Reliability of diagnosis

Abstract

Reliability of a diagnosis can be measured as either agreement with a reference criterion or agreement between clinicians as to a diagnosis. Most psychiatric and some general medical diagnostic constructs are identified on the basis of a clinical picture, not using a reference criterion such as laboratory tests. Most psychiatric diagnostic constructs have moderate to substantial reliability in research studies. They are likely to be less reliable in clinical practice. Measures such as standardizing interviews can improve reliability. General medical diagnostic constructs have similar reliability to psychiatric diagnostic constructs in research studies and are also likely to be less reliable in clinical practice. Even with laboratory tests, some medical conditions are hard to distinguish due to similarities in their clinical pictures. For alternative mental health classifications, psychological formulation—except psychodynamic formulation—is less reliable than psychiatric diagnosis. Symptom-based classification has at least equal reliability to psychiatric diagnosis. Dimension-based classification has equivalent reliability to psychiatric diagnosis. These may be combined with diagnosis but may have less usefulness on their own for other functions, e.g. administrative.