What Is an Intracluster Correlation Coefficient? Crucial Concepts for Primary Care Researchers

Primary care research often involves clustered samples in which subjects are randomized at a group level but analyzed at an individual level. Analyses that do not take this clustering into account may report significance where none exists. This article explores the causes, consequences, and implications of cluster data.

Using a case study with accompanying equations, we show that clustered samples are not as statistically efficient as simple random samples.