STATISTICAL APPROACHES FOR MULTIDIMENSIONAL SLEEP HEALTH: APPLICATION TO ALL-CAUSE MORTALITY IN OLDER ADULTS. (11th November 2018)

• Record Type: Journal Article
• Authors: Wallace, M L
• Is Part Of: Innovation in aging. Volume 2(2018)Supplement 1
• Page Start: 596
• Page End: 596
• Publication Date: 2018-11-11
• Subjects: Aging -- Periodicals; Gerontology -- Periodicals; 612.67
• DOI: 10.1093/geroni/igy023.2214
• Languages: English
• ISSNs: 2399-5300

Abstract

Abstract: Sleep is multidimensional, characterized by domains such as satisfaction, alertness, timing, efficiency, and duration. However, few studies use a multidimensional sleep health framework in relation to health outcomes. One possible reason is because this approach presents more methodological challenges. Dr. Wallace will demonstrate a range of statistical approaches to studying multidimensional sleep health, including: (1) composite scores quantifying the number of "good" or "bad" sleep domains; (2) multivariable regression; and (3) random forests that can be used to establish the importance of multidimensional sleep health relative to other established risk factors. These approaches will also be compared to a more traditional "unidimensional" approach focusing on individual dimensions of sleep. The benefits and disadvantages of each approach will be demonstrated using a large (N=8588) cohort of older adults aggregated from three publicly available datasets, each of which collected self-report sleep data and up to 15 years follow-up for mortality.