Efficient sampling to enhance studies with longitudinal and clustered data

Deviations from simple random sampling are ubiquitous in epidemiological study design because they concentrate resources on informative subjects, thereby improving study efficiency.  Cost and resource efficient sampling strategies have not, however, permeated into longitudinal and correlated data settings due to the lack of methods to guide sampling and to analyze the enriched samples.  This symposium will highlight recent developments in efficient epidemiological study designs for longitudinal and clustered data.  We will first discuss the promise of enriched sampling designs in these settings.  We will then introduce several broad classes of designs, along with specific instances, that seek to improve power while acknowledging costs by identifying key features of available data to inform sampling strategies.  Designs discussed will include those that sample directly on the outcome data, those that sample on auxiliary data related to the outcomes, those that sample at the level of the cluster (e.g., subjects in longitudinal studies or clinics in multi-clinic studies), and those that sample within clusters (e.g., time-points within subjects in longitudinal studies or patients within clinics in multi-clinic studies). Whereas emphasis will be on study design, we will also introduce and connect various analytical approaches to data arising from these novel designs.