Chapter 7: Evaluating Clinical Evidence

A significant portion of the chapter is dedicated to evaluating the validity of diagnostic tests by comparing them against a gold standard using 2x2 contingency tables. This process allows for the calculation of critical statistics, including sensitivity (the true positive rate) and specificity (the true negative rate), which are reinforced by the mnemonics SnNOUT and SpPIN to help clinicians rule diseases in or out. The discussion advances to predictive values (PPV and NPV), noting their dependence on disease prevalence, and introduces Likelihood Ratios (LR) as a more robust metric for revising pre-test probabilities into post-test probabilities, often utilizing tools like the Fagan nomogram or natural frequencies. The chapter also addresses the reproducibility of clinical findings, using Kappa scores to measure inter-observer agreement beyond chance and defining precision through the coefficient of variation. Furthermore, the text provides a rigorous guide for critically appraising medical literature, identifying potential sources of systematic error such as selection, performance, detection, and attrition bias. It explains how to interpret results from therapeutic and prevention trials by calculating experimental and control event rates, relative risk, absolute risk differences, and the Number Needed to Treat (NNT) or Harm (NNH). Finally, the chapter underscores the importance of effectively communicating these statistical concepts to patients to minimize framing effects and facilitate shared decision-making.