Abstract Introduction Sleep health evolves in late adulthood, yet the relative contributions of short- and long-term fluctuations in sleep difficulties to dementia risk remain unclear. Using a nationally representative dataset, we sought to inform whether patterns of sleep difficulty were prospectively associated with incident dementia risk among older adults in U.S. Methods We analyzed 6,376 adults ≥65 years in the nationally representative National Health and Aging Trends Study (NHATS; 13-year follow-up, with three sample refreshes). Sleep difficulties were characterized as: (1) difficulty falling asleep; or (2) maintaining sleep each year. Frequency was collected on 5-point Likert scales from rarely to every night. We derived: (1) cumulative sleep reserve, defined as the time-updated proportion of prior study years without sleep difficulties (score from 0% to 100% good-sleep years); (2) recent sleep trajectory, categorized as improving (less frequent), stable (no change), or worsening (more frequent) in the preceding three years; (3) sleep stability, defined as the within-person standard deviation of sleep difficulty frequency. Incident dementia was determined by self-reported diagnosis or performance on validated measures. We used Cox proportional hazard modeling to ascertain incident dementia risk. Results Dose–response analyses showed a smooth, graded decline in dementia risk with increasing cumulative sleep reserve. Each additional 10% of accumulated good-sleep years was associated with ~5–6% reduction in dementia hazard (difficulty falling asleep: HR=0.95,95%CI:0.94–0.96,p 0.001; difficulty maintaining sleep: HR=0.94,95%CI:0.93–0.95,p 0.001). Among adults with high cumulative sleep reserve, a recently worsening trajectory was associated with increased dementia risk (falling asleep: HR=1.26,95%CI:1.13-1.39,p 0.001), relative to a recently stable trajectory. Finally, greater within-person variability in reported sleep difficulties was associated with higher dementia risk (falling asleep: HR=1.82,95%CI:1.56–2.13,p 0.001; maintaining sleep: HR=1.43,95%CI:1.18–1.75,p 0.001). Conclusion Cumulative sleep reserve showed a strong, graded prospective dose–response association with dementia risk, with each incremental increase in proportion of good-sleep years providing additional protection. The finding that recent worsening in those with high reserve increases dementia risk may reflect the preclinical/prodromal stage of dementia. Our analysis revealed that greater within-person fluctuation in reported sleep difficulties was associated with higher dementia risk, underscoring the importance not only of how good sleep is but how consistently good it remains over time. Support (if any)
Goel et al. (2026) studied this question.