To advance sleep science, let's study change

Abstract

Sleep is critical for physical, cognitive, and mental health, but how sleep supports these domains likely fluctuates across the lifespan. While traditional observational and experimental study designs-often cross-sectional or limited (two-wave) longitudinal designs-have provided valuable insights, they fall short of capturing the dynamic nature of sleep and its effects over time. To fully understand these complex and evolving relationships, multimethod, multi-time point longitudinal designs are required. These approaches can illuminate the temporal dynamics of sleep and its outcomes, offering stronger and potentially causal conclusions. In this article, we aim to empower sleep scientists, clinicians, and trainees with research methods focused on studying change-methods that can be applied across both observational and experimental designs. To truly advance the field, it is critical to examine sleep throughout the lifespan, from infancy through older adulthood, with repeated and nuanced assessments of sleep and its related outcomes. We outline a variety of statistical analysis approaches and corresponding design considerations that support the rigorous study of change in sleep. Finally, we offer forward-looking recommendations for scientific training, research program evaluation and funding, and the development of research infrastructure and collaborations. Together, these strategies have the potential to propel the field of sleep research forward, generating richer insights and change-based conclusions. Statement of Significance As a field, we are driven by a fundamental question: does sleep temporally precede and cause changes to our physical, mental, and cognitive health? While many existing studies use cross-sectional or limited (two-wave) longitudinal designs, these approaches often fall short of capturing the full picture needed to understand the timing and impact of sleep. Encouragingly, we have the tools and methods needed to pursue this important work. In this article, we highlight statistical approaches and research designs that can help move the field forward. With thoughtful application of these methods, we can strengthen our conclusions, generate more impactful findings, and bring us closer to understanding the role of sleep across the lifespan.

Keywords: aging; behavioral sleep medicine; causality; data analysis; longitudinal studies; multivariate analysis; public health; research design; statistics.

Figure 1

(A) Graphical abstract illustrating longitudinal trajectories of sleep duration, physical activity, and body fat [26]. Within- and between-person trajectories are depicted based on previous publications whenever possible [27–30]. Examples of key life events are shown at the bottom of panel (A), representing potential influences that could be incorporated into longitudinal research designs. (B) The annual relationship between sleep (black) and physical activity (purple). Solid lines represent group-level averages, while dashed lines illustrate an example individual’s data. (C) Example of a 7-day consecutive data collection approach. Light columns represent days and dark columns represent nights. A cross-sectional study using only one or two time points might incorrectly conclude there is no relationship between the three variables. However, with three or more time points, a clearer pattern emerges, revealing day-to-day coupling of sleep and physical activity. (D) Example of evaluating sleep duration and body fat by age (tracking an individual over time), by cohort (comparing generational changes), or by period (evaluating historical influences).