The Value of Psychiatric Diagnoses.

Are we using the best diagnoses to make progress in psychiatric research? In the current issue of JAMA Psychiatry, we ask whether dimensional diagnoses should replace categorical diagnoses. Yee and coauthors1 argue in favor of such a change, and Weinberger and coauthors2 argue against it. Our statistical editor, Helena Chmura Kraemer, PhD, reflects on this debate.3 These invited articles were stimulated by the decision of the National Institute of Mental Health to focus psychiatric research on the Research Domain Criteria (RDoC), a new nosology for psychiatric disorders. There is unanimous agreement that DSM-5 is far from perfect. But what should be done? Do we need to reboot psychiatric research with a radically different diagnostic system? Or should we make progress from within the current system? Many critics point to the poor validity of the DSM-5 diagnoses. They are molded with a template pioneered by Emil Kraepelin. He conjectured that 3 different validators of psychiatric diagnoses (ie, clinical features, brain pathology, and etiology) converge on diseases defined by nature. He assured researchers that they could divide the labor of discovery as they set out to conquer mental illness. In the end, nature would reveal the psychiatric diseases to us. When he introduced this paradigm in 1896, it generated considerable enthusiasm. He was able to secure a $1 million grant from the Rockefeller Foundation to build the first Psychiatric Research Institute. His optimism has fueled psychiatric research ever since. But psychiatric research has turned out to be more complex. The validators of psychiatric diagnoses (several others have been added to the 3 proposed byKraepelin4) donot convergeonnatural entities.What is valid for a clinicianmightnot be valid for a brain researcher or a gene hunter. So far, nature has not revealed psychiatric diseases to us. But we can developdiagnostic constructswith increasingly greater utility in predicting future outcomes. The truevalueof apsychiatric diagnosis is the ability to predict course of illness, response to treatment, and, ultimately, quality of life and level of function in society. Good clinicians use diagnoses in the service of best patient care; they balance a paternalistic focus on outcomes with a respect of personal agency and encouragement for recovery. Are dimensions better than categories in getting us to our goal? Some clinicians are at ease with diagnostic categories: the psychiatrist who needs to manage a psychiatric crisis, or the forensic expert who is asked to assess the mental state of adefendant.However, othersmightpreferdimensions: theclinicianwho aims to understand the first-person experience of mental states or the psychologist who wants to explain humanbehavior.Suchpersonalpreferencesdonotneedtobemutually exclusive. Categories can be embedded in a dimensional systemanddimensionscanbeconverted intocategories. In psychiatry, we do not have to carve nature at its joints. But we do need to identify transition points, the pointswhere behavior shifts from normal to pathological, from one syndrome to another, and from one severity level to another. DSM-5 has included several dimensional approaches to psychopathology: symptomandseveritymeasures that cut across diagnoses; dimensions of psychosis that capture transitions between the main categories of functional psychoses; and a detailed, dimensional approach to personality. It is not clear how the new domains of the RDoC matrix map onto the current dimensions of psychopathology. There is the strong hypothesis that the RDoCmatrix is a better map for explorers of the humanmind. But the details of the RDoC domains (ie, thecoordinatesof thenewmap)need tobe tested. We have not reached a Copernican turn, when we replace a faulty clinicalmodelwith a correct brain-basedmodel. In fact, the RDoC domains are conjecture, ready for stringent experimental testing. These tests need to establish reliability inmultiple settings andpopulations,withproper correction formultiple comparisons. The experimental designs should probe causative relationships. Only then canwe consider RDoC as a viable option for a new psychiatric nosology. In the end, progress in psychiatric research will depend on our goals, not just the choice of dimensional or categorical approaches. The proponents of RDoC are correct that we need a much deeper exploration of neuroscience and genetics to advance a mechanistic understanding of mental illness. But the prediction of future outcomes will remain our North Star. In this journal, we hope to publish experiments that test how validators improve the clinical utility of a psychiatric diagnosis. Ideally, such studies include head-to-head comparisons of categorical anddimensional approaches. Suchvalidators (categorical or dimensional) can give rise to measures of clinical complexity and quality of care. Ultimately, we need methods that help us to capture clinical heterogeneity and move us toward more personalized psychiatric care. Whether we define an abnormal mental state categorically as brain disease or dimensionally as neurodiversity is a matter of personal preference. We know that we have made progress when our research generates new knowledge that leads to better outcomes for persons with a psychiatric diagnosis. Viewpoints pages 1159, 1161 and 1163 Opinion