The proposed project aims to address this major gap in suicide research by combining computational modeling and dynamic real-time data to capture suicidal states. Aim 1 of the project is to identify the number of within-person suicidal states. To achieve this aim, Hidden Markov Models, a form of computational model that identifies hidden discrete states in dynamic data, will be applied to multiple real-time measures of suicidal thinking from an ongoing NIMH-funded intensive longitudinal study of suicidal thoughts and behaviors (N = 300). Aim 2 is to capture the duration of suicidal states. The temporal dynamics of suicidal states will be operationalized as when participants transition between states and on average how long participants stay in a given state. Aim 3 is to test which suicidal states are predictive of near-term risk of suicidal behavior. An existing event-time prediction model framework will be applied to generate interpretable and precise event-time predictions of suicide attempts for each type of suicidal state. The proposed study’s greatest potential impacts are to provide foundational information on suicidal thinking and to generate suicidal states that could be used in future Just-in-Time-Adaptive Interventions for suicide prevention. The proposed study also promotes a program of research that includes two major NIMH priorities of suicide prevention and computational psychiatry. If successful the proposed project would advance the understanding of suicidal thinking which could one day improve the prediction and prevention of suicidal behavior.