The Research Domains Criteria (RDoC) initiative has proposed to overcome limitations in the existing diagnostic taxonomy by investigating new ways of classifying mental disorders based on dimensions of observable behavior and neurobiological measures. A fundamental challenge is determining the validity of the implied relations among these measures both within and across levels of analysis. Using an existing database, this project aims to implement novel data-analytic strategies to examine the validity of selected RDoC domains: working memory (maintenance, updating), and cognitive (effortful) control (response inhibition/suppression). We will accomplish this in two separate Aims: (1). Examine the cross-level relations of selected genetic variants, self-report, behavioral, and MRI measures using Bayesian network models. We propose related analytic approaches for each construct, first identifying measurement models at each available level, and then using exploratory methods (ESEM, MIMIC) to interrogate relations across dimensions. (2). Examine the cross- level relations of selected genetic variants, self-report, behavioral, and MRI measures using Bayesian network models. In the RDoC framework identified by the NIMH workgroups, there is an implied hierarchical structure among different levels of measurement. Using Bayesian network models, we will create cross-level models to investigate whether the hierarchical structure proposed by the RDoC working group is validated in the data, using both observed and latent measures within each level. The existing database includes extensive phenotyping of these RDoC dimensions at diverse levels of analysis including: self-reports, clinical rating scales, clinical diagnostic interview schedules, neuropsychological measures, experimental cognitive measures, and genome-wide genotyping assays. All these data types were acquired in 153 patients, including those with schizophrenia (SZ, n = 58), bipolar disorder (BP, n = 49) and ADHD (n = 46). Healthy volunteers (n =1,137) received all personality, neurocognitive measures and genotyping, along with the ASRS for ADHD screening and the SCID for diagnosis. Additional fMRI and MRI neuroimaging data were obtained in a subset of 128 healthy people and all 121 patients. Among the deliverables of this research will be objective determination about whether selected measures of neural circuit integrity (from structural and functional MRI methods) are truly intermediate phenotypes (i.e., do they mediate relations from genetic to behavioral measures) for both working memory and cognitive control. We will establish whether the dimensions of working memory and response inhibition are consistent across healthy controls and patient groups (ADHD, BP, SZ).
We propose using novel analytic methods to determine the psychometric validity of constructs identified by the RDoC Workgroups for both working memory and cognitive control using exploratory structural equation modeling (ESEM) and Bayesian network models, from the genome to the self-report levels. Public Health Relevance: The proposed research implements novel data-analytic strategies to examine the validity of selected RDoC domains, developing methods for cross-level modeling that could be deployed in the future in larger knowledge bases. Creating innovative models across levels will become more necessary as new data are accumulated across multiple levels of analysis, and as these are deposited into national and international repositories.
