LAUSR

Objective To estimate the proportion of late-life cognitive loss attributable to impending death. Methods Older persons (n = 1,071) in a longitudinal cohort study without dementia at enrollment underwent annual cognitive assessments (mean 10.6 years, SD 4.6, range 4–24) prior to death. We estimated the onset of terminal acceleration in cognitive decline and rates of decline before and after this point in change point models that allowed calculation of the percent of cognitive loss attributable to terminal decline. Outcomes were composite measures of global and specific cognitive functions. We also estimated dementia and mild cognitive impairment (MCI) incidence before and during the terminal period. Results A mean of 3.7 years before death (95% credible interval [CI] −3.8 to −3.5), the rate of global cognitive decline accelerated to −0.313 unit per year (95% CI −0.337 to −0.290), a more than 7-fold increase indicative of terminal decline. The mean global cognitive score dropped 0.377 unit (SD 0.516) assuming no terminal decline and 1.192 units (SD 1.080) with terminal decline. As a result, 71% (95% bootstrapped CI 0.70, 0.73) of overall global cognitive loss was terminal. In subsequent analyses, terminal decline accounted for 70% of episodic memory loss, 65% of semantic memory loss, 57% of working memory loss, 52% of perceptual speed loss, and 53% of visuospatial loss. MCI incidence in the preterminal and terminal periods was similar, but dementia incidence was more than 6-fold higher in the terminal period than preterminal. Conclusion Most late-life cognitive loss is driven by terminal decline.