Childhood traumatic brain injury and mid-to late-life risk of dementia and stroke: Neural network and traditional survival analyses in the Health and Retirement Study

Background Childhood traumatic brain injury (CTBI) is common but its long-term impact on neurological outcomes remains unclear. While previous studies suggest links between traumatic brain injury and dementia or stroke, most have examined adults or military cohorts. This study assessed whether CTBI is associated with later-life dementia and stroke using data from a large U.S. population-based cohort. Methods We used the Health and Retirement Study (HRS) Waves 9-15 to follow 10,768 adults free of dementia and stroke at baseline (2008). Childhood TBI (before age 16) was self-reported. We modeled dementia and stroke using IPTW-adjusted Cox/Fine–Gray analyses and a neural network. Results Among 10,768 HRS participants, 966 (8.9%) reported CTBI. Over 12 years, dementia occurred in 10.5% with CTBI versus 8.3% without (RR 1.27; RD 2.2%). Stroke showed a similar, modest pattern (8.1% vs 6.6%; RR 1.23). Cox/Fine–Gray estimates were imprecise overall, whereas the neural model discriminated well for both outcomes (AUC 0.827 dementia; 0.781 stroke). Conclusion In this nationally representative cohort, retrospectively reported childhood traumatic brain injury was associated with modest differences in later-life dementia and stroke risk, with signals most apparent in flexible neural network risk models. Findings should be interpreted cautiously and warrant replication in independent cohorts, particularly studies that capture injury severity, recurrence, biomarkers, and genetic susceptibility.