Introduction to the ABAI Practice Community Blog and Practice Board member co-author bios. Please note that this blog is co-written by members of the ABAI Practice Board. Special thanks to Tom Walz for his contributions to this entry.
In Part 2 of this series on making sense of diagnosis categories, we provide a brief overview of three current research efforts to build a better system for understanding clinical presentations (i.e., alternatives to DSM and ICD categories, see Part 1: What is a diagnosis?) with the hope that improved strategies for classification will lead to improvements in matching assessment to the supports and treatment resources needed to improve functioning.
Are the Diagnostic and Statistical Manual of Mental Disorders (manual) and the International Classification of Diseases (ICD) psychiatric systems the best approach for (a) detecting problems that are severe enough to interfere with social, academic, and occupational functioning, (b) connecting them to effective interventions, and (c) advancing our understanding to provide more people with improved quality of life? You might be surprised that former directors of the National Institute of Mental Health (NIMH) have criticized the DSM saying, “The weakness is the lack of validity (Insel, 2013) and “…totally wrong…in fact what they produced was an absolute scientific nightmare (Hyman, as cited in Belluck & Carney, 2013, May 6).”
Shortcomings of the prevalent diagnostic systems, many of which are acknowledged in the latest editions of the manuals themselves, are:
- Diagnoses are based on observed behavior rather than physiological or etiological patterns. Although the model is medical, evidence does not support a disease model.
- Although the manual lists conditions that contribute to social, academic, or occupational difficulties, such as financial or housing insecurity, history of interpersonal violence, barriers to education, and meaningful employment, diagnostic codes that characterize such conditions are seldom utilized.
- Similarly, the manual outlines steps for assessing sociocultural factors in clients’ lives, but they receive short shrift within the diagnostic process.
- Evidence suggests that there are no special processes (or processes different in kind from other, everyday behavioral processes) through which behavioral health conditions arise.
- The current system is non-dimensional. It’s a threshold model – either a case/client meets the diagnostic criteria, or it does not. Such a threshold model is not supported by evidence.
- The classification system, while using some specifiers (e.g., “mild,” “major”), does not allow or account for a continuum of behavioral health.
- Many people suffer and do not meet criteria for diagnoses. Their presentations fall under the “other/not otherwise specified” category.
- Sociocultural or sociopolitical influences dictate “clinical significance.” There is a risk that diagnostic categories and thresholds to meet them reflect sociocultural norms, coerce conformity, and limit diversity.
- The manuals are works in process – categories are emerging or redefined over successive editions.
- Between diagnostic categories, there is substantial overlap (also called “co-morbidity” in the medical model), calling into question the categories themselves.
- Within diagnostic categories, presentations are so heterogeneous that the categories poorly characterize individual situations.
- Diagnostic categories are not supported by other sciences (e.g., neuroscience); there is little cohesion between different levels of scientific analyses.
- Current diagnostic systems have not advanced our comprehensive knowledge and understanding.
You might find it difficult to imagine a world in which DSM diagnoses are irrelevant, yet we already have one foot in this world. With regard to research, NIMH has not required DSM and ICD categories for research funding since 2013. Instead, efforts to build a comprehensive and cohesive system that spans the sciences (e.g., behavioral sciences, neuroscience, molecular sciences) are underway. We will describe three efforts below, explain how behavioral sciences fit within each of these, and what their implications might be for clinical practice.
- The Research Domain Criterion (RDoC) Program
In 2009, NIMH launched an ambitious framework designed to promote interdisciplinary collaboration, connecting physiology, biology, chemistry, neurosciences, and the behavioral sciences to promote potential new points of intervention linked to the assessment of relevant mechanisms targeted by the intervention. Given that the diagnostic categories of the current classification systems do not hold across levels of analysis, researchers organized new domains based on behavioral constructs and multi-level units of analysis. For example, the domain positive valence includes processes such as reward satiation, which in turn can be investigated with molecular, cellular, brain pathway or circuit, behavioral, or self-report research strategies. Unlike diagnosis-driven research that focuses on assumed pathologies, the RDoC program emphasizes our understanding of routine, everyday processes. Researchers assume that processes at each level of analysis reflect a continuum of variation, and some of these processes are more likely to be involved in clinically significant patterns that disrupt social, academic, or occupational aspects of life.
Intersection with behavior science
Shepherd noted, “Nothing in the brain makes sense except in the light of behavior” (2015). The RDoC assigns behavioral research a strong organizing role, identifying and characterizing reliable behavior-environment relationships that serve as the context in which biological processes are investigated.
Implications of the RDoC for clinical practice
One of the aims of the program is to support the identification of reliable mechanisms, components, and processes that are involved in behavioral variability, contributing to cases we identify as intervenable. The hope is that future behavioral and/or physiological interventions will address these mechanisms or processes and, as they can pinpoint specifics, will be more effective than current approaches to healthcare.
In the case of the RDoC and the other two-dimensional systems that follow below, we need to rethink reimbursement practices. Third-party payers have historically demanded a categorical determination of intervenable cases (i.e., clients must meet diagnostic criteria for services to be covered). Thinking about healthcare access when cases are described dimensionally will require changes in healthcare gatekeeping and funding models.
- The Hierarchical Taxonomy of Psychopathology (HiTOP) Program
Some scientists are developing a new categorization system with psychometrics. Their focus is on latent constructs that are inferred from patterns (think of “vulnerability,” for example). These patterns are investigated at hierarchical micro and macro levels using quantitative analyses of self-report data. The goal is to identify co-occurring or clustering patterns that might suggest novel points of intervention. Similar to the RDoC, the HiTOP Program assumes that behavioral health is dimensional and spans a range of variability.
Intersection with behavioral science
The hierarchical organization via HiTOP categories does not force clinical cases (e.g., being fearful, apprehensive) into a diagnosis. Instead, the HiTOP Program asks, “what patterns do these cases follow?” Researchers can then use the HiTOP clusters to recruit their participants using a level of specificity that fits their research question. Questions can then address any level of specificity, from the macro level (e.g., general distress) to the situational micro level (e.g., fear of spiders).
Implications clinical practice
Because the HiTOP program relies on existing questionnaires and other quantifiable measures, these tools are available for identifying individuals who could benefit from resources or therapeutic supports. However, the HiTOP Program faces challenges. The HiTOP Program does not have a clear way of accounting for conditions that contribute to social, academic, or occupational difficulties. Researchers have examined how life experiences (e.g., significant loss, poverty) can increase risk for difficulties, but the model does not distinguish contributions from life events from assumed psychopathology. For example, a person who is a member of a community that recently or persistently has been subjected to violence may obtain high scores on a measure of “paranoid” thoughts; however, the sense of danger and the accompanying vigilance can be understood as reasonable given their context. Additionally, most measures that inform the HiTOP approach have been designed and normed for verbally fluent individuals, mostly adults. The same breadth of measures is not available for individuals who require proxy reporters.
Network approaches represent novel quantitative strategies that characterize behavior and its context. In a network analysis, the observed correlations among variables (e.g., client difficulties, life events) reflect elements of a causal system that interact with one another. Within this framework the network structure reflects the problem and not some hypothesized underlying cause or structure.
The network analysis family consists of different temporal approaches. Using measures taken at a single time point, researchers and clinicians examine relationships among selected measures from a “snapshot-within-subject” perspective and also ask from a “between-subjects” stance whether multiple subgroups within the dataset emerge (i.e., studying strengths of relationships among the measures). It is thus a bottom-up approach that holds promise for identifying relationships when variations would be lost in (and contribute noise to) the overall sample’s data trends. Network analyses can also be used with longitudinal/time series data, including behavioral assessment data collected two or more times daily. These analyses describe how changes in one or more variables are accompanied by changes in other variables. The logic of this type of analysis is the same as the logic of functional analysis (time series-within-subject). Complex between-subject analyses are also used to identify subgroups of individuals where the measures indicate shared causal relationships.
Because of the current characteristics of statistical models, the network approach has difficulty capturing the contributions of stable contextual variables that occur over multiple time points. If there is no variation in the measures of context (e.g., social oppression is persistent), then it is difficult to quantify the influence of these variables.
Intersection with behavior science
Network analyses can include behavior monitoring data and support analysis of meaningful relationships among variables at the individual level. As noted above, it can identify the presence of subgroups within a dataset, allowing for a more nuanced exploration of individual differences. This approach aims to improve our ability to characterize causal systems that reflect psychological distress and when these are better understood, interventions can be better tailored to individual needs.
Implications for clinical practice
Researchers are currently exploring the use of network analyses to inform process-based behavior therapy (see Moskow et al., 2023). This is a work in progress and the program of research will take some time to mature to determine the types of variables and measurement practices that are most informative to treatment selection and clinical progress monitoring. Similar to HiTOP, the lack of clinical cutoffs within the approach means it does not fit with current third-party billing models that require a categorical diagnosis for determination of service eligibility and payment.
“It’s been a long time coming, but I know a change is gonna come.” (Sam Cooke – “Ain’t that good news!”)
Despite the well-known limitations of the DSM and ICD, the three efforts described above are not ready to replace these tools for clinical diagnosis and healthcare gatekeeping. In fact, substantial changes in existing power systems in healthcare will be required to overcome the existing inertia of current structures and practices. Our next blog will discuss navigating mental health treatment with and without the use of diagnostic labels.
References & Resources
Belluck, P., & Carey, B. (2013, May 6). Psychiatry’s guide is out of touch with science, experts say. The New York Times. https://www.nytimes.com/2013/05/07/health/psychiatrys-new-guide-falls-short-experts-say.html?smid=url-share
Insel, T. (2013, April 29). Transforming Diagnosis. NIMH Director’s Blog. https://web.archive.org/web/20130724080927/http://www.nimh.nih.gov/about/director/2013/transforming-diagnosis.shtml
Kotov, R., Krueger, R. F., Watson, D., Cicero, D. C., Conway, C. C., DeYoung, C. G., Eaton, R. R., Forbes, M. K., Hallquist, M. N., Latzman, R. D., Mullins-Sweatt, S. N., Ruggero, C. J., Simms, L. J., Waldman, I. D., Waszczuk, M. A., & Wright, A. G. (2021). The Hierarchical Taxonomy of Psychopathology (HiTOP): A quantitative nosology based on consensus of evidence. Annual Review of Clinical Psychology, 17, 83-108. https://doi.org/10.1146/annurev-clinpsy-081219-093304
Moskow, D. M., Ong, C. W., Hayes, S. C., & Hofmann, S. G. (2023). Process-based therapy: A personalized approach to treatment. Journal of Experimental Psychopathology, 14, Article 20438087231152848. https://doi.org/10.1177/20438087231152848
Shepherd, G.M. Neuroenology: how the brain creates the taste of wine. Flavour 4, 19 (2015). https://doi.org/10.1186/s13411-014-0030-9