Comparison Groups Matter in Traumatic Brain Injury Research: An Example with Dementia

Abstract

The association between traumatic brain injury (TBI) and risk for Alzheimer disease and related dementias (ADRD) has been investigated in multiple studies, yet reported effect sizes have varied widely. Large differences in comorbid and demographic characteristics between individuals with and without TBI could result in spurious associations between TBI and poor outcomes, even when control for confounding is attempted. Yet, inadvertent control for post-TBI exposures (e.g., psychological and physical trauma) could result in an underestimate of the effect of TBI. Choice of the unexposed or comparison group is critical to estimating total associated risk. The objective of this study was to highlight how selection of the comparison group impacts estimates of the effect of TBI on risk for ADRD. Using data on Veterans aged ≥55 years obtained from the Veterans Health Administration (VA) for years 1999–2019, we compared risk of ADRD between Veterans with incident TBI (n = 9440) and (1) the general population of Veterans who receive care at the VA (All VA) (n = 119,003); (2) Veterans who received care at a VA emergency department (VA ED) (n = 111,342); and (3) Veterans who received care at a VA ED for non-TBI trauma (VA ED NTT) (n = 65,710). In inverse probability of treatment weighted models, TBI was associated with increased risk of ADRD compared with All VA (hazard ratio [HR] 1.94; 95% confidence interval [CI] 1.84, 2.04), VA ED (HR 1.42; 95% CI 1.35, 1.50), and VA ED NTT (HR 1.12; 95% CI 1.06, 1.18). The estimated effect of TBI on incident ADRD was strongly impacted by choice of the comparison group.