County-Level Associations Between Social and Sleep Deprivation Conditioned by Regional Effects

Insufficient sleep leads to numerous negative health out-comes, and yet, approximately one-third of the adult U.S. population receives less than the recommended amount of sleep for optimal health. Consequently, sleep deprivation is now recognized as a public health crisis. Strategies to encourage optimal sleep must consider the social, environmental’ and cultural factors linked to insufficient sleep, although studies exploring these links are limited. This paper utilizes machine learning to address a gap in epidemiological modeling by conducting a joint county-level analysis of PLACES data and social deprivation measures across four U.S. regions. The innovation of this research stems from the use of random forest models, which include spatial autocorrelation structures within the modeling framework, commonly applied in ecology and earth science. Our findings reveal that incorporating spatial structures enhances the models' explainability by approximately 20%. Social deprivation measures can account for roughly 70% of the prevalence of sleep deficiency in the South and Midwest, but only about 55% in the Northeast and West, indicating that additional factors, such as the cultural glorification of sleep deprivation and urban lifestyles with lengthy commutes, may influence these regions. An analysis combining all four regions indi-cates that unemployment and educational attainment, rather than poverty as previously suggested, are the primary social deprivation factors affecting sleep duration. These insights could inform intervention strategies to promote optimal sleep, aligning with the objectives of Healthy People 2030.