Strategies for Establishing Clinical-Decision Thresholds in Psychiatry: A Review

Nearly all psychopathology is inherently continuous, but both clinical practice and research often require categorical decisions (eg, treat or wait, pay for treatment or decline, enroll in study or not); therefore, thresholds are applied to continuous phenomena to guide these decisions. The most common strategy is to select cutoffs by focusing on diagnoses, but this may be unhelpful in many decisions (eg, selective prevention, choice between treatment options). This review evaluates strategies for developing thresholds for psychopathology dimensions that may enhance clinical decision-making, drawing on methods used in psychiatry, psychology, internal medicine, and health economics.

One strategy used extensively in neuropsychology, child psychiatry, clinical psychology, and laboratory medicine is to define multiple thresholds (eg, mild, moderate, and severe), using statistical deviance vis-à-vis the general population. A second strategy is to select thresholds in reference to functional impairment, as has been done for some psychopathology measures. A third approach, common in internal medicine but underused in psychiatry, references probability of a negative outcome (eg, mortality). A fourth option, gaining momentum in European psychiatry, is to select thresholds based on costs and benefits of a clinical action at different levels of severity or risk. Common examples of these 4 strategies are reviewed, comparing their strengths and limitations. They are relevant not only to clinical measures but also research instruments. Value judgments are inherent in approaches 2, 3, and 4; thus, selection of such thresholds requires stakeholder input. Importantly, clinicians make decisions by considering all information, and thresholds are only 1 factor.

Although diagnosis-focused thresholds are sometimes valuable, the field would be served best by also considering the aforementioned alternative strategies. These approaches have been greatly underused in psychiatry, and studies are urgently needed to address this gap. Herein, recommendations are offered for this research.