ObjectiveEvidence concerning whether sleep disturbances in older adults predict mortality is mixed. However, data are limited to self-reported sleep problems and may be confounded with other comorbidities. We examined whether electroencephalographic (EEG) sleep parameters predicted survival time independently of known predictors of all-cause mortality.MethodsA total of 185 healthy older adults, primarily in their 60s through 80s, with no history of mental illness, sleep complaints, or current cognitive impairment, were enrolled in one of eight research protocols between October 1981 and February 1997 that included EEG sleep assessments. At follow-up (mean [SD] = 12.8 [3.7] years after baseline, range = 4.1–19.5), 66 individuals were positively ascertained as deceased and 118 remained alive (totalN= 184).ResultsControlling for age, gender, and baseline medical burden, individuals with baseline sleep latencies greater than 30 minutes were at 2.14 times greater risk of death (p= .005, 95% CI = 1.25–3.66). Those with sleep efficiency less than 80% were at 1.93 times greater risk (p= .014, CI = 1.14–3.25). Individuals with rapid eye movement (REM) sleep percentages in the lowest 15% or highest 15% of the total sample’s distribution (percentage of REM <16.1 or >25.7) were at 1.71 times greater risk (p= .045, CI = 1.01–2.91). Percentage of slow-wave sleep was associated with time to death at the bivariate level, but not after controlling for potential confounders.ConclusionsOlder adults with specific EEG sleep characteristics have an excess risk of dying beyond that associated with age, gender, or medical burden. The findings suggest that interventions to optimize and protect older adults’ sleep initiation, continuity, and quality may be warranted.