Harmonizing Veteran Longitudinal Cohorts to Identify Prognostic Factors in Post-Traumatic Brain Health

Portland Va Medical Center, Portland OR

Investigators

Abstract

The TBI literature provides preliminary evidence supporting associations between TBI exposures and poor health and functional outcomes, particularly as they occur in the context of the traumatic and stressful circumstances of military deployment. Ongoing longitudinal cohort studies of service members and Veterans with military combat and training exposures offer researchers access to extant data addressing factors that modify risks for developing and/ or recovering from a range of brain disorders, including TBI, PTSD, pain, depression, and suicidality. These longitudinal studies of participants with TBI present an opportunity to accurately categorize this risk by harmonizing the overlapping elements from two or more VA-funded resources. Pooling individual participant level data from longitudinal TBI research studies will result in a large enough dataset to consider relevant moderators, mediators, and confounders in analyses and allow for more impactful and clinically meaningful findings. In order to address the present knowledge gaps and harmonize largescale, multi-modal data from varied sources, well- planned and reproducible standardization, curation, and dissemination is needed to allow for meaningful analyses. Recent advances in FAIR (Findable, Accessible, Interoperable, Reusable) data methods can guide these efforts and ensure efficient and accurate results with large and similar enough data. Our multiple-investigator proposal team has identified two longitudinal cohort studies that can address the above-mentioned Post-traumatic Brain Health knowledge gaps in the relatively near term. Longer-term prospects from this work include incorporation of other potential largescale datasets that can be added once the initial harmonization and FAIR data methods development are established. The Long-term Impact of Military-relevant Brain Injury Consortium's Prospective Longitudinal Study (PLS) is a 10-year, 17-site cohort of >2,500 service members and veteran participants with combat-exposure who are well-characterized initially and then have annual reassessments, and the Translational Research Center for TBI and Stress Disorders is a 14-year, 2-site longitudinal cohort of >925 veterans with combat-exposure who are deeply characterized initially and then undergo comprehensive reassessment at 2, 5 and 10 years. Our proposed leadership team has experience pooling and harmonizing data from large studies that measure similar concepts with a variety measurements and disparate variable names. We have developed a model data harmonization system to combine data from multiple heterogenous studies and facilitate the analysis of a single pooled dataset. The system combines study files into a single data model and leverages natural language processing techniques to facilitate data pooling and harmonization using algorithmic approaches. Notably, this team's preliminary success resulted in being the first and only team to create and disseminate FITBIR metadata and resulted in ongoing funding to continue this work from DoD/ CDMRP. While this work is similar to the work proposed here, TRACTS data are not yet incorporated into FITBIR, and the preliminary FAIR data methods we developed will be applied to TRACTS data as part of this proposal to increase efficiency and accuracy of our proposed LIMBIC/TRACTS harmonization work. The research team will develop a 2-year plan to: 1) define a standardized approach and create a crosswalk for the modalities and domains of LIMBIC and TRACTS longitudinal cohort to create a unified assessment profile, 2) identify a rigorous harmonization approach to allow for overall data analyses using the unified data set, 3) identify an multi-modal, analytic approach for the harmonized data sets, 4) explore the use of a VA-supported machine-learning approach as a possible tool-set in identifying multimodal patterns relevant to predicting posttraumatic brain health, 5) develop and propose a research program to standardize, harmonize and analyze the prospective, longitudinal dataset to identify risk factors associated with brain disorders and recovery and 6) develop, pilot test, and propose follow-up FAIR data methods to apply the systems developed in this proposal to incorporate additional, largescale, longitudinal military data to this robust data resource.