Abstract
Health services research frequently focuses on variation in the structure, process, and outcomes of clinical care. Robust approaches for detection and attribution of variation are foundational to both quality improvement and outcomes research. Describing care in structured healthcare systems across hospitals in which clinicians work to provide care for patients as a multileveled structure allows the impact of organization on practice and outcome to be ascertained. Mixed-effect statistical models can describe both the partitioning of variation among levels of these structures and by inclusion of explanatory variables the valid estimation of the features of health systems, clinicians, or patients, with observed differences in processes or patient outcomes. In this Readers' Toolbox, the authors describe the rationale for considering healthcare structures when assessing clinical practice, outcomes, and sources of variation. They describe statistical considerations and methods for the estimation of analysis of structured data and assessment of variance.