Mental Disorders as Homeostatic Property Clusters: A Narrative Review

Psychiatric classification faces longstanding challenges, including heterogeneous diagnostic categories, high comorbidity rates, limited interrater reliability, and modest clinical utility. Successive revisions of systems such as the DSM since 1980 have done little to resolve these issues. How to move forward?

This narrative review draws an analogy to biology, which has grappled for centuries with the hard problem of species classification. Following Darwin's insights that species are not fixed categories, contemporary theorists of classification have moved away from the natural kind view toward understanding species as homeostatic property clusters (HPCs): sets of properties contingently clustered in nature because the presence of some properties favors the presence of others. Probabilistic associations among these properties lead to imperfect aggregations and gray areas between species. This work adapts the HPC view for mental disorders, where probabilistic associations among biopsychosocial mental health properties form statistical aggregations: property clusters. These clusters are just as messy as in biology and usually lack sharp boundaries. Similar to species, diagnostic structures cannot be straightforwardly discovered in this space-they must be superimposed. To advance this view, a research agenda is outlined for mapping out a mental health atlas by identifying properties, their associations, and their dynamics and illustrating this idea using example data.

The HPC view accounts for many robust phenomena in mental health science, turning classification challenges from isolated anomalies into natural consequences of superimposing structure on the landscape of mental health problems. It aligns with major clinical and research frameworks-including Engel's biopsychosocial model, network and systems approaches, the Hierarchical Taxonomy of Psychopathology, and the Research Domain Criteria project-highlighting its role as a mental health science meta-framework. Doing so, it helps sidestep unproductive debates over the best universal classification system. Clinicians, researchers, and policymakers have different priorities and constraints, and no single classification system will optimally serve all stakeholders.