Aging
Subjective visuospatial cognition and dementia risk in adults over age 90: a comparison of time-to-event methods Alexander Ivan B. Posis* Alexander Ivan Posis Posis Posis Posis Posis Posis Posis Posis University of California, Davis
Background: Visuospatial deficits are common in dementia. Yet, associations of visuospatial cognition and incident dementia are unknown in adults 90+. In this population with high mortality rates, it’s unclear if dementia risk estimates are sensitive to underlying time-to-event model assumptions. We estimated associations of subjective visuospatial cognition and dementia risk in adults 90+ with 4 different time-to-event methods.
Methods: We included 778 LifeAfter90 participants (mean±SD age=92±2; 63% women; 36% ≥college education). Baseline subjective visuospatial cognition was assessed using a 7-item visuospatial Everyday Cognition scale (ECogVis) subscale. Scores were summed and averaged where greater scores suggest greater visuospatial impairment (range=1-4). Time-to-dementia was assessed every 6 months from baseline. Associations of ECogVis and dementia risk were estimated with 4 multivariable models: Cox proportional hazards (PH), Fine-Gray (FG), Aalen additive (AA), and Weibull accelerated failure time (AFT).
Results: Over 1983 person-years of follow-up, there were 101 dementia diagnoses (60/1000 person-years) and 272 deaths (137/1000 person-years). Worse baseline ECogVis scores yielded higher hazard of dementia in PH models, with similar estimates whether mortality was treated as a censoring event (HRCox PH=1.40, 95%CI 0.95,2.05) or competing risk (HRFG=1.37, 95%CI 0.96,1.96; Table 1). The AA model showed an absolute increase of 60 more dementia cases per 1000 person-years (AA SErobust=0.70 per 1000 person-years) for every 1 unit increase in the ECogVis score and the AFT model showed faster time-to-dementia (time ratio=-0.16; 95%CI -0.35,0.03).
Conclusions: While estimates were in the expected direction, low subjective visuospatial cognition was not associated with greater dementia risk in those with exceptional longevity. In this very high mortality population, findings from four time-to-event models were qualitatively similar in direction of estimates.
