Background: Key parameters that underlie the association between sleep and cognitive decline are still controversial. Furthermore, reversing the impact of sleep on cognition is yet to be explored. Objective: To investigate the prevalent sleep disturbances that induce cognitive changes over two years in non-demented elderly. Patients and Methods / Material and 

Methods: Sleep-related parameters (bedtime, sleep duration, latency, quality, efficiency, and daytime dysfunction) were evaluated in a nationwide, population-based cohort (1,485 with normal cognition [NC] and 628 with mild cognitive impairment [MCI]) by Pittsburgh Sleep Quality Index, and was related to the risk of the cognitive decline or the chance of its recovery over 2 years. We stratified participants according to the baseline cognitive status and used logistic regression model in a full-model fit. 

Results: In participants with MCI, prolonged sleep increased the risk of dementia by approximately 4-fold (odds ratio [OR], 3.62; 95% confidence interval [CI], 1.02–12.80). Newly developed prolonged sleep within 2 years increased the risk of dementia by 8-fold (OR, 8.11 [95% CI, 1.89–34.82]), while the odds of reversion to NC were almost doubled if sleep duration decreased within 2 years (OR, 1.85 [95% CI, 1.07–3.20]). In participants with NC, however, long sleep latency, not prolonged sleep, increased the risk of cognitive decline by approximately 60% (OR, 1.63 [95% CI, 1.11–2.40]). 

Conclusion: Prolonged sleep may be a modifiable risk factor of dementia in people with MCI, but not in those with NC. Long sleep latency may be an early sign of cognitive disorders in people with NC.