Cognitive resilience: our cognitive health's strongest predictor that is the least understood?

This project studies whether "cognitive resilience" — operationalized as the residual of observed cognition minus the cognition predicted from plasma Alzheimer's disease biomarkers and demographics — carries prognostic value for incident cognitive impairment, dementia, and death, beyond what the biomarkers and a person's own observed test performance already indicate. The population is the Health and Retirement Study's HCAP–Venous Blood Study subsample (n=2,132), a nationally representative U.S. panel with linked plasma biomarkers, neuropsychological assessment, and up to 8 years of follow-up once the 2024 wave is available. The team has already analyzed the 2016–2022 portion of this cohort and found that participants in the top quartile of the resilience residual had substantially lower 6-year cumulative impairment than the bottom quartile, an association that survived adjustment for baseline cognition, biomarker burden, death as a competing event, and an extensive robustness suite, and that held under an independent re-derivation of the resilience construct itself (an "unexplained component" analysis modeled on total-factor-productivity residuals in growth economics). Those results — including two data-quality findings from an independent biomarker-verification review that is disclosed in full within the registration itself as prior evidence, not withheld. What this registration commits to prospectively is the 2024 extension: whether the same top-vs-bottom-quartile contrast, for both the original and the input-orthogonal resilience residual, continues to predict 8-year impairment, newly powers a dementia-specific endpoint that was inconclusive through 2022, and replicates under independent R and Stata implementations of the estimator — all specified here before any team member accesses the 2024-wave files. The purpose of registering now, rather than after the fact, is to make that specific boundary — what was already known versus what is genuinely being tested — verifiable by anyone reading this project.