Harmonization of social and physical health measures across prospective clinical studies of combat exposed service members and veterans: the total brain diagnostics program.

Introduction: Traumatic Brain Injury is prevalent during military service and is associated with short- and long-term psychosocial and functional changes, though comprehensive longitudinal data on Veterans and active-duty service members (SMs) are lacking. To address this research gap, we utilized data from the Long-term Impact of Military-relevant Brain Injury Consortium (LIMBIC) and the Translational Research Center for TBI and Stress Disorders (TRACTS), two prospective longitudinal cohort studies of veterans and active duty SMs, each containing a wide range of symptom scales, objective assessments, and health-related outcomes. Methods: This paper describes the innovative methods used to achieve the initial proof-of-concept harmonization for baseline psychosocial and physical health data from these two large cohort studies. To achieve harmonization, we gathered a multidisciplinary team with clinical and research expertise. We created a list of measures utilized by each study and organized them into larger clinically meaningful domains. When possible, we harmonized full measures, or single items directly, while others needed to be indirectly harmonized, by recoding, aligning categorical levels, and categorizing scales based on established cut scores. We calculated descriptive statistics to summarize and compare data. We then conducted Principal Component Analysis (PCA) for all continuous measures to assess whether site-level effects were observed in the shared variance. Results: A total of 73 variables capturing psychosocial function, sensorimotor, pain, and clinical health factors were harmonized across the LIMBIC and TRACTS studies. There were no differences in sex or ethnicity distributions between the studies. Sensory, social health, and health related clinical data were broadly comparable across cohorts, while pain intensity and headache disability were higher in LIMBIC. PCA analysis suggests data is suitable for pooled analysis. Discussion: We were able to directly harmonize multiple self-report measures of social well-being and indirectly harmonize other functional and demographic variables. While this initial effort focused on baseline data, the included principles can be employed to harmonize longitudinal data to increase the ability to detect clinical phenotypes to be applied in precision medicine approaches in future research.