Diagnostic accuracy describes how reliably a system identifies the correct condition or clinical interpretation. It is assessed by comparing its conclusions with an appropriate reference standard, such as a confirmed diagnosis or expert-reviewed clinical outcome. Measures may include sensitivity and specificity, but the right metrics depend on the task and the risks of errors.
In clinical AI, accuracy matters because an incorrect interpretation can lead to missed conditions, unnecessary tests, or unsuitable treatment. It should be evaluated alongside cost and clinical appropriateness: a cheaper recommendation is not necessarily more accurate or safer. In practice, healthcare organizations validate systems on representative cases, monitor performance after deployment, and require clinician review where needed. Changes to prompts or model versions can alter recommendations, so those changes should be controlled and retested.