RDoC at Ten Years - Part 2: The Next Ten Years
In a previous message, I discussed how the Research Domain Criteria (RDoC) framework has changed the conversation in mental health research since its initiation in 2010. RDoC is based on the notion that studying psychopathology in terms of specific behavioral/psychological functions, and measuring these functions dimensionally, would result in more accurate descriptions of our patients and a more robust, straightforward pathway to understanding mental illnesses. Subsequent studies organized around RDoC principles have supported this notion, demonstrating the potential of the framework and identifying neurobiological mechanisms with greater fidelity. Building off these successes, I see a bright future for RDoC, as it embraces theoretical and computational approaches and strives for more precise and individualized ways to improve the lives of people living with mental illnesses.
The RDoC framework is built on the notion that various aspects of behavior can be represented as domains of function, like cognition or social function. Behavior within these domains can be further broken down into specific constructs — like working memory or cognitive control (specific forms of cognition), or attachment and facial recognition (specific forms of social function). These domains and constructs, as I have noted before, were determined through a consensus process in which experts in a given domain carefully considered existing data in defining constructs that were jointly determined by their behavioral functionality and implementing neural systems. The idea was not only to create an initial set of functions for research, but to provide a set of principles for the framework to grow with the rapid progress of science. This flexibility is exemplified in the RDoC positive valence domain, which has its origins in the brain “reward center” identified in 1954 and has since grown to encompass three constructs and nine sub-constructs. Now as RDoC passes its 10th anniversary, we seek to increase the focus on flexibility, testing and refining these constructs and exploring their clinical utility.
Evolving RDoC Through Theory and Computation
NIMH is pursuing this refinement in multiple ways. For example, we are interested in using theory-driven approaches to help investigate the neural mechanisms that implement these functional constructs. Theoretical accounts, accompanied by formal, mathematical descriptions, can help specify the details of a given behavior and uncover the neural circuits responsible for planning and executing that behavior. Working memory, for example, can be broken down into encoding, maintenance, and retrieval phases, each of which can be quantified precisely. Mathematical formulas can describe these phases; in these equations, specific variables represent parameters that need to be computed in order to carry out the behavior. One can then look to see how and where in the brain these variables are represented, and whether they might be represented differently in different individuals or in those with memory dysfunction due to a mental illness. NIMH supports research aimed at creating such mathematically formal descriptions of behavior, through a specific funding announcement .
At the same time, we are supporting data-driven efforts to measure function within and across domains. For example, we encourage studies that propose computational approaches to validate dimensional constructs relevant to psychopathology. Applied to RDoC, this initiative supports work that explores the relationships between measures of function — such as behavioral tests, neurophysiological variables, and self-reports — within and across constructs in order to identify convergent and divergent mechanisms. Measures of performance within a domain — comparing, for example, tests of working memory and cognitive control — should correlate better than across different domains — comparing working memory and attachment. Put in a simpler way, if cognition and social function are really separate domains, comprising different functions subserved by different brain systems, then individuals with poor working memory could be expected to also have (on average) relatively poor cognitive control, but not necessarily impaired attachment behaviors. Moreover, working memory deficits might map onto neural circuitry that overlaps with the circuits for cognitive control, but differs from that subserving attachment. Over the next several years, these theory- and data-driven approaches have the potential to refine RDoC constructs, explicate the relationships among the constructs, and help us explore the relationships between biology and behavior in novel and powerful ways.
In the meantime, we continue to work toward the development and optimization of tools to assess RDoC constructs . These efforts include such research as developing variants of tasks optimized for use in particular settings (for example, brief versions for high-throughput screening), creating versions that yield comparable construct validity across developmental stages, and adapting tests that have particular applicability for clinical trial outcome measures or biomarkers.
Translating RDoC: Towards Clinical Utility
Thinking a little bit further into the future, both data-driven and theory-driven approaches could be combined to evolve the RDoC framework. We envision developing theoretically defined, mathematically quantified behavioral tests that span the full panoply of RDoC domains and constructs. If designed properly, these tests could then be rolled out to a large number of people, generating datasets of unprecedented scale. These datasets could be used to further validate and refine the framework. Though we’re not quite ready to embark on a full-scale version of this project, NIMH staff have been overseeing the first cautious steps towards this vision. We’re developing a small test battery that can accommodate multiple behavioral tests of various RDoC constructs. Each test is “gamified”: short and engaging, and deliverable over a smartphone or web browser. Working with our colleagues in the NIH All of Us Research program , we plan on rolling out these tasks to hundreds of thousands of volunteers to pilot the process we’d like to take, eventually, with a more comprehensive battery of RDoC-relevant tasks.
But the real reason to work with the All of Us cohort is not just the ready availability of volunteers or the program’s robust patient privacy and data security safeguards. The data from these volunteers, which will eventually include 1,000,000 people living in the U.S., are linked to clinical and genomic datasets. Participants can also be recontacted, allowing us to invite them to join additional studies if they are willing. Thus, we can link these behavioral data to information about mental health, as well as real-world function, genetics, demographics, social determinants of health, and more. With additional studies, we could also link to biological factors like brain structure and activity information from neuroimaging and neurophysiology, epigenomic and biochemical evidence of environmental factors, and more. Such a dataset would have the potential to fulfill the RDoC goal of translating an understanding of mental function and dysfunction into a clinical knowledge base that could give us more useful information about our patients — information like what might be causing their illness, what their future course of illness might look like, and what treatments might work best. In other words, used this way, RDoC could pave the way to precision psychiatry.
The RDoC framework has driven considerable change over the past decade. To achieve the vision articulated here would mean even greater progress. It will require NIMH to continue to expand the scope of RDoC as an experimental framework that was always intended to change over time. It will require RDoC enthusiasts and skeptics alike to design the studies and gather the data necessary to refine this framework and its utility for generating and testing new models of psychopathology. It will require the increased involvement of our colleagues in computational and data science to ensure these studies are as informative as possible. And it will require the participation and engagement of our stakeholder communities, most notably those living with mental illness, to ensure that these studies are aimed squarely at developing needed solutions. Put these elements together, and we have a lot to look forward to in the next ten years of RDoC.
Olds, J., & Milner, P. (1954). Positive reinforcement produced by electrical stimulation of septal area and other regions of rat brain. Journal of Comparative and Physiological Psychology, 46(6), 419-427. doi: 10.1037/h0058775