SUMMARY Bipolar disorder (BD) is a mood disorder with high recurrence and disability rates, high economic burden, and an estimated suicide risk 20 times higher than the general population. While efficacious treatment is available, BD patients spend a large proportion of their life symptomatic. Predicting the onset of episodes is a valuable strategy to decrease suicide and disability rates and to optimize healthcare costs. The overall objective of this (R21) Exploratory/Developmental study is to obtain pilot data to support the feasibility and potential value of a new approach to predict mood episodes in stable adult patients with BD. This proposal aims to develop new data modeling and inference techniques that will enable more tailored clinical signal detection: examining changes within each individual, over time. To do so, we propose integrating multimodal, high-dimensional telemonitoring data, nonlinear techniques and artificial intelligence classification systems. This approach builds on our preliminary work on: (i) nonlinear techniques for the study of mood regulation in BD; (ii) an award-winning simulation using a machine learning technique (Markov Brains) for episode prediction in BD. AIMS: Aim 1 (feasibility): To obtain and integrate multimodal data to perform time-series analysis and to calculate entropy levels in 90 euthymic BD adults. Exploratory Aim 2 (potential value): To use machine learning techniques (Markov Brains) to distinguish participants at higher risk for a depressive or manic relapse based on their time-series and entropy levels (from Aim 1). HYPOTHESES: H1: We will be able to collect enough data in 80% of our participants and to integrate multimodal data to perform time-series analysis and to calculate entropy levels. H2: Markov Brains will identify participants at higher risk for a mood episode based on high (vs. low) auto-correlated time-series and low (vs. high) entropy levels. SIGNIFICANCE: This R21 application challenges more traditional prediction models by conceptualizing inter- and intra-individual variability as a dynamic property of biological systems. By leveraging densely-sampled objective and subjective data, autonomic, clinical and demographic data, this proposal aims to develop inference techniques that will examine changes within each individual, over time, in order to enhance the estimation performance. Ultimately, if we develop the capacity to predict mood episodes, we should be able to prevent them.

NARRATIVE The proposed research is clinically relevant as it could develop new data modeling and inference techniques that will enable more tailored clinical signal detection: examining changes within each individual, over time, in order to predict mood changes indicative of an imminent episode of illness in bipolar disorder. The project supports the goals of strategy 2.2 of NIMH Strategic Research priorities by creating predictive algorithms at the individual level (what is an individual’s digital signature like, and how well does deviation from that signature predict an event?)