Treating diagnostic frameworks as information systems: normative implications and opportunities

The DSM-5 and ICD-11 are not just clinical manuals; they are information architectures. Every time a clinician applies diagnostic criteria, every time a researcher aggregates symptoms into a diagnostic variable, and every time a committee drafts a revision, they operate on the same basic object: a coded system for transmitting information about mental states. Yet no one has applied information theory [ 1 ] to evaluate the architecture of diagnostic systems themselves. Existing applications of information-theoretic thinking in psychiatry — model-selection criteria such as AIC and BIC — assess how well statistical models fit covariance data; they say nothing about the efficiency of the diagnostic code itself. This Viewpoint argues that information theory (henceforth, IT) — the elsewhere highly influential mathematical framework developed by Shannon to characterize the capacity and efficiency of communication systems [ 1 ] — provides what is currently missing: a principled, quantitative basis for designing, evaluating, and reforming diagnostic frameworks and the instruments that operationalize them. The case for this reframing is not merely theoretical. Diagnostic quality, the product of validity and reliability in the DSM-5 and ICD-11 is, by available estimates, remarkably poor -essentially entailing that the majority of variability stems from measurement error or other irrelevant factors [ 2 , 3 ].