Unpacking the 2 × 2 table

Abstract

Statistics estimated from a 2 × 2 contingency table, such as sensitivity, specificity, and likelihood ratios, are usually presented in contexts of diagnostic decision making or evaluation of screening tests. However, their use is not restricted to those contexts. The purpose of this article is to summarize for clinicians and researchers in nursing and other health care disciplines the conceptual basis, calculation, application, and limitations of statistics derivable from a 2 × 2 table and to provide a worked example based on previous research. It will be shown that these statistics can potentially be used for analyzing any binary outcome (e.g., occurrence vs. nonoccurrence of some event) with a binary predictor (e.g., present vs. absent or above vs. below some cut point). Most can be estimated with a hand-held or online calculator, and, apart from chi-square statistics, without first converting observed data to a test-statistic. Clearer understanding of the derivation and clinical meaning of these statistics will encourage their wider use in health research and clinical practice.