0286 Presence and Potential Impact of Demographically Based Attrition in Pediatric Sleep Manipulation Research

Abstract

Abstract Introduction Pediatric sleep manipulation studies have mostly enrolled youth from dominant-culture, higher-income families. Studies with more diverse samples have been underpowered to detect differential attrition and demographic effect moderators. Here we pool data from two adolescent sleep manipulation protocols in a Midwest US city to better examine demographic differences in attrition rates and effect sizes. Methods Samples were pooled from studies detailed in Beebe et al. (2017; SLEEP, zsw035) and Duraccio et al. (2021; JSR, e13054). Both involved a sleep stabilization week, followed by 5-night periods of sleep restriction (6.5 hours/night in bed) and healthy sleep (9.5-10 hours/night in bed) in randomized counterbalanced order. Primary caregivers and 14-17 year-old adolescents each completed attention and sleepiness questionnaires for both conditions. Here we compare adolescents who were caregiver-identified as Black vs. White, the two largest local racial groups. Caregivers also reported their own education, family income, and household structure (single- vs. two-parent). Non-parametric tests looked for differential attrition and MANCOVA tested for racial differences in effects. Results Of the 257 initially enrolled, Black adolescents and those from households with one parent, lower income, or lower caregiver education were differentially lost to attrition (all p<.001), even though the racial makeup of the final sample approximated the regional population (36% Black, 64% White). In the final sample, Black and White youth were equally able to change their sleep (p>.90). Manipulation effects were significantly smaller for Black than White adolescents for inattention (self-report p=.026; parent-report p=.017) and sleepiness (self-report p=.002, parent-report p<.001), but these differential effects were non-significant after controlling for family income, household structure, and caregiver education (p>.05). Conclusion Even when a final study sample seemingly approximates the diversity of the local population, differential attrition may affect results. In this case, it superficially appeared that being in a non-dominant group (self-identified Black) was protective against the impact of short sleep. However, this effect disappeared after controlling for demographic risk factors. Participation in sleep manipulation studies can be challenging, so families facing higher burdens may need more support; otherwise, only the most resilient of those families may succeed, which could distort findings. Support (If Any) NIH (R01HL120879, R01HL092149)