Rethinking psychiatric diagnosis on the eve of DSM-5: A new paradigm for NIMH

I haven’t even ordered my copy of DSM-5 (for $199, mind you) and it may already be outdated.

As we near publication of DSM-5, the National Institute of Mental Health (NIMH) is reshaping the future of psychiatric diagnosis. NIMH Director Thomas Insel, MD, recently outlined a conceptual framework (first introduced in 2010) for diagnosis that is not necessarily based on current DSM-5 categories (currently defined by symptoms) and begins to transition psychiatry toward a diagnostic model that supplements symptom measures with genetics, objective indicators of brain circuitry and physiology, as well as behavioral and self-report data. This concept is based on the success of “precision medicine” strategies in cancer research, where cancer subtypes are quickly identified using a range of tests so that interventions can be more readily and effectively targeted, thus improving outcomes.

In a bold and controversial decision, NIMH is shifting its research funding away from traditional DSM categories (e.g., depression and schizophrenia) and toward cross-cutting categories of mental functioning and their measures, called the Research Domain Criteria (RDoC). This will be driving the NIMH research agenda, but it is unlikely to impact clinical practice anytime soon.

See NIMH Research Domain Criteria (RDoC) image.

Is this a mistake?

Translation to practice: One problem associated with NIMH divergence from DSM has to do with the already astounding gap that exists between research and practice in mental health. Since most of our treatment interventions are organized around DSM diagnoses, there is risk of further distancing research based on the RDoC framework from practice based on the DSM. If the RDoC model proves to be useful, we will need a solution to translate from RDoC to DSM and vice versa.

Prematurely dismissing what works:  While remarkable advances are being made in genetic and physiological measures of brain functioning, the current toolset of subjective standardized psychological assessments still provides superior validity compared to currently available objective measures, unlike cancer research. In fact, even in more “objective” spheres of medicine, there has come to be a growing appreciation for outcome measures based on the patient’s perspective.

DSM may not be the problem: While diagnostic accuracy in common practice is undeniably bad and unlikely to improve noticeably with changes planned for DSM-5, this unreliability does not appear to be due to the diagnostic framework itself. It is already possible to achieve high levels of reliability and validity using symptom-based structured interviews. This suggests that the DSM framework has value, but it is difficult to use in practice.

Despite Risks, RDoC is the Right Agenda for Mental Health

Fundamentally, our conceptualization of psychiatric diagnosis has been limited to the tools (and knowledge) we had available at the time. It was not long ago that psychiatry conceptualized most mental disorders within a psychodynamic and psychoanalytic tradition that had little to do with biology. Evidence today suggests that the symptoms of mental illness are manifestations of underlying biological processes, and as Dr. Insel suggests, it is time to challenge our current notions of diagnosis and seek a better understanding of the biological foundations of the brain so that we may develop more effective treatments and even cures.  If successful, this agenda will move mental health into the mainstream of health, where it belongs.

A Call for Big Data in Mental Health Research

In pursuing this bold research agenda, NIMH must be willing to consider more research of an exploratory nature. Due to the competitiveness of NIMH funding, research support usually goes to “sure bets,” i.e., experienced researchers running rigorous experiments to make incremental advances in knowledge. This works well, albeit slowly, in areas where the theory is well-developed, but there is no guarantee that the five RDoC domains are any more accurate reflections of biology than DSM. This is a different problem and a different era – an era of “big” data.

The effort to establish a new diagnostic foundation necessitates a different scientific approach, one perhaps better suited to philanthropic foundations that are willing to take big risks. One good example is the partnership between Google’s Sergey Brin, the Parkinson’s Institute, the Michael J. Fox Foundation, and 23andme. They are pooling their resources to combine vast quantities of genetic and other data to accelerate Parkinson’s research. The Fox Foundation is now experimenting with crowdsourced data analysis to machine learning teams with very promising results.

We need a similar effort in psychiatry. Instead of 100’s of discrete studies, can we feed the various RDoC data sources into a single research warehouse of psychiatric symptoms, behaviors, physiology, and genetics to facilitate exploratory data analysis? (Full disclosure: my organization, Centerstone Research Institute is building a repository of community mental health data to advance mental health research and policy.) It is about time for a data-driven reboot of psychiatric diagnosis, but it needs to leverage modern analytics.