Abstract
OBJECTIVE: To investigate the association of objective long sleep duration (LS) and insomnia with objective short sleep duration (ISSD) with mortality in older persons. METHODS: In 3,054 men (average age 76.4±5.5; mean follow-up=12.1 years) and 3,048 women (average age 83.6±4.8; mean follow-up=5.4 years), Cox proportional hazards models examined the association of LS (actigraphy-estimated sleep duration>8h) and ISSD (insomnia [difficulty initiating or maintaining sleep and/or sleep medication use ≥3/week] and concurrent actigraphy-estimated sleep duration<6h) with mortality. Other phenotypes (insomnia with normal sleep duration [INSD; insomnia and sleep duration 6-8h]; asymptomatic short sleep [AS; no insomnia and sleep duration<6h]) were also examined. Participants with normal sleep (NS; no insomnia and sleep duration 6-8h) served as the reference group. Models were adjusted for demographics and comorbidities. RESULTS: In unadjusted models, LS was associated with increased mortality in men and women when compared with NS. In women only, LS was associated with higher mortality after adjustment for demographics and comorbidity compared with NS (HR 1.30 [1.07, 1.59]). In demographic-adjusted models and across cohorts, ISSD was significantly associated with an increased hazard of mortality compared with NS (HR 1.25 [1.10, 1.43] for men; 1.36 [1.11, 1.67] for women). This association was not significant in either cohort after adjusting for comorbidity. Persons with INSD or AS did not have increased mortality risk compared with NS. CONCLUSION: LS and ISSD are at-risk phenotypes in older persons. Associations with mortality may be mediated by chronic diseases. Future work should examine whether sleep improvements decrease mortality in older persons.