Abstract
The Lifestyle for Brain Health (LIBRA) index, a weighted measure of modifiable risk factors for dementia, has been widely applied in European cohorts but less explored in US populations. We applied it to a cohort aged 65+ years, sampled from a relatively low-socioeconomic US region, and compared its predictive validity for incident mild cognitive impairment (MCI) and incident dementia risk. We used both reference factor weights derived from published meta-analyses and cluster-specific factor weights based on estimation results using our own cohort data. We also compared the original LIBRA index with an updated version (LIBRA2), which added three new risk factors including hearing impairment, sleep disturbance, and low social contact. We used joint modeling to account for informative attrition over time. Higher baseline LIBRA indexes were consistently associated with increased dementia risk across factor weight estimation methods. LIBRA2 demonstrated better predictive validity for incident dementia compared to MCI, further improving its performance. Cohort-specific factor weights strengthened associations with cognitive impairment risk but may limit comparability across studies. These findings contribute to valid use of LIBRA for dementia prediction in a late-life US cohort and support the incorporation of additional factors and cohort-specific adjustments to enhance its utility in diverse populations.