Prediction of Risk and Resilience in Psychosis and Bipolar Spectrum Disorders: A Translational Multimodal Study

About This Grant

This K99/R00 award will position the candidate to become an independent clinical researcher with expertise in individualized novel phenotyping and prediction of risk and resilience to psychosis and bipolar spectrum disorders (PBSD). Background. PBSD are among the most disabling conditions worldwide, evidenced by poor quality of life and premature mortality. These disorders demonstrate pluripotentiality and heterotypic continuity across clinical, cognitive, and neural phenotypes. The ability to predict transdiagnostic functional outcomes is critical for implementing precision-based interventions. Despite advances in identifying shared risk factors and pathophysiological mechanisms, translating research findings into clinical practice remains a challenge. Specific Aims. This project synthesizes data from NIMH-sponsored clinical high-risk (CHR) cohort, the North American Prodrome Longitudinal Study 2 and 3 (NAPLS-2, NAPLS-3) and Accelerating Medical Partnerships – Schizophrenia (AMP-SCZ), and translates empirical findings to electronic health records (EHR). Aim 1.1 will leverage the NAPLS cohorts to identify novel combinations of demographic, social determinants of health (SDOH), clinical, cognitive, and biological factors associated with risk, remission, and resilience using machine learning. Aim 1.2 will externally validate these models in AMP-SCZ and investigate the predictive power of digital phenotyping measures. Aim 2.1 will apply temporal deep learning and explainable artificial intelligence (XAI) to test these predictive models and align CHR variables with unique XAI-derived common data elements in the demographically-diverse Epic EHR using a longitudinal retrospective design. Aim 2.2 will design a clinician-facing nomogram for future deployment as an automated real-time predictor of PBSD as preparation for an R01 application. Training. The candidate will achieve these goals through a resource-rich institutional environment and cohesive training plan in: (1) PBSD etiology and course, including SDOH and immunological biomarkers; (2) advanced statistical modeling and machine learning techniques; and (3) optimization of EHR tools and registries. This training will support the development of an independent research program integrating novel digital and EHR phenotyping with clinical practice. Mentorship. The candidate will be supported by an expert interdisciplinary team: Robert Bilder, Ph.D. (primary mentor), Carrie Bearden, Ph.D. (co-mentor), David Miklowitz, Ph.D. (co-mentor), Steven Cole, Ph.D. (consultant), and Douglas Bell, Ph.D. (consultant). Impact. This project directly aligns with the NIMH’s Strategic Goals related to the pressing need to improve assessment platforms within healthcare to screen, detect, and treat mental illnesses; optimizing real-world data collection systems with computation modeling; evaluating the role of social determinants of health in the onset and course of mental illness; and developing decision-support tools for interventions and stepped care. The research outcome will develop innovative methods to prospectively identify individuals likely to demonstrate risk, remission, or resilience, enabling real-time individual- and population-level detection and interventions.