Using multilevel analysis in patient and organizational outcomes research.

BACKGROUND Outcomes research often compares patient and organizational outcomes across institutions, dealing with variables measured at different hierarchical levels. A traditional approach to analyzing multilevel data has been to aggregate individual-level variables at the institutional level. OBJECTIVES To introduce the conceptual and statistical background of multilevel analysis and provide an example of multilevel analysis that was used to examine the relationship between nurse staffing and patient outcome. METHODS A two-level model was presented employing multilevel logistic regression analysis. RESULTS Outputs from multilevel analysis were interpreted. Other statistics were presented for model specification and testing. CONCLUSION Researchers should consider multilevel modeling at the study design stage to select theoretically and statistically sound research methods.