As noted, WHO member states are responsible for reporting health statistics based on ICD to WHO (e.g., causes of mortality and morbidity, prevalence). An example of how these national health statistics are used is the ongoing Global Burden of Disease (GBD) study. The most recent wave of this study showed that mental and behavioral disorders account for 6.6% of total global disease burden as measured by disability-adjusted life years (GBD 2015 DALYs and HALE Collaborators, 2016), which represent a combination of prevalence, premature mortality, and disability. Vigo, Thornicroft, and Atun (2016) argued that the 6.6% total is an underestimate because the study attributes many outcomes of mental and behavioral disorders to other causes, and that the true percentage is closer to 13.0%. The GBD study also showed that mental and behavioral disorders account for an even larger percentage (18.9%) of global disability and that the top 20 causes of global disability include anxiety disorders, schizophrenia, autism spectrum disorder (ASD), and substance use disorders, as well as other mental and behavioral disorders (GBD 2015 Disease and Injury Incidence and Prevalence Collaborators, 2016).

Over the past several decades, governments of WHO member states have increasingly integrated ICD into clinical processes and policies related to health care coverage and reimbursement, social services, and disability benefits (International Advisory Group for the Revision of ICD-10 Mental and Behavioural Disorders [IAG], 2011). In most countries, provision of medical services other than routine examinations or preventive services is contingent on a qualifying diagnosis. Thus, diagnoses are used to facilitate access to appropriate health care, and the lack of a qualifying ICD diagnosis is used (often appropriately, but sometimes inappropriately) to deny services. This integration of ICD with clinical processes has increased as a result of more intensive administrative management of health services, the development and integration of electronic-information infrastructure for health care (e.g., electronic health records), automated treatment algorithms and care pathways, and increased implementation of standards of care and “evidence-based clinical guidelines,” most of which are based on diagnostic categories (e.g., National Institute for Health and Care Excellence, U.K., 2015). The match between diagnosis and services received is also used to evaluate treatment delivery and outcomes. Likewise, national and private health-insurance policies often use ICD (again, in both appropriate and inappropriate ways) to define the scope of practice and reimbursement policies for specific groups of health professionals.

These applications of ICD in health care systems are governed by national and local laws and regulations as well as by health-system policies and are not directly determined by WHO. However, many countries, including the United States, legally require the use of ICD (or national modifications of it) for the collection and reporting of diagnostic information as a part of health encounters. Country-level modifications usually include additional codes for hospital records, but they may also involve the addition or, less commonly, the deletion of diagnostic categories or subtypes. Changes in ICD, therefore, have major implications for national and health-system policies and processes and may require legislative or regulatory changes. As a result, WHO member states view themselves as major stakeholders in ICD revisions.

The U.S. government has published ICD adaptations as frameworks for government morbidity and mortality statistics beginning with ICD-7. The National Center for Health Statistics and the Centers for Medicare and Medicaid Services are currently responsible for developing and maintaining the adaptations, which contain additional information to facilitate indexing by hospitals and to enable the coding of morbidity and utilization data from patient records. These U.S. adaptations of ICD, called “Clinical Modifications” (CMs), initially had little direct relevance to U.S. health professionals. However, they became substantially more important when federal regulations required health professionals to use ICD-CM codes on Medicare claims and, after the passage of the Health Insurance Portability and Accountability Act (HIPAA) in 1996, on all electronic transactions for billing and reimbursement (U.S. Centers for Medicare and Medicaid Services, 2015). Effective October 1, 2015, 25 years after the approval of ICD-10 by the World Health Assembly, the U.S. government began requiring U.S. health professionals and health systems to use a new CM, ICD-10–CM (U.S. National Center for Health Statistics, 2015). The overlapping time frames of DSM-5, the U.S. ICD-10–CM, and WHO’s forthcoming ICD-11 have created substantial confusion among U.S. health professionals.

An obvious question is why the U.S. would require this expensive and time-consuming transition to ICD-10 just before the completion of ICD-11. Not surprisingly, the answer is complex. The ICD-10–CM had been under development for years and was intended for substantially earlier implementation, but objections from health systems, health insurers, and professional associations (e.g., the American Medical Association) regarding expense and burdensome administrative requirements delayed implementation to the point that many suggested that the United States skip implementation of ICD-10 and move directly to ICD-11 (Natale, 2014; Sullivan, 2012). However, the U.S. Department of Health and Human Services argued that, first, the ICD-10–CM implementation was urgently needed to correct many ICD-9–CM deficiencies, and second, the development of a U.S. ICD-11-CM would require several additional years after the ICD-11’s approval by the World Health Assembly and its initial release, which is expected in 2018. Congress eventually passed legislation requiring ICD-10-CM implementation.

When the United States will move to ICD-11 remains to be seen. On the one hand, those who resisted ICD-10–CM almost certainly will resist a subsequent change. On the other hand, implementing ICD-11 after its completion and approval would be consistent with the United States’ responsibilities as a WHO member state and with the country’s leadership role in many WHO activities, including key aspects of ICD-11’s development. The U.S. government has already modified ICD-10–CM to make its categories and terminology more compatible with their DSM-5 counterparts. Some of these changes effectively move ICD-10-CM closer to ICD-11. However, a piecemeal process may result in a patchwork classification system that is far from ideal, as demonstrated in past efforts to update ICD-9-CM to accommodate the categories and terms of DSM-IV, so it would be better if a full update were made sooner rather than later.

Thus far, we have introduced the major concepts that we address in this article and provided a historical context for the remainder of the article by reviewing the development of ICD and DSM. We now turn to describing the current (or, in the case of ICD, impending) instantiations of the three descriptive systems of psychopathology that are our primary focus.

The priorities of WHO’s Department of Mental Health and Substance Abuse in developing the ICD-11 classification of mental and behavioral disorders are rooted in WHO’s public-health mission and objectives. The department has long been concerned about an unacceptable global “mental health gap” (WHO, 2008). Despite the major contribution of mental disorders to global disease burden and disability, treatment for them remains unavailable or woefully inadequate in most of the world. The 2004 World Mental Health Survey found that in low- and middle-income countries, less than 25% of individuals with serious mental illness (e.g., schizophrenia, bipolar disorder, MDD) had received any treatment in the past year, and even in high-income countries (e.g., the U.S., Western European countries), the proportion was only a third to a half (World Health Organization World Mental Health Survey Consortium, 2004). In that same year, Kohn, Saxena, Levav, and Saraceno (2004) estimated that the global treatment gap between those who needed treatment for mental disorder and those who received it ranged from 32% to 78%, depending on the disorder.

Moreover, it is well established that people with serious mental illness have a much higher prevalence of nonmental health conditions (e.g., cardiovascular, metabolic, and respiratory diseases; De Hert, Correll, et al., 2011; Svendsen, Singer, Foti, & Mauer, 2006), which contributes to disproportionately higher rates of mortality and a substantially reduced life expectancy (Cuijpers & Smit, 2002; Thornicroft, 2011, 2013; Walker, McGee, & Druss, 2015). There may be direct relations between mental disorders and specific health conditions, such as cardiovascular disease (e.g., Davidson, 2012; Scott et al., 2013). However, paradoxically, a possible indirect relation is that traditional institutional treatment for mental disorders may aggravate modifiable risk factors such as smoking and obesity (Shin et al., 2012), which are the greatest contributors to poor physical health and excess mortality among people with mental disorders (e.g., Cerimele & Katon, 2013; de Leon & Diaz, 2005; Scott & Happell, 2011). In addition, common pharmacological treatments for mental disorders may have side effects that are damaging to overall health (Correll, 2007).

Major contributors to the mental-health treatment gap are specific patient, provider, treatment, and system factors that act as barriers to the recognition and management of physical diseases in people with mental disorder (De Hert, Cohen, et al., 2011). System factors include deficient resources in mental-health-care systems, issues of stigma, inadequate prevention programming, and lack of parity in health financing, including through insurance coverage (Saxena, Thornicroft, Knapp, & Whiteford, 2007). Not surprisingly, the greatest health disparities are found among underserved population groups, such as African Americans in the United States, who face increased morbidity as a result of their minority status when they experience mental disorder (Carliner et al., 2014; Das-Munchi et al., 2016; Mookhoek, deVries, Hovens, Bouwers, & Loone, 2001; Stecker, Fortney, Steffick, & Prajapati, 2006; Voruganti et al., 2007). Thornicroft (2011) has suggested that physical-health disparities among people with mental disorder contravene international conventions regarding the “right to health.”

Another contributing factor to the mental-health treatment gap is the absence of specialized expertise and qualified mental-health service providers in most parts of the world. Data from WHO’s Mental Health Atlas (WHO, 2011) indicate that high-income countries are reasonably well supplied with mental-health professionals, with nearly 44 for every 100,000 people. But less than 20% of the world’s population lives in these countries (World Bank, 2015). In upper-middle-income countries (e.g., Brazil, China, Mexico), which represent about 33% of the world’s population, there are fewer than one-third as many mental health professionals—less than 14 for every 100,000 people. Lower-middle-income countries (e.g., India, Indonesia, Nigeria), which represent about 39% of the world’s population, have fewer than four mental-health professionals for every 100,000 people and, on average, nearly three of these are nurses. In low-income countries including much of sub-Saharan Africa, which represent 9% of the world’s population, mental-health professionals are essentially nonexistent. Simply put, these statistics mean that most people with mental disorder in the world are unlikely to see a mental-health professional at any time in their lives.

Clearly, the current status of health care provided to people with mental disorder around the world is an urgent public-health priority. To begin to address this situation, the World Health Assembly approved the global “Mental Health Action Plan 2013–2020” (WHO, 2013a). The action plan and accompanying resolution represented a formal recognition by WHO member states of the importance of mental health and committed them to specific and measurable actions to improve mental health, including increasing service coverage for people with severe mental illness by 20%. The plan further moved away from “a wholly medical model to address income generation and education opportunities, housing and social services, and other social determinants of mental health” (Saxena, Funk, & Chisholm, 2013, p. 1971).

The ICD-11 classification of mental and behavioral disorders serves as a primary means for identifying individuals in need of services. As such, it is integrally connected to the WHO’s “Mental Health Action Plan,” which supports WHO’s Department of Mental Health and Substance Abuse’s overarching goal for revising the classification: that ICD-11 provide a better tool for helping clinicians to identify not only the people in need of mental-health services but also the treatments and management strategies most likely to be effective, with the ultimate goal of helping WHO member states reduce the disease burden associated with mental disorder.

Another WHO product that is integrally related to the action plan is the Mental Health Gap Action Programme’s “mhGAP Intervention Guide for Mental, Neurological, and Substance Use Disorders in Non-Specialized Health Settings,” now in its second edition (WHO, 2016a). This document includes evidence-based treatment guidelines designed to be implemented in primary-care settings in low- and middle-income countries for “priority” mental disorders, which are defined on the basis of disease burden, prevalence, and the extent to which evidence-based treatments can be implemented in such settings. Designed to be consistent with the ICD-11, the intervention guide provides guidance for a range of conditions including depression, psychoses, child and adolescent mental and behavioral disorders, dementia, disorders due to substance use, and self-harm and suicide, as well as guidance on more broadly targeted psychosocial interventions. WHO has also developed similar guidelines for conditions specifically related to stress (e.g., posttraumatic stress disorder, or PTSD; WHO, 2013b).

The public-health aims described above have substantially influenced the methods and priorities of the developers of ICD-11. To be a better tool for reducing the disease burden of mental disorders, the new system will need to be useful and usable throughout the world at the points where people with mental-health needs are most likely to come into contact with opportunities for care. For this reason, WHO is particularly focused on issues of clinical utility and global applicability in developing the revision (Reed et al., 2013).

Since the release of DSM-III in 1980, several problems with its diagnostic system (which for the most part also apply to the ICD classification; Sturt, 1981) had become evident. It also gradually became clear that the rapid and widespread adoption of DSM-III and its successors as the standard for funding federal grants, drug trials, and reimbursement for mental-health services had had several negative effects. One effect was reification of DSM categories and criteria as a direct reflection of the true nature of psychopathology, and a second was their almost exclusive use in research evaluating treatment efficacy, which we discuss further below. The rapid pace of mental-health research in the 1990s and increasing concerns about correcting the DSM’s perceived deficiencies led to a meeting of the leadership of APA and NIMH in 1999 to discuss the possible next steps for DSM. This meeting set the DSM-5 development process into motion.

Epidemiologic studies, starting with the Epidemiologic Catchment Area program (Boyd et al., 1984) and, later, the National Comorbidity Survey (Kessler, 1994) and its replication (Kessler & Merikangas, 2004), revealed high rates of comorbidity among DSM disorders, especially when diagnostic-hierarchy requirements (in which Disorder X could not be diagnosed if it was “due to” Diagnosis Y) were ignored. Some comorbidity patterns were expected on the basis of presumed etiologic mechanisms and clinical observation; for example, various anxiety disorders were found to be comorbid with each other, perhaps as the result of a hypothesized common underlying dimension of sensitivity and hyperreactivity to threat stimuli (e.g., T. A. Brown & Barlow, 1992). However, much of the comorbidity was unexpected or nonspecific—for example, 56% of National Comorbidity Survey respondents with at least one mental disorder had two or more (Kessler et al., 1994), and anxiety disorders were comorbid not only with each other but also with various depressive (L. A. Clark & Watson, 1991) and alcohol-use disorders (Kushner, Sher, & Beitman, 1990). Although diagnostic hierarchies were largely discontinued in DSM-III-R, concerns about the validity of DSM’s categorical diagnostic boundaries persisted for other reasons.

One particular issue is the specificity with which diagnoses direct treatments—a major rationale for any diagnostic system. For example, emerging research evidence strongly suggested that the targets of psychopharmacologic treatments did not fit neatly into specific DSM categories. New antidepressants, such as selective serotonin reuptake inhibitors (e.g., fluoxetine, better known as Prozac) and second-generation antipsychotic medications, were introduced in the 1980s and 1990s. As these therapies were subjected to further investigation, results showed that “antidepressants” were effective in the treatment of many disorders beyond depression, such as eating disorders and anxiety disorders. Likewise, several of the second-generation antipsychotic medications were found to be effective alone or as adjunctive treatments for nonpsychotic mood disorders.

Concurrently, psychotherapy research led to the introduction of “manualized” treatments—in particular, cognitive-behavioral and interpersonal therapies—that were standardized and could be disseminated for clinical and research purposes. Cognitive-behavioral therapy, originally developed to treat depression, has since been used successfully in a very wide range of mental disorders, including anxiety, obsessive-compulsive disorder, eating disorders, PDs, and substance use disorders (e.g., see https://www.div12.org/psychological-treatments/treatments/). In sum, the nonspecificity of both pharmacological and psychosocial treatments called into question the specificity of the DSM disorders and their purported underlying mechanisms. Major questions were raised about the boundary between MDD and generalized anxiety disorder (GAD), and those among schizophrenia, bipolar disorder, MDD, and schizoaffective disorder, as well as between and among other pairs and groups of often co-occurring disorders (e.g., L. A. Clark, Watson, & Reynolds, 1995).

Similarly, another negative effect of the rapid and widespread adoption of the diagnostic categories of DSM-III and its successors was their almost exclusive use in research evaluating treatment efficacy, with the highest level of evidence seen as that coming from randomized controlled trials testing the effects of specific treatments in patient populations defined by specific DSM categories. Initially, DSM-III was heralded as providing an empirical basis for mental-disorder treatments. However, as evidence of the nonspecificity of both pharmacological and psychosocial treatments accrued, the nearly exclusive reliance on DSM diagnoses in treatment research was criticized for limiting the ability of such research to cut across category boundaries (Cuthbert & Insel, 2013; Hyman, 2010). It also systematically disadvantaged treatments with broad effects by requiring repetitive demonstrations of their efficacy in specific, narrowly defined patient populations, despite the fact that most treatments’ mechanisms of action are likely the same across many disorders (Tucker & Reed, 2008).

The requirement of repetitive demonstrations presented a particular barrier for psychological treatments (e.g., cognitive-behavioral therapy) as opposed to pharmacological ones, given that psychological treatments generally lack the financial backing of organized commercial interests that would support treatment trials with large samples (Tanenbaum, 2005). Compared with trials of psychological treatments, large-scale drug trials can be completed in relatively short time frames, because any licensed physician can prescribe a medication and “treatment sessions” can be brief and intermittent. Thus, the labor costs of pharmacological treatment trials are relatively low, and such studies are readily funded by pharmaceutical companies, which stand to profit considerably if a trial supports a drug’s efficacy. In contrast, training clinicians to administer a psychological treatment with a high level of reliability is time-consuming, and treatment sessions typically require a commitment of an hour or two weekly, so the labor costs are high and there are few to no financial incentives for companies to fund such studies.

To the extent that people tend to believe that response to a pharmacological agent means that a disorder has a biological cause, this situation also contributed to a self-perpetuating emphasis on biological causality. Although it is reasonable to assume that a complete and accurate understanding of biological causality will be important to developing more effective treatment strategies for mental and behavioral disorders, an overemphasis on biological causality runs counter to increasing evidence regarding the multicausal nature of mental illness.

The growth in epidemiological studies worldwide since 1994 (Kessler, 1999) also revealed substantially more cross-national epidemiological variation in DSM-IV disorder prevalence than expected. For example, social anxiety disorder showed a 34-fold variation in prevalence between the country with the highest prevalence and the country with the lowest prevalence despite use of the same instrument (the Composite International Diagnostic Interview for DSM-IV) and the same time frame of measurement (12 months; for other examples, see Hinton & Lewis-Fernández, 2011, and Lewis-Fernández et al., 2010). These findings raised the question of whether the diagnostic criteria underlying the epidemiological instruments were overly specific and thus could fail to identify different cultural expressions of the same disorders (Lewis-Fernández et al., 2010). It is always possible, of course, that the prevalence of particular disorders is highly variable or that lack of familiarity with the disorders resulting from lower access to mental health care and information in some settings leads to lower endorsement of symptoms. The APA’s DSM-5 Task Force opted to revise disorder criteria if the cross-cultural data warranted it. For example, one criterion for the diagnosis of social anxiety disorder, fear of negative evaluation, was broadened to include alternative presentations that are typical in Asia. A separate section, “Culture-Related Diagnostic Issues,” was added to most DSM-5 disorders to facilitate cross-cultural application of the nosology.

Some questions about the underlying mechanisms of DSM disorders were prompted by the inability to find laboratory markers or tests for the vast majority of disorders. A few initially promising tests were found subsequently to be of little or no practical value for clinical practice. Currently, narcolepsy is the only mental condition that has a biomarker (hypocretin deficiency) as a criterion, although robust biomarkers for various causes of neurocognitive disorder have been found (e.g., Alzheimer’s disease) or are likely to be found in the relatively near future (e.g., prion disease) and may be added as criteria in future versions of the DSM.

Generally, the search for indicators of the construct validity of diagnoses as outlined in E. Robins and Guze’s (1970) classic article has been frustrating. Initially promising findings, such as a gene mutation associated with DSM-III bipolar disorder (Egeland et al., 1987), were later disproved (Ginns et al., 1992), deflating the optimism that we might soon find “the gene for” various mental illnesses (Kendler, 2005, p. 1243). Endophenotypes—measurable, typically biological, characteristics “along the pathway” between genes and clinical disorders—were the next object of enthusiasm (e.g., Gottesman & Gould, 2003, p. 636), as it was hoped that they would “more clearly reflect the impact of specific genes” (Patrick, 2014, p. 1333). However, endophenotypes are proving to be no simpler genetically than the clinical phenotypes with which they are associated (Iacono, Vaidyanathan, Vrieze, & Malone, 2014).

Nonetheless, the processes of seeking specific genetic influences and examining why early findings were disproved have led to more sophisticated research designs, technologies, and statistical methods. These improved research methods, in turn, have led to findings that reveal significant genetic overlap among many DSM-IV diagnoses and may “provide valuable insights into the psychological and neural mechanisms” that give rise to psychopathology (Iacono et al., 2014, p. 1339)—a hope that is the foundation of NIMH’s RDoC project. For example, researchers recently found significant variation in a particular cluster of genes that they were able to link both to a specific immune-system function—synaptic pruning—and to schizophrenia (Sekar et al., 2016). Related findings in the study showed that variation in the function’s expression affected mouse brains in a way that was consistent with observed characteristics in schizophrenia, indicating a potential neural mechanism contributing to the disorder’s development.

The limitations of DSM-IV diagnostic categories were also seen in clinical use. One of the most striking manifestations was the high rate of nonspecific diagnoses used in general clinical practice, expressed as disorders “not otherwise specified” (NOS). Certain NOS diagnoses were used particularly frequently, including NOS variants of PD (Verheul & Widiger, 2004), eating disorder (Smink, van Hoeken, & Hoek, 2013), dissociative disorder (Johnson, Cohen, Kasen, & Brook, 2006), and pervasive developmental disorder (Lauritsen, Mortensen, & Pedersen, 2004 The use of NOS diagnoses was acceptable for cases in which (a) a named syndrome was not classified in the DSM (e.g., a diagnosis of eating disorder NOS for cases of binge eating disorder); (b) a broad diagnosis could be made, but diagnostic criteria were not met for any specific diagnosis (e.g., PD-NOS); or (c) insufficient information was available for a specific diagnosis (e.g., a diagnosis of psychotic disorder NOS as a placeholder until general medical and substance-related causes could be ruled out). Although such usage was permitted, the high use of NOS categories indicated problems in diagnostic coverage (e.g., should binge eating disorder be added to the classification?) or criteria (e.g., is there a more useful way to describe the PDs and pervasive developmental disorders seen in clinical practice?).

Clinicians were also given limited ability to document clinically significant symptoms that occurred outside the criteria of the patient’s primary diagnoses. For example, anxiety symptoms are frequently seen in patients with MDD and are associated with poorer outcomes (Walker & Druss, 2015). However, there was no way to acknowledge these symptoms if they did not meet criteria for a specified anxiety disorder, despite their clear clinical effect. Likewise, sleep difficulty is a frequent experience of persons with a wide range of mental disorders and often is treated, despite being a symptom of only a few diagnoses (e.g., GAD, bipolar disorder, and MDD) outside sleep disorders themselves. The limited availability of ways to document these commonly associated symptoms hindered the overall ability of practitioners using DSM-IV to describe the diversity of patient presentations, to justify treatment decisions, and to follow important treatment outcomes.

The overall consensus of the DSM-5 developers was that despite acknowledged advantages, the specificity of DSM diagnostic criteria was starting to hinder progress in the search for underlying mechanisms of mental disorders, and the criteria were not describing the clinical realities experienced by patients and treated by clinicians. Initially, it was thought that knowledge generated since the development of DSM-IV might be useful in improving the validity of DSM disorders—particularly, that findings in genetic epidemiology, molecular genetics, and functional and structural imaging might help the manual move toward a more etiologically based classification. Accordingly, dimensional models of diagnosis, a topic of interest that was abandoned as premature for DSM-IV development, was resurrected in early DSM-5 discussions (Kupfer, First, & Regier, 2002), and DSM-5 development started with commitments to examine research evidence from multiple scientific fields and an openness to complementary and alternative classifications systems (e.g., the Spectrum Project; Maser & Patterson, 2002).

The revision of DSM-IV was undertaken with several guiding principles. First, as with ICD-11, the highest priority in modifying DSM-IV was to optimize clinical utility. Second, revisions were to be guided by research evidence. A 5-year, NIH-funded series of research-planning conferences was held before the appointment of the DSM-5 Task Force to synthesize research evidence in important areas and to identify gaps in research. Third, it was felt that continuity with DSM-IV should be maintained in DSM-5 to prevent upheaval in the field, although there were no a priori constraints on the degree of change that could be made between DSM-IV and DSM-5 if the supporting research evidence was strong.

After the planning conferences, the DSM-5 Task Force, work groups, and cross-cutting study groups met for approximately 5 years. Proposals for changes were reviewed for the strength of their supporting research evidence and, when indicated, for their clinical, public health, and forensic implications. They also were reviewed by the APA Assembly and, finally, by the APA Board of Trustees, the two governing bodies of the APA. The Assembly’s members are selected by geographically based psychiatric societies across the United States, its territories, and Canada, with additional members representing various constituencies (e.g., early-career psychiatrists, underrepresented minority groups); approved allied organizations also sit on the Assembly. The group acts in an advisory capacity to the Board of Trustees, representing the needs of its constituents. The Board of Trustees is composed of officers elected by the APA general membership, representatives elected by members in geographic areas, a limited number of past presidents, and additional members representing the Assembly and various other constituencies. The Board of Trustees is responsible for determining APA’s priorities, policies, and budgets. The leadership and members of the APA Assembly and Board of Trustees tend to reflect the diverse membership of the organization and include researchers, academicians, and full-time clinicians in a range of practice settings. However, participation of the general membership in electing officers has been less than representative (e.g., fewer than 15% of members cast a vote for President-Elect of the Board of Trustees in 2016).

In the end, this process resulted in a revised manual that was considerably more conservative than had been envisioned more than 10 years earlier. The consensus from the early research-conference series was that, in general, current neuroscience findings would not be translatable to clinically useful diagnostic criteria, a conclusion that shut down hopes for a “paradigm shift” toward more etiologically based diagnostic criteria. Dimensional assessment measures of symptom and diagnostic severity and of disability were recommended by the DSM-5 Task Force for clinical use but were rejected by the Board of Trustees because of a lack of evidence on their utility in improving patient care and outcomes. The DSM-5 Personality and PD Work Group’s proposal of a radically different method to diagnose PDs, a hybrid categorical-dimensional model, was also rejected, largely because the Board of Trustees was not convinced that its clinical utility had been sufficiently well established despite many iterations in its development to maximize usability and its acceptable performance in the DSM-5 field trials.

In the end, the changes in DSM-5 were relatively conservative. Nonetheless, many evidence-based innovations were introduced that were expected to lay the groundwork for future diagnostic improvements. We discuss those involving reorganization, introduction of the spectrum concept, elimination of the multiaxial system of diagnosis, and emphasis on developmental, gender-related, and cultural aspects of mental disorder.

The WHO Disability Assessment Schedule 2.0 was recommended by the DSM-5 Task Force as an alternative to the Global Assessment of Functioning, but it was not approved by the APA Board of Trustees, pending further data on its utility in improving patient care and outcomes. Likewise, the APA Board of Trustees did not approve the many patient-reported dimensional assessment measures that the DSM-5 Task Force recommended for clinical use. Specifically, a number of cross-cutting symptom measures with follow-up questions to probe domains that screening identified as relevant were compiled to provide a comprehensive overview of commonly seen psychiatric symptoms that could guide psychiatric assessment and treatment, and highlight potentially overlooked problems. For many disorders, diagnostic severity measures that focus on syndromes and were designed primarily for administration after a diagnosis had been made also were recommended. Both of these types of measures are discussed in a later section. It is notable that DSM-5 does not include a quality-of-life measure, which may be considered an oversight due to the relatively low level of patient involvement in the overall DSM-5 development process. Future testing of the DSM-5’s assessment measures should involve patients and their families to ensure that the measures are comprehensive and include domains that are important to them and not just to clinicians. If the published measures are found to be lacking in this regard, they should be supplemented.

Another assessment measure in Section III is the Cultural Formulation Interview (CFI; Lewis-Fernández, Aggarwal, Hinton, Hinton, & Kirmayer, 2016). The CFI converts the DSM-IV Outline for Cultural Formulation (OCF) into a set of standard questions and explicit instructions that can be used easily by busy clinicians. The main goal of the OCF in DSM-IV was to help clinicians identify how culture and context affect patients in ways that are relevant to diagnosis and treatment (Mezzich, Caracci, Fabrega, & Kirmayer, 2009). The CFI operationalizes the OCF framework as three sets of semistructured questionnaires: a core 16-item interview that can be used during a routine initial evaluation with any patient by any clinician in any setting; the CFI–Informant version of the assessment, which is used to gather information from close associates of the patient (e.g., family members); and 12 supplementary modules that expand on these basic assessments if a comprehensive cultural assessment is needed (these supplementary modules are available through the APA Web site at http://www.psychiatry.org/psychiatrists/practice/dsm/dsm-5/online-assessment-measures). The core interview aims to enhance the person-centeredness of an evaluation by grounding it from the outset in the patient’s own experience of illness (Lewis-Fernández & Aggarwal, 2013).

The CFI was included in DSM-5 because culture affects every aspect of clinical care, including patients’ and families’ concepts of illness, the patterning of symptoms, help-seeking choices and treatment expectations, and even the models that clinicians use to interpret and understand symptoms in terms of psychiatric diagnoses (Lewis-Fernández et al., 2016). All individuals, not only nondominant groups such as racial/ethnic minorities, are influenced by their cultural backgrounds and people draw upon different aspects of their backgrounds at different times. To emphasize these issues and help guide implementation of the CFI, “culture” was defined in the instructions in a way that would apply to all people—as “the values, orientations, knowledge, and practices that individuals derive from membership in diverse social groups,” “aspects of the person’s background that may affect his or her perspective,” and “the influence of family, friends, and other community members” (APA, 2013, p. 750). The CFI was included in Section III to help clinicians include basic information on the social context of individuals and their networks in diagnostic assessments, thereby helping to reduce misdiagnosis, to calibrate the assessment of severity, and to facilitate patients’ engagement in treatment (Adeponle, Thombs, Groleau, Jarvis, & Kirmayer, 2012; Aggarwal, Nicasio, DeSilva, Boiler, & Lewis-Fernández, 2015).

The APA Board of Trustees retained the DSM-IV system for personality diagnosis in the main Section II and placed the Alternative DSM-5 Model for Personality Disorders (AMPD) developed by the Personality and PD Work Group in Section III “to preserve continuity with current clinical practice, while also introducing a new approach that aims to address numerous shortcomings of the current approach” (APA, 2013, p. 761). The two main criteria of the AMPD are “impairments in personality functioning and pathological personality traits” (APA, 2013, p. 761; emphasis in original). The first criterion represents the core dysfunction of PD and consists of impairment in self-functioning (defined as problems in identity, self-direction, or both) or in interpersonal functioning (defined as problems in empathy, intimacy, or both). The second criterion describes the way in which personality impairment is expressed, operationalized as any one or more of 25 specific traits (e.g., submissiveness, manipulativeness) or the five broad domains in which they are organized: negative affectivity, detachment, antagonism, disinhibition, and psychoticism. Both personality functioning and traits are dimensional, and the model provides guidelines to help clinicians and researchers determine whether a given criterion is above threshold for impairment, in the case of functioning, or pathology, in the case of traits.

In addition, the AMPD describes six specific PD types (of the 10 in Section II, the main section of DSM-5)—antisocial, avoidant, borderline, narcissistic, obsessive-compulsive, and schizotypal—using particularized versions of the four subareas of personality impairment (i.e., identity, self-direction, empathy, and intimacy) and specific sets of traits. For example, the avoidant PD type is characterized by anxiousness and at least two of three other specific traits: withdrawal, anhedonia, and intimacy avoidance. These six PDs were chosen for inclusion because each had a considerable empirical and clinical literature, whereas the other four did not. Moreover, the remaining PDs (paranoid, schizoid, histrionic, and dependent) could be characterized very simply using one or two of the AMPD’s specific traits. For example, submissiveness and insecure attachment together form dependent PD.

The AMPD also includes a “PD-trait specified” diagnosis, which the text states can be used “when a personality disorder is considered present, but the criteria for a specific disorder are not” (APA, 2013, p. 761). It also has the potential to describe any combination of personality impairment and pathological traits (even when the criteria for a specific disorder are met), and preliminary evidence suggests that an expanded definition of “PD-trait specified” would provide both comparable coverage of the PD domain and greater specificity and flexibility in describing individuals’ personality pathology (L. A. Clark et al., 2015), as well as eliminate within-PD comorbidity by allowing for a single diagnosis, albeit one with myriad trait manifestations.

The AMPD is broadly congruent with what is proposed for PD diagnosis in ICD-11, as described further in a later section, “Categories and Dimensions.”

The final chapter in Section III contains criterion sets and supporting text for eight syndromes, disorders, or conditions that were proposed for DSM-5 but judged by the Task Force to have insufficient empirical evidence to warrant their inclusion in the main Section II (e.g., attenuated psychosis syndrome, internet gaming disorder, and nonsuicidal self-injury). They were included in Section III to facilitate research that will provide data for their consideration in a future DSM version.

Of the various criticisms of DSM-IV that were considered by DSM-5’s developers, virtually all apply to ICD-10 and are being considered in ICD-11’s development as well. ICD-11’s developers have had the advantage of being able to consider both the initial DSM-5 proposals and its final published version, as well as the evidence used to support changes from DSM-IV to DSM-5. As with the DSM-5, the development of ICD-11 has involved a model of working groups, but these groups were required to be globally representative and multidisciplinary and typically included one or more members from parallel DSM-5 work group to facilitate harmonization of the proposals. Therefore, in many instances, changes proposed for ICD-11 have been similar to those made in DSM-5.

The overall architecture of DSM-5 and the proposed “Mental, Behavioural, and Neurodevelopmental Disorders” chapter of ICD-11—that is, the broad groupings they contain and the order of those groupings—were the subject of a series of meetings between representatives of APA and WHO; the documents’ similarity is a tangible result of these harmonization efforts. For example, the new DSM-5 groupings of “Obsessive-Compulsive and Related Disorders” and “Disorders Specifically Associated with Stress” (in ICD-11) or “Trauma- and Stressor-Related Disorders” (in DSM-5) are part of a broader grouping in ICD-10 that includes anxiety disorders. Some proposed ICD-11 groupings have names slightly different from those of their DSM-5 counterparts, but they are conceptually similar. Further, the integration of forms of disorders that are seen as continuous across youth and adult psychopathology has been accomplished in both classifications. Diagnostic-hierarchy requirements were eliminated in a previous version of the DSM and have been eliminated in the proposed ICD-11, although in a few specific instances, disorders that should not be simultaneously diagnosed are explicitly noted. Conversely, ICD has never had a multiaxial system, so this change in DSM-5 made it more similar to ICD-11.

In some cases, the combination of substantive structural changes from ICD-10 to ICD-11 and related changes from DSM-IV to DSM-5 increased the similarity between the manuals. For example, the ICD-10 classification of sleep disorders reflects an inaccurate mind-body dualism: Ostensibly “nonorganic” sleep disorders were in the chapter on “Mental and Behavioural Disorders,” whereas ostensibly “organic” sleep disorders were included in such chapters as “Diseases of the Nervous System” or “Diseases of the Respiratory System.” In contrast, a new and separate chapter has been proposed for ICD-11 that includes all sleep-wake disorders. Likewise, sexual dysfunctions that were divided in ICD-10 into “nonorganic” and “organic” forms are combined in a new ICD-11 chapter called “Conditions Related to Sexual Health.” Parallel changes in DSM-5—the elimination of DSM-IV’s primary/secondary and organic/nonorganic distinctions in both of these disorder groups—were less obvious but effectively rendered the two manuals more similar.

ICD-10’s gender identity disorders have been substantially reformulated, renamed “gender incongruence” (Reed et al., 2016), and moved out of the mental disorders chapter.⁸ The ICD-11 Working Group deemed this a substantial improvement over the current classification of categories related to transgender identity as mental disorders. Impulse control disorders and disruptive behaviour and dissocial disorders are grouped in their own separate chapters in ICD-11 but are combined in DSM-5. PDs are placed earlier in the proposed ICD-11 classification structure than in DSM-5 (where they are one of the last chapters) to reflect their more developmental sequence. Finally, mental and behavioral syndromes due to particular types of diseases or disorders are organized differently in ICD-11 than in DSM-5 because of differences in the way that the two classifications handle primary and secondary disorders.

Several categories have been recommended for inclusion in the ICD-11 that were considered but not included in the DSM-5, such as olfactory reference disorder (Stein et al., 2016), complex PTSD, and prolonged grief disorder (Maercker et al., 2013). Conversely, DSM-5 contains several categories that have not been recommended for inclusion in ICD-11, such as disruptive mood dysregulation disorder (Evans et al., 2017). ICD-11 has retained ICD-10’s separate categories for harmful use of substances and substance dependence in order to reflect the public-health importance of harmful use, particularly for population-based prevention programs and early clinical interventions, such as in primary-care settings (Poznyak, Reed, & Clark, 2011).

The origins of ICD and DSM as public-health-focused classification systems account for some of the characteristics that have contributed to subsequent problems in classifying mental disorders. We discuss these problems more fully in our four main topic sections, but it is fitting to describe the characteristics themselves here, in the context of the systems’ descriptions. Specifically, a main characteristic of both ICD and DSM is that they are classic hierarchical categorical systems, such as Carl Linnaeus’s biological classification set forth almost 3 centuries ago.⁹ Hierarchical categorical classification systems are exhaustive, their elements are mutually exclusive, and lower levels of the structure are related systematically to specific higher levels. Every entity is distinct, must be classified in one and only one place in the system, and is more closely related to its near-neighbor entities than to those in more distant “branches.” The ICD and DSM systems are intended to have these qualities but, importantly, as applied to diseases, disorders, or health conditions—that is, unlike the biological classification of living beings, these systems categorize not people with disorders but rather the disorders that people may develop. (In ICD, these disorders include all health conditions, whereas DSM is confined to mental disorders.)

Several characteristics of both ICD and DSM that stem largely from these structural requirements have received criticism. First, disorders can be represented in one and only one place, but the basis for assigning certain disorders to particular places in either system is arbitrary and not fully consistent with all available evidence. For example, in both systems, major depression and GAD are classified in different places, even though they are highly comorbid and the genetic contribution to these disorders is essentially identical (Kendler, 1996, p. 68). Both DSM-5 and ICD-11 modified this structural requirement and cross-listed a few disorders so that the systems would better reflect the complexity of mental-disorder diagnosis. For example, DSM-5 placed schizotypal PD and antisocial PD with the other PDs; moreover, it also placed schizotypal PD with other schizophrenia-spectrum disorders and antisocial PD with its required childhood precursor, conduct disorder. Similarly, ICD-11 classified Tourette syndrome both in the chapter “Diseases of the Nervous System,” with other tic disorders, and with the “Obsessive-Compulsive and Related Disorders,” because of comorbidity and shared familiality.

Second, categorical systems are ideal when the entities being classified are taxa (i.e., unique, nonoverlapping kinds), which, for the most part, mental disorders clearly are not (Haslam, Holland, & Kuppens, 2012). Such systems can classify dimension-based entities only by imposing cut points, which may be relatively arbitrary. This is not particularly problematic in the case of “simple” health conditions that are based on one or two dimensions, such as obesity or blood pressure, because it is easy to understand, for example, that a person with a body mass index (BMI) of 29.9 will be classified as “overweight” and one with a BMI of 30.0 will be classified as “obese” for practical reasons, not because these two BMIs are fundamentally distinct. However, classifying nontaxonic, multidimensional phenomena such as mental disorders, which can overlap in myriad ways, is not straightforward.

Third, because each disorder must be classified in one and only one place, there is a tendency to make each entity as distinct as possible, resulting in highly—indeed, overly—specified categories. This issue is manifested in several ways, each of which has associated problems that are among the core criticisms of ICD and DSM. For example, major depression can present with marked anxiety, psychotic features, or both, so to keep these three disorder presentations distinct, a comprehensive categorical system must do one or more of the following:

1. Specify exclusion rules (e.g., exclude a diagnosis of GAD if a person meets the criteria for major depression); this choice comes at the expense of incompletely capturing the clinical picture of many individuals.
2. Create either specified intermediate categories (e.g., “mixed anxiety–depressive disorder” and “schizoaffective disorder”) or use disorder specifiers (e.g., “major depression with psychotic features”). This tactic is problematic because such disorders also have to be classified in one and only one place, but placement near one disorder imposes a distance from the other, violating the hierarchical organizational principle according to which entities should be closely related to their near neighbors but not to those in more distant “branches” of the system. For instance, schizoaffective disorder is placed with schizophrenia-spectrum disorders, so it cannot also be placed with mood disorders; conversely, major depression with psychotic features is placed with the mood disorders, so it cannot placed with other psychotic disorders. This approach also results in the proliferation of diagnostic categories, because every important overlap between dimensions then requires the creation of specific diagnostic label.
3. Define disorders’ criteria specifically but permit multiple comorbidities (i.e., allow a person to meet the criteria for major depression, an anxiety disorder, and a psychotic disorder). Resulting diagnoses may misleadingly suggest that the individual has three disorders rather than a single, complex mental disorder. We discuss these issues further in our sections on “Multidimensionality” and “Comorbidity.”
4. Define disorders overly specifically, with the result that many individuals’ conditions fall into “wastebasket” categories, such as “other depressive episodes” (ICD-10), “depressive disorder, unspecified” (ICD-10 and ICD-11), “depressive disorder, NOS” (DSM-IV), or “other specified depressive disorder” (DSM-5). Research literature reviews and meta-analyses have shown that in some domains of psychopathology, a greater percentage of individuals are more accurately characterized as having these types of diagnoses than meet the criteria for any specific disorder in the domain (e.g., Thomas, Vartanian, & Brownell, 2009, for eating disorder; Verheul & Widiger, 2004, for PD).

In addition to these various problems, an overarching unintended consequence of overspecification is that it contributes to the reification of disorders. That is, when disorders are highly specified, a natural assumption is that there is an empirical basis for the precision with which they are defined, so they come to be regarded as representing real and distinct natural phenomena. In turn, thinking about mental disorder in this way obscures the arbitrary nature of and lack of scientific evidence for many distinctions among categories or between disorder and health, as well as inhibits inquiry into their interrelationships and common substrates.

In summary, despite the many changes and improvements in ICD-11 and DSM-5 from their predecessors and the various differences between them, they both remain categorical classification systems that are fundamentally descriptive in nature, based primarily on self-reported symptoms, clinically observed signs, and a few tests (e.g., of intellectual functioning). In contrast, the perspective adopted by NIMH for its RDoC project, to which we turn next, diverges markedly from the conventions developed by ICD and DSM.

As discussed, the DSM-5 developers’ hope for a more neuroscience-based manual was quickly dashed by the facts that research findings indicated that biological factors not only cut across multiple disorders but also varied within disorders (Hyman, 2007) and there was insufficient research to support an alternative, empirically based organization. In no small part this was because, since the early 1980s, NIMH and other funding agencies had virtually mandated the use of DSM or ICD diagnostic categories. Thus, although the criterion-based approach of DSM-III and ICD-9 led to voluminous research and important scientific advances that ultimately made RDoC’s alternative approach possible, this very research also showed that the two classifications’ methods and overreliance on narrow, symptom-based categories ran counter to the study of mental disorder with full consideration of its multiple interacting etiologies (L. A. Clark & Watson, 1991; Hyman, 2010; Kapur, Phillips, & Insel, 2012; Owen, 2014; Sharma, Markon, & Clark, 2014).

The RDoC project was initiated in early 2009 in response to this changing landscape, including feedback received in late 2008 from researchers in the field as NIMH prepared its new strategic plan. An experimental approach was deemed necessary, given the still-nascent state of the science of mental disorder and the conceptual and empirical constraints of research based on current classifications. It was apparent that if developments in basic and translational science were to be applied to the science of mental disorder, a long-term approach would be needed. Such an approach would need to examine psychopathology with reference to behavioral and brain mechanisms rather than in terms of existing disorder categories. It was not that decades (or, indeed, centuries) of clinical observation and research should be discarded wholesale, but rather that an alternative research framework was needed to understand more fully the basic mechanisms that are disrupted in mental disorders.

The RDoC project thus was developed to respond to both of the difficulties that the groups revising DSM-5 encountered: the hitherto predominant reliance on the DSM and ICD diagnostic categories to guide research on the causes and mechanisms of disorder, and the lack of a clear alternative in the face of the failure of this research to support existing diagnostic systems. Accordingly, RDoC differs from the DSM and ICD in two important ways. First, it is based upon a fundamentally different approach to dimensionality, one more in line with DSM-5’s AMPD than with current diagnostic procedures; the bulk of this section elaborates RDoC’s dimensional approach. Second, the RDoC project was not intended for practical clinical use in the near future. Rather, it provides a framework for research. It does not formally incorporate any current ICD or DSM disorders and, in fact, does not define mental disorder or any specific disorders. It was thought that such a radical departure was needed to create a research framework that could address the problems discussed above in trying to understand, diagnose, and treat mental disorder.

It is important to contextualize the RDoC project not only within the “political” domain of institutional taxonomies but also in response to prior and ongoing intellectual advances, particularly those in developmental psychopathology. Many behavioral and psychological problems in children and adolescents reflect early stages in the development of later psychopathology, yet there is reluctance to label still-maturing individuals as having a mental disorder. Taken together, these points may explain why reification of mental disorder in younger age groups (as opposed to in adults) developed more slowly and less strongly. Accordingly, the work of developmental psychopathologists was a rich intellectual source for many aspects of the RDoC project. For example, as early as the mid-1960s, Achenbach (1966) had proposed that psychopathology has a hierarchical structure; moreover, such concepts as equifinality (the idea that diverse developmental pathways can lead to the same or similar outcomes) and multifinality (the idea that a given risk factor may result in different outcomes depending on the environment) had been used for several decades to understand the development of psychopathology as an outcome of person-environment interactions (e.g., Cicchetti & Rogosch, 1996; Sroufe & Rutter, 1984).

What is new about RDoC is the idea that to understand mental illness in all its complexity, the field needs a research framework that accommodates the study of all causal factors together. Although it is of course impossible to incorporate all factors in any given study, what the RDoC project hopes to do is break researchers away from studying a few relevant factors in isolation—that is, without consideration of how they fit together with all the other relevant factors. Similarly, given the work of developmental psychopathologists, by the time the RDoC project was being designed, almost all mental illness had come to be regarded as neurodevelopmental—as being affected by, and also affecting, neural structures and processes throughout development. The RDoC perspective, therefore, is that it is vital to incorporate a neurodevelopmental aspect in research programs, and yet the problems delineated above (e.g., overspecification and comorbidity) indicate that current disorder criteria miss or insufficiently reflect fundamental mechanisms that are important in the development and maintenance of mental illness. Until recently, it was difficult to investigate what these mechanisms might be. Now, new methods in neuroimaging, behavioral neuroscience, psychometric analyses of behavior, and for examining the effects of various events and social contexts upon behavior and the brain are transforming our understanding of normal and abnormal functioning.

These considerations provide important contexts for the RDoC project, which shares with DSM and ICD the quest for optimal classification—identifying the appropriate groupings and/or dimensions for study and, ultimately, for treatment. However, rather than continuing the research tradition of seeking deeper understanding of the mental disorders in our current classification systems, the RDoC project is intended to provide a research framework that will deepen our understanding of the neurological, biological, psychological, social, and cultural structures and processes that underlie mental illness broadly speaking. In turn, this information can guide the evolution of the classification of mental illness into more valid groupings and dimensions, much as Linnaeus’s two-kingdom system of plants and animals has evolved into a six-kingdom system. The hope is that this approach also will facilitate development of improved methods for preventing and treating mental illness. Rather than a wholesale overturn of current diagnostic systems, RDoC is intended to provide an alternative perspective that eventually will inform the field as to which aspects of current approaches should be retained or minimally modified, and which should be replaced or greatly revised. Importantly, the current RDoC scheme is considered “Version 1.0” and is explicitly intended to evolve as research progresses.

The RDoC framework is intended to facilitate the study of brain-behavior relations across developmental and environmental contexts. In recent decades, research has increasingly specified major functional aspects of behavior and mental operations—fear, executive functioning, and social attachment—and the neural systems that implement them. Given this trend, a research framework with a dual emphasis on brain and behavior seemed the most promising point of departure for an experimental program. However, there is no claim that the RDoC framework has “got it right.” Both the specific aspects of a neural-circuit-based framework and its overall organizational scheme are likely to change—perhaps dramatically—as research in this new direction accumulates.

The RDoC research framework, depicted in Figure 1, has four major components intended to provide guidance for clinical research. Two components—neurodevelopmental and environmental factors—provide the broad context for the framework as a whole. The other two—functional domains and their units of analysis—form the “RDoC matrix,” which is situated in the context of the neurodevelopmental and environmental components. Although RDoC’s contextual components have not received as much attention as its matrix, they are essential aspects to consider in programs of research. That is neurodevelopment’s inclusion in the research framework does not mean that every study must include a longitudinal component, but investigators are encouraged to address potential antecedent or consequent aspects of development and to relate their results to these associated data. In turn, environmental factors are intended to encompass the totality of external influences—both the effects of the broader environmental context, such as culture, family, class, and other social groupings and hierarchies, and also the effects of particular events, such as accidents and traumatic experiences. Environment also interacts significantly with neurodevelopmental processes, and its effects may differ by the age at which events occur or milieus are experienced. Thus, these two components constitute the broad—and indispensable—contexts that must be considered in studying one or more aspects of behavior.

Within this essential context, RDoC calls for studying psychopathology with respect to basic functional dimensions, organized into five major domains. For example, in the domain of “negative valence systems,” dimensions such as the fear system perform the biological function of enabling the organism to respond to aversive or threatening stimuli and contexts. These functional dimensions are conceived in terms of the continuous span of functioning from the normal, adaptive range to increasing degrees of abnormality in each system (e.g., mild fear to full-blown panic). They are assessed by measurements taken across several units of analysis, which range from measures of genetics and neural-circuit activation to measures of overt behavior and self-report assessments. The goal of studying these various units of analysis is not to explain behavior or cognitive activity in terms of molecular or cellular processes (i.e., reductionism), but rather to understand more thoroughly the relations among the different activities of various systems. The five domains and their seven units of analysis, which together form the RDoC matrix, are currently the most fully articulated aspects of the RDoC project, but it is important to remember that the matrix must be considered within the essential frame of the neurodevelopmental-environmental context.

RDoC-based research projects that are focused on further articulating the matrix explore psychopathology through the study of one or more functional dimensions, which are hypothesized to be related to specific kinds of symptoms that typically cut across two or more ICD or DSM disorders. For example, anxiety symptoms are hypothesized to be related to impairment in the dimension that has the function of responding to potential threat, whereas impaired social functioning and problems with concentration are hypothesized to be related to functional dimensions in the domains of, respectively, social processes and cognitive systems. Because cross-cutting symptoms such as anxiety, impaired social functioning, and concentration problems appear to be a major reason for much of the comorbidity that is particularly problematic in research studies, RDoC is agnostic to current diagnostic categories. RDoC’s vision for the matrix is to generate a research literature regarding relations among physiological, behavioral, cognitive, and symptomatic measures that can inform future versions of ICD and DSM, with a particular potential to incorporate biological and/or psychometrically advanced measures of behavior and cognition into a precise diagnostic assessment that points to more tailored treatments. Again, however, we must not lose sight of the fact that this research literature will be incomplete until it is understood in its neurodevelopmental-environmental context.

The way in which the four major RDoC components just described are developed relates to RDoC’s role as a framework for research. The five major domains of functioning comprise the “rows” of the two-dimensional RDoC matrix:

1. Negative valence: systems that enable response to aversive stimuli or contexts (e.g., threat, loss, aggression due to frustration);
2. Positive valence: systems that mediate reward-related activity (e.g., approach motivation, reward responsiveness);
3. Cognitive systems, such as attention, perception, and memory;
4. Social processes, such as affiliation and attachment, facial expressions and other social communication, and perception and understanding of the self and others; and
5. Arousal and modulatory systems, such as circadian rhythms, sleep-wakefulness, and brain-stem activation and arousal systems.

Each of these five domains contains three to six specific dimensions, termed constructs (the customary term in psychology for a functional element of behavior or cognition). For instance, the cognitive-systems domain contains constructs for attention, perception (visual, auditory, etc.), working memory (the short-term storage of relevant information—e.g., remembering a phone number long enough to dial it), declarative memory (memory for general and personal facts and events), language behavior, and cognitive control (the ability to focus on goals and to inhibit unwanted behavior).

The constructs are the heart of the RDoC system, and each was defined in a series of workshops (one per domain) by subject-matter experts using three criteria: first, evidence for the validity of the construct as an affective, behavioral, and/or cognitive dimension; second, an implementing neural system or circuit; and third, a presumed relation to one or more specific symptoms or other aspects of psychopathology. For instance, fear is a classic psychological construct that has been measured in many ways (e.g., with self-reports, observations of behavior, and physiological assessments); has a neural circuit that has been well delineated over the past 30 years, involving (at a minimum) a potential-threat-recognition signal, a memory component, and an interpretation of the signal’s meaning, which originate in the amygdala, hippocampus, and cortex, respectively (e.g., LeDoux & Pine, 2016); and is clearly important for fear-relevant behavior in many anxiety and posttraumatic disorders (Lang, Davis, & Öhman, 2000).

The units of analysis represented by the seven columns of the RDoC matrix denote measures of genes, molecules, cells, neural-circuit activity, physiology (e.g., heart rate or cortisol level), behavior (quantitative ratings of either spontaneous behavior or behavior in laboratory tasks), and self-reports (including clinician-rated interviews). The aim is that experimental studies will use measures from several of these columns to study the constructs in an integrative way, relating the measurements to each other to achieve a more in-depth understanding of the particular dimensions being studied. In addition, there is an eighth column termed “paradigms” (not shown in the figure) to reflect the fact that many measures in the other columns are gathered in the context of particular tasks or self-report assessment instruments.

As mentioned, the constructs and units of analysis are to be examined within the critical frame of neurodevelopment and environment influences, both of which interact with the constructs and the units of analysis. The overall intent of RDoC is to increase attention to neurodevelopmental and environmental effects and to free investigators from having to use descriptively oriented ICD/DSM categories, encouraging them instead to focus on basic mechanisms that are more relevant to etiology. However, the RDoC matrix does not explicitly delineate specific elements for the neurodevelopmental and environmental components in the way that the domains/constructs and the units of analysis are specified. Although this was intended to provide investigators with maximum flexibility in creating research designs that pursue the particular questions of their research, the lack of specification has been interpreted by many researchers as indicating less NIMH interest in these components; accordingly, NIMH is considering providing guidelines that may prove useful to investigators and that will rectify the false impression that the contextual components may be ignored.

It is important to examine differences between RDoC and current approaches to mental-disorder classification, as well as the implications of shifting to an alternative framework. First, in ICD-11 and DSM-5, mental disorders are characterized primarily by a description of current symptoms and associated features, such as age of onset, duration, and course. In contrast, as an experimental framework that is intended to generate new research hypotheses and data, RDoC is not bound by the need to produce a clinically usable document. Rather, it can temporarily set aside the premises of the classical approach to diagnosis to try to address the limitations of this approach for mental illness that research has revealed. In essence, RDoC inverts the usual paradigm for understanding mental illness. Its emphasis starts with elucidating basic dimensions of behavior and cognitive/affective processing together with the neural systems that play a major role in implementing them, and then considers psychopathology in terms of abnormalities of various degrees in these systems.

The goal is not to explain current syndromes in terms of these dimensions; rather, it is to characterize the kinds of symptoms or disabilities that result from abnormality in a given dimension or interacting set of dimensions, such as the effects of fear upon cognitive processes, in their neurodevelopmental and environmental contexts. The approach is fully and explicitly dimensional—not simply across the severity range of a diagnosed disorder but across the entire span of normal to abnormal functioning. This approach is currently recognized in the diagnosis of some domains of mental disorder, such as intellectual functioning and PD, and the cross-cutting symptom assessments in DSM-5’s Section III, but not in most others.

Second, although DSM and ICD still use a classical approach, clinicians and researchers today do tend to accept that the causes of mental disorder are multifactorial in nature, and yet etiological considerations are typically not a major focus in either diagnosis or treatment formulation. In contrast, RDoC is specifically focused on etiology: A major goal of RDoC is to foster research that elaborates the complex interactions among genetic vulnerability, perturbations in neurodevelopment, and bidirectional influences with various facets of the environment (broadly conceived), the effects of which also depend upon the time or times at which they occur. Similarly, RDoC encourages investigators to acquire data across multiple units of analysis (biological, cognitive-behavioral, self-reported) and to examine the results in a multivariate way (e.g., with computational modeling; Marquand, Rezek, Buitelaar, & Beckmann, 2016). The aim is to develop an integrated understanding of causal and other relations among neural systems, psychological operations, and relevant symptoms or impaired functioning. The long-term goal is not to pursue a reductionist model that explains all phenomena in terms of neural activity but rather to arrive eventually at a more sophisticated account of how symptoms of psychopathology relate to quantifiable measures in other domains.

Finally, although—to their developers’ credit—ICD-11 and DSM-5 have begun to include information concerning developmental and environmental factors, these are ancillary to their main purpose, whereas RDoC explicitly encourages theoretical models that include neurodevelopmental and environmental components. From an RDoC perspective, the first aim is to develop a comprehensive literature about the convergent and interacting roles of neurodevelopment and the environment as well as their relations with biological, psychological, and social variables in the onset and course of impairment in various functional domains. At that point, we will be in a stronger empirical position to discuss more abstract questions such as how best to define mental illness and to classify mental disorder. Thus, for the present, the RDoC approach sets aside issues related to definition and classification, as well as how best to describe patients’ current status in the context of their development and past and current environments. It is anticipated that RDoC will be able to offer relevant data to address these issues as it accrues information that can guide and be incorporated into future iterations of ICD and DSM.

Both the third and fourth editions of the DSM and the eighth through tenth editions of the ICD included a few disorders for which evidence of a psychological causation was required. For example, conversion disorder was based on the psychoanalytic concept of defense mechanisms and involved the expression of unconscious psychological conflicts as somatic symptoms (e.g., paralysis). The DSM-IV required that “psychological factors are judged to be associated with the symptom or deficit because the initiation or exacerbation of the symptom or deficit is preceded by conflicts or other stressors” (APA, 2000, p. 492), whereas ICD-10 was even more direct, requiring for a definitive diagnosis of dissociative (conversion) disorders “evidence for psychological causation, in the form of clear association in time with stressful events and problems or disturbed relationships (even if denied by the individual)” (WHO, 1992b, p. 123). This criterion for conversion disorder was rewritten in DSM-5 to eliminate one of the last remaining vestiges of “purely” psychologically defined etiology, now requiring simply that “clinical findings provide evidence of incompatibility between the symptom and recognized neurological or medical conditions” (APA, 2013, p. 318). Similar changes are proposed for ICD-11.

In addition, in recent versions of ICD and DSM-III to –IV, environmental causation in the form of exposure to one or more stressful life circumstances or traumatic events is a required part of several diagnoses (e.g., PTSD, acute stress disorder). Both ICD-11 and DSM-5 have taken this criterion a step further and include a new section devoted to disorders specifically related to stress (trauma- and stressor-related disorders), which also includes adjustment disorder, reactive attachment disorder, and disinhibited social engagement disorder and, in ICD-11, complex PTSD and prolonged grief disorder. Exposure to a traumatic event or a stressor is a diagnostic requirement for these disorders and therefore a necessary element in their etiology. However, how the stress or trauma fits into a larger etiological framework that includes both pathophysiological processes and cultural factors requires further investigation, as do relations among the various disorders specifically associated with stress, and even certain disorders for which stress is not a diagnostic requirement but that have overlapping phenomenology (e.g., mood and anxiety disorders). In contrast to the relatively few diagnoses that include explicit psychological or environmental etiologies, ICD-11 and DSM-5 contain multiple examples of etiological thinking based on biological causation. The specific ways in which these are organized are somewhat different in the two manuals, so we discuss them separately below.

Another proposed revision for ICD-11 is elimination of the problematic organic/psychogenic dichotomy for sexual dysfunctions, which partly involved moving these to a separate chapter on “Conditions Related to Sexual Health.” DSM-5 made similar changes within the limits of the fact that it remains a classification of mental disorders. Each classification implemented an approach that recognizes the potential role of many factors in contributing to the development and maintenance of sexual dysfunctions. Likewise, ICD-11 and DSM-5 share an emphasis on the importance of illicit substances, prescribed medications, and general medical conditions in the causation of mental disorders, and practitioners should consider these factors as causes before making a definitive diagnosis and commencing treatment, because substance intoxication and withdrawal, adverse reactions to medication, and general medical conditions can result in symptoms indistinguishable from those of “primary” mental disorders.

For the most part, the organization of ICD across all health conditions is not based on etiology. Rather, most of its chapters are organized according to organ systems (e.g., diseases of the circulatory system, diseases of the respiratory system) or their most characteristic symptoms (e.g., sleep-wake disorders, mental and behavioral disorders). Some chapters have multiple organization schemes. For example, the “Infectious Diseases” chapter contains groupings of disease categories based on the types of organisms that cause them (e.g., bacteria, viruses, fungi), their mode of transmission (e.g., predominantly sexually transmitted infections), the organ system they primarily affect (e.g., viral infections of the central nervous system), or their presenting symptoms (e.g., viral infections characterized by skin and mucous-membrane lesions).

Many diseases and health conditions in ICD are characterized, like mental and behavioral disorders, by multiple, interacting causes (e.g., acute myocardial infarction, type 2 diabetes mellitus). Where they are placed in ICD may reflect only one of those causal factors or another organizing principle that is considered to be clinically important. For example, diabetic retinopathy is classified with other forms of retinopathy under diseases of the visual system, even though it is known to be a consequence of diabetes mellitus, which is classified under endocrine, nutritional, and metabolic diseases. In fact, type 2 diabetes itself, even though insulin resistance is essential in its etiology, has treatment ramifications that resemble those of cardiovascular disease. Given our considerable knowledge of the mechanisms or pathophysiology of many of these disorders, however, their placement causes little to no difficulty in ICD. In contrast, even though we understand that mental processes and mental events all have substrates in the brain, the chapters on “Mental, Behavioural, and Neurodevelopmental Disorders” and “Diseases of the Nervous System” are separate in ICD, likely for two main reasons: First, brain substrates are only one aspect of the etiology and phenomenology of these disorders, which primarily involve impairments in the higher order functions of cognition, emotion, and behavior and are influenced by interpersonal, social, and cultural factors; and second, our understanding of their causal mechanisms is still rudimentary.

Further, the “Mental and Behavioural Disorders” chapter in ICD-10 is one of those with multiple organization schemes. Whereas some groupings are based on causation (e.g., mental and behavioral disorders due to psychoactive substance use), most are based on similarity of symptoms and evidence of shared validators such as familiality (the tendency for mental disorders to run in families) and temperamental antecedents. Thus, mood disorders form one grouping, whereas schizophrenia, schizotypal, and delusional disorders constitute another. Still others are now seen as unhelpful conglomerations of entities based on outdated theoretical perspectives and will be reorganized in ICD-11. For example, ICD-10’s grouping of neurotic, stress-related, and somatoform disorders is proposed in ICD-11 to be reorganized into several narrower groupings, none of which is referred to as “neurotic.”

Like ICD-10, ICD-11 will incorporate the classification of specific causal factors when they are clearly relevant to treatment strategies. For example, separate categories are provided for mental and behavioral syndromes that are symptomatically similar but caused by substances (illicit or prescribed) or an underlying medical condition (e.g., a brain tumor). Delirium is classified according to its etiology because the particular cause of a patient’s delirium is a critical factor in the immediately necessary treatment response; likewise with dementia and other neurocognitive disorders, in that etiology guides the prediction of a case’s course and outcome and the selection of management strategies.

In some areas, the classification proposed for ICD-11 goes considerably further than that of ICD-10 in incorporating etiology. For example, the ICD-10 classification of sexual dysfunctions relies on an artificial dichotomy between “organic” sexual dysfunctions, classified mostly in the chapter “Diseases of the Genitourinary System,” and “nonorganic sexual dysfunctions,” classified under mental and behavioral disorders. This mind-body split is not consistent with either current research or best practices, which are based on a view of sexual response as a complex interaction of psychological, interpersonal, social, cultural, physiological, and gender-influenced processes, any or all of which may contribute to the development of sexual dysfunctions. For ICD-11, an integrated classification of sexual dysfunctions has been proposed and a system of etiological qualifiers provided because of their relevance to treatment selection (Reed et al., 2016).

Further, as a broad classification of health conditions, ICD encompasses a variety of ways in which causal influences in mental disorders can be recorded. For example, one may note toxic environmental factors using categories from the chapter on “Injury, Poisoning and Certain Other Consequences of External Causes.” The chapter “Factors Influencing Health Status and Contact With Health Services” contains a wide range of categories for documenting contributory factors to an individual’s illness, including potential health hazards related to socioeconomic and psychosocial circumstances such as education and literacy, unemployment, problems related to the physical environment (e.g., occupational exposure, noise, pollution), the social environment (e.g., acculturation difficulty, social exclusion), housing, and negative events in childhood (WHO, 2016b). Further expansion of these categories has been proposed for ICD-11 (WHO, 2017). Moreover, ICD’s sibling classification, the International Classification of Functioning, Disability and Health (WHO, 2001), contains a comprehensive classification of environmental factors that may affect functioning and disability in the context of a given health condition, such as human-made environmental changes, supportive or nonsupportive relationships, and services, systems, and policies. The categories of the two classifications were designed to be used together to provide a more comprehensive picture of individuals’ health status and functioning (Reed, Spaulding, & Bufka, 2009).

In sum, WHO classifications offer a relatively comprehensive framework for identifying factors that may contribute to the etiology and expression of mental disorders and other health conditions. These categories could be used as a framework for additional epidemiological and clinical research and further refined on this basis, but national health-data systems and reimbursement policies generally do not facilitate systematic collection and reporting of this type of information (e.g., by reimbursing health professionals to record it as a part of standard health-encounter documentation), so the availability of these data for analysis on a global level is extremely limited. In general, WHO classification systems do not restrict the range of causes and contributory factors that may be considered in conceptualizing mental disorder. Rather, a specific subset of factors is prioritized in allocating resources for public-health data collection, research, and health-service reimbursement. These decisions are most frequently made at the level of national governments.

By relying on a mix of etiological views that incorporate both biological and psychodynamic factors, pre–DSM-II U.S. classification explicitly acknowledged multiple causal factors in the development of mental disorders. A major emphasis in developing DSM-III was to be atheoretical with regard to etiology or pathophysiological process unless one or both of these were well established (e.g., in the “Organic Mental Disorders” section for adjustment disorder, which stated “the disturbance is a reaction to psychosocial stress”; APA, 1980, pp. 6–7). This stance was taken partly to shed earlier references to psychodynamic causation, as well as to acknowledge that the cause of most mental disorders was unknown. Moreover, the possibility of multicausality was acknowledged: “Undoubtedly, with time, some of the disorders of unknown etiology will be found to have specific biological etiologies, others to have specific psychological causes, and still others to result mainly from a particular interplay of psychological, social and biological factors” (APA, 1980, p. 7). Identifying either specific biological causes or a particular interplay of psychological, sociocultural, and biological factors has proved to be difficult, although there are a few disorders in DSM-5 that have subtypes with specific, identifiable causes (e.g., narcolepsy, major and mild neurocognitive disorders).

Beginning with DSM-IV, etiological considerations were introduced into descriptive text sections devoted to risk and prognostic factors and culture- and gender-related issues. The sections on risk and prognostic factors include explicit references to environmental causation when appropriate (e.g., the relevant section on PTSD references the severity of the trauma, personal injury, etc.) but typically indicate more general risk factors (e.g., season of birth and urban rearing in schizophrenia). The culture section includes discussion of the causal contribution of cultural systems. For example, the relevant PTSD section states the following:

Most such references, however, are purely probabilistic and offer little utility in diagnosing individuals.

In sum, as knowledge on the causes of mental disorders has advanced, DSM-5 has been able to progress from the atheoretical stance of DSM-III. To the extent allowed by current knowledge, it describes biological (e.g., genetics, neurodevelopment) and both general (e.g., culture) and specific (e.g., traumatic events) environmental factors that play a role in the onset and prognosis of mental illness. It also acknowledges the complexity of these causal factors, stating “the range of genetic/environmental interactions over the course of human development affecting cognitive, emotional and behavioral function is virtually limitless” (APA, 2013, p. 19) and cautions that “a diagnosis does not carry any necessary implications regarding the etiology or causes of the individual’s mental disorder” (APA, 2013, p. 25).

RDoC is based in etiological thinking. Indeed, a major emphasis of the project is to learn more about the causes and mechanisms of mental illness, ultimately integrating knowledge relating to all four of its components (i.e., functional domains and units of analysis in the contexts of neurodevelopment and all that “the environment” encompasses).

Recent studies on psychosis provide an example of an RDoC-themed approach to identifying and understanding etiological factors in a particular disorder spectrum. Schizophrenia has for some time been recognized as a neurodevelopmental disorder, with the overt symptoms of psychosis being the end state of an extended process (Rapoport, Giedd, & Gogtay, 2012). Various lines of investigation have explored possible avenues to understanding the aberrant development that leads to psychosis. One long-standing hypothesis is that synaptic pruning—the reduction in cortical synapses that occurs as a part of normal development across adolescence (e.g., Feinberg, 1982)—is excessive in schizophrenia, such that there is an aberrant reduction in these synapses. Support for this hypothesis is accumulating: Cannon et al. (2015) found accelerated loss of gray matter in the years leading up to an overt episode of psychosis (i.e., the schizophrenia prodrome) was a critical factor in disease onset, and a genetics study has provided evidence for one mechanism by which this excessive pruning may occur (Sekar et al., 2016).

Although these reports focused on schizophrenia, a recent groundbreaking study suggests a broader and more nuanced picture. The Bipolar and Schizophrenia Network on Intermediate Phenotypes is investigating patients diagnosed with schizophrenia, psychotic bipolar disorder, or schizoaffective disorder, an intermediate category between the other two disorders (Clementz et al., 2016). Rather than comparing the three disorders against one another, the investigators set aside the patients’ different diagnoses and sought other measures that could sort them in novel ways. A clustering analysis returned two major factors of “cognitive control” (cognitive and self-regulatory functioning) and “sensorimotor reactivity” (brain activity in response to simple stimuli such as tones and lights). Various combinations of these two factors resulted in three groups, labeled “biotypes,” each of which included patients from all three diagnostic groups. The amount of gray-matter loss differed systematically across the three biotypes, but not as a function of their DSM diagnostic categories, and one biotype was associated with increased use of marijuana. Of course, no one study is conclusive, but these data illustrate the potential for biological and behavioral measures to identify intermediate phenotypes more directly related to interacting neurodevelopmental and environmental factors relevant for etiology.

As awareness has grown that virtually all mental disorders result from many different factors, ICD and DSM have both acknowledged this complexity and moved to incorporate etiological factors in the relatively few instances in which they are known. In contrast, RDoC was developed in large part to support research into the etiologies of mental disorder. Its ambitious goal is to understand how functional deviations in various brain and behavioral response systems interact to result in mental disorder, while emphasizing that these processes are developmental rather than static and that they occur in the context of individuals’ interpersonal, social, and cultural environments.

The dimensional nature of many, perhaps most, phenomena underlying mental disorders has long been clear to careful readers of the scientific literature and observers of clinical phenomenology. However, ICD remains structured as a categorical taxonomic system because this format is necessary for its application as the classification system for global health statistics and, to a large extent, for its use in clinical systems (e.g., in treatment selection and the determination of eligibility for health care services). For this reason, ICD follows particular rules and conventions that have deep historical roots and are well accepted as the basis for classification in other areas of medicine, even though one can point to many aspects of health conditions across diverse areas of medicine that are more accurately and precisely conceptualized as dimensions (e.g., blood pressure).

The ICD-11 classification of mental and behavioral disorders is required to follow the same set of structural and taxonomic rules as those used in the rest of ICD’s classification system. This requirement imposes different and much stricter restrictions on its classification model for mental and behavioral disorders than is the case for RDoC or, theoretically, even for the DSM, the taxonomic focus of which is limited to mental and behavioral disorders. At the same time, WHO’s IAG (2011) has pointed out that the inclusion of mental disorders in ICD facilitates coordination with classification of other disorders, including neurological and other medical conditions that are frequently comorbid with mental and behavioral disorders, and facilitates the search for related mechanisms of etiology, pathophysiology, and comorbidity of disease processes. The representation of mental and behavioral disorders alongside other health conditions in the ICD-11 also provides a solid basis for the parity of the mental health field with the rest of medicine for clinical, administrative, and financial functions in health care.

One way to integrate dimensional constructs into a categorical system is to divide a given dimension into ordinal subcategories. For instance, ICD-10 included subcategories corresponding to the severity of a current depressive episode in depressive and bipolar disorders, which have been retained in ICD-11 with some refinements. Other ICD-11 diagnoses have subcategories created by imposing clinically important cutoffs on dimensional phenomena, such as dangerously low body weight in anorexia nervosa and extent of functional language impairment in ASD.

Likewise, subcategories of “mental retardation” in ICD-10 were based on the severity of intellectual impairment: mild, moderate, severe, and profound. Determination of these levels was based primarily on standardized tests of intellectual functioning (i.e., IQ), using cutoffs that were relatively well accepted when ICD-10 was developed in the late 1980s. The equivalent diagnostic categories in ICD-11, now called disorders of intellectual development, provides an example of a second strategy for incorporating dimensional information in a categorical system: being more explicit about the dimensions that clinicians must consider to arrive at a particular diagnostic determination. Specifically, to derive the appropriate severity-based subcategory for a disorder of intellectual development in ICD-11, the clinician must make judgments on multiple dimensions, considering both intellectual functioning and adaptive behavior across the domains of conceptual, social, and practical skills.

The proposed changes in the ICD-11 diagnostic guidelines for paraphilic disorders, called disorders of sexual preference in ICD-10, provide another example of categories that incorporate multiple dimensional judgments. In ICD-10, the diagnostic guidelines for these disorders often merely described the behaviors they involved. In ICD-11, in keeping with the ICD’s central function as a global public-health tool that provides the framework for international public-health surveillance and reporting, a distinction has been made between conditions that are relevant to public health and indicate the need for health services and those that involve private behaviors without any appreciable public-health impact and for which treatment is neither indicated nor sought (Krueger et al., 2017). The core proposed diagnostic requirements for a paraphilic disorder in ICD-11 are, first, a sustained, focused, and intense pattern of sexual arousal—as manifested by persistent sexual thoughts, fantasies, urges, or behaviors—that involves others whose age or status renders them unwilling or unable to consent, and, second, the individual’s having acted on these thoughts, fantasies, or urges or having been markedly distressed by them. However, a paraphilic disorder diagnosis may also be assigned when the pattern of sexual arousal is associated with marked distress or significant risk of injury or death, even if it does not focus on nonconsenting individuals. This formulation thus implicitly requires that clinicians who are applying the guidelines assess several components of the diagnosis, each of which can be conceptualized along a dimension—degrees of arousal, consent, action, distress, and harm—to determine whether a diagnosis of a paraphilic disorder is warranted.

More fully dimensional characterizations of disorder entities proposed for ICD-11 have been made possible by specific structural innovations. (For a description of structural changes from ICD-10 to ICD-11, see First et al., 2015.) One example is the proposal to eliminate specific PDs in ICD-11 because of well-established problems with their validity and application in clinical systems and to replace these with a dimensional classification (Tyrer et al., 2015). Specifically, the proposed ICD-11 model contains, first, a set of essential features (i.e., features that must be present to make a diagnosis) that are centered on self- and interpersonal dysfunction. If the essential features are met, then the clinician makes a determination regarding how severe the disturbance is and assigns a diagnosis of mild, moderate, or severe PD. (A subclinical level—personality difficulty—is also included.)

PDs may then be described further through the use of six qualifiers. Five of these are trait-domain qualifiers, which are a set of dimensions that correspond to the underlying structure of the full range of adaptive-to-maladaptive personality traits: negative affectivity (the tendency to experience, and to have difficulty regulating, a wide range of distressing emotions and related cognitions), detachment (the tendency to maintain emotional and interpersonal distance), dissociality (the tendency to disregard social obligations and conventions and the rights and feelings of others), disinhibition (the tendency to act impulsively in response to immediate external or internal stimuli without consideration of longer term consequences), and anankastia (the tendency to maintain a narrow focus on controlling one’s own and others’ behavior and situations to ensure conformity to one’s own “correct” standards). As many of these trait domains may be noted as are judged to be prominent and contributing to the PD and its severity. A borderline qualifier (the last of the six) is also included and may be used if a certain characteristic pattern of maladaptive functioning is evident.

Likewise, the ICD-10 subtypes of schizophrenia (e.g., paranoid, hebephrenic, catatonic) have been proposed for elimination in ICD-11 because of their lack of validity. They are to be replaced by a set of symptom ratings that may be applied not only to individuals with schizophrenia but also those with other primary psychotic disorders (Gaebel, 2012). The rated dimensions include positive symptoms (delusions, hallucinations, disorganized thinking and behavior; experiences of passivity and control); negative symptoms (constricted, blunted, or flat affect; alogia, or paucity of speech; avolition; anhedonia), depressive mood symptoms, manic mood symptoms, psychomotor symptoms, and cognitive symptoms.

These proposals to incorporate more sophisticated dimensional elements in ICD-11 address a series of specific problems. For disorders of intellectual development, for example, the proposal addresses an inadequate measurement model that takes insufficient account of how people actually function in daily life (Tassé, Luckasson, & Nygren, 2013). For paraphilic disorders, the previous formulation led to overpathologizing private behaviors that lacked clinical relevance or public-health importance (Reed et al., 2016). The classification of specific PDs and subtypes of schizophrenia in ICD-10 (and similar elements in DSM-IV) had produced reified categories commonly seen as unchanging, lifetime diagnoses that identified specific types of people. However, these ideal types were insufficiently informative for effective management of actual patients with personality pathology or schizophrenia, and the proposals for ICD-11 better represent current scientific evidence regarding the nature of these disorders (Gaebel, 2012; Tyrer et al., 2015).

At the same time, these proposals for dimensional classification are in some ways more complex than the purely categorical approach they are intended to replace, and they may impose increased clinical and administrative burdens on their users, at least initially, when the new system is unfamiliar. After a period of adjustment, however, clinicians may find that, overall, the new system is actually simpler than the one it is replacing. For example, ICD-10 defines 10 specific personality disorders, whereas the ICD-11 proposal has only three levels of severity and six optional specifiers, requiring a maximum of nine. Nonetheless, these new proposals have little hope of being adopted and widely implemented in clinical practice unless they provide useful information at the level of clinical encounters that justifies the time and effort that learning the new systems will involve. Moreover, they will be of little use for health statistics or other policy applications based on aggregated patient-encounter data if clinicians cannot apply them consistently (Reed et al., 2013), Whether clinicians can apply these dimensional assessments appropriately and consistently and whether they find that doing so yields clinically important information is currently being tested in field studies that will influence the final form of the ICD-11 diagnostic guidelines (Keeley et al., 2016).

The limitations of the categorical diagnostic system were clearly recognized by the developers of DSM-5. Early in the planning process, it was decided that the categorical system, despite its flaws, had an intrinsic appeal to clinicians—especially psychiatrists, DSM’s key target audience—who are trained to determine patients’ diagnoses. Nonetheless, the loss of information inherent in categorical diagnoses was seen as a deficit worth ameliorating with supplemental dimensional approaches. Two key candidates for incorporating dimensions into a categorical diagnostic system were diagnostic severity and cross-cutting symptoms.

The DSM-IV contained a brief section defining diagnostic severity in its introduction and, like ICD-10, provided specific guidance for conduct disorder, MDD, and “mental retardation.” The importance of being able to note the severity of individuals’ symptoms when making diagnoses is particularly well exemplified by the latter two disorders. Individuals whose MDD is of mild severity compared with those whose symptoms are moderate or severe have been shown to respond differentially to certain treatments (Hollon & Ponniah, 2010). Further, treatment research has long used change in severity (e.g., as measured by the Hamilton Depression Rating Scale; Hamilton, 1960), in addition to the less informative categorical presence or absence of the disorder, to assess outcomes. The severity levels of “mental retardation” have been used widely to gauge the level of support needed for individuals with this diagnosis. Not surprisingly, the service needs of an individual with profound or severe intellectual deficits are considerably more intensive than those of an individual with mild deficits. Thus, diagnostic severity has implications for research, provision of treatment and rehabilitation services, service planning, and resource allocation.

Under the guidance of the Diagnostic Assessment Instruments Study Group, the DSM-5 work groups were tasked with identifying or developing measures to assess severity of a wider range of diagnoses. In conjunction with the Impairment and Disability Study Group, the members of the former Study Group proposed that these be based on the symptoms of the disorder and not on the extent of disability resulting from those symptoms. Disability—the effect of individuals’ symptoms on their ability for self-care and engagement in social and other life activities—was recognized as an important consequence of disorder severity, but a domain to be measured separately. Likewise, the level of distress caused by symptoms was felt to be a consequence of symptom severity and disability, and thus not a suitable indicator of diagnosis-specific severity. The study groups proposed that measurements of severity generally should either take the form of a symptom count, as did the guidelines for assessing severity in DSM-IV, or be based on a more fine-grained assessment of symptom frequency, intensity, and/or duration.

To reduce clinicians’ burden, measures were to have a patient-administered format, although it was recognized that clinician-completed measures might be necessary for some disorders and symptoms. To be included, severity measures were to be freely available (i.e., not proprietary), short enough for use in busy clinical settings (containing approximately 10 items at most), and, for clinician-rated measures, able to be used without formal training in their administration. When severity instruments that met these requirements were not already available for a particular diagnosis, a DSM-5 work group could develop one. Most of the severity measures recommended by the work groups were based on symptom frequency, intensity, or duration; some (e.g., the Patient Health Questionnaire–9; Kroenke, Spitzer, & Williams, 2001) were based directly on the diagnostic criteria for the disorder. Other instruments measured severity as manifested through specific aspects of a disorder, such as BMI for anorexia nervosa. As proposed for ICD-11, severity of schizophrenia in DSM-5 is assessed through several of its associated symptoms, some of which (e.g., impaired cognition, depression, and mania) are not included in the disorder’s diagnostic criteria. Finally, because individuals can have clinically significant symptoms that do not meet full diagnostic criteria, the instructions for the diagnosis-specific severity measures indicate that they may be used to assess individuals with such symptoms regardless of whether their symptoms are above threshold for diagnosis.

Soon after the publication of DSM-III, it was recognized that rigid diagnostic categories often do not correspond to clinical reality. Rather, patient presentations often include any number of clinically significant symptoms that are not among the criteria of a particular diagnosis. Some of these symptoms (e.g., sleep problems and anxiety) are seen frequently across a wide range of mental disorders. Others (e.g., suicidal ideation, illicit drug use) are less frequent but are of high clinical significance when they do occur. Increasingly, these nondiagnostic, co-occurring symptoms (e.g., presence of anxiety symptoms in MDD or depressive symptoms in schizophrenia) are viewed as important predictors of patients’ treatment response and prognosis (Conley, Ascher-Svanum, Zhu, Faries, & Kinon, 2007; Fava et al., 2008).

The DSM-5 work groups developed measures of many of these nondiagnostic, cross-cutting symptoms for such symptom domains as depression, somatic complaints, and substance use, and they serve two purposes. First, they provide documentation of the presence and severity of extra-diagnostic symptoms to guide clinicians’ treatment decisions and allow them to follow patients’ outcomes. Second, when administered before a clinical visit (e.g., via patient self-ratings), the measures draw attention to symptoms that might otherwise be missed because they are outside patients’ formal diagnoses. The DSM-5 work groups and study groups provided input as to which cross-cutting symptom domains to include.

For ease of administration, the Diagnostic Assessment Instruments Study Group recommended a two-stage process using patient-rated—or parent-rated, for children—measures. The first stage, Level 1, includes a few screening questions for each symptom domain. For several domains, for example, a score of at least 2 on one 5-point rating scale (i.e., corresponding to at least “mild” severity) indicates the need for a Level 2 assessment. The Level 2 assessments are more complete symptom assessments, which allows them to be used easily to assess outcomes, including change over multiple time points. Table 1 lists the Level 1 cross-cutting symptom domains and the corresponding Level 2 measures for adults, adolescents (ages 11–17), and children (aged 6–17, assessed through parent or guardian reports). Some Level 1 domains do not have associated Level 2 assessments. For these domains, scores above a predetermined threshold are “flags” for the clinician to conduct further individualized follow-up assessments. It is noteworthy that many of the Level 2 measures are relatively pure assessments of clinically important symptoms rather than of diagnostic syndromes, which often include heterogeneous groups of symptoms. As such, they may have use in bridging clinical practice and RDoC research on basic neuroscience and behavioral domains.

The cross-cutting symptom measures were tested in the DSM-5 field trials and had generally good to excellent test-retest reliabilities in all three age groups (Narrow et al., 2013). Importantly, the diagnostic approach that was tested in the field trials—namely, using dimensional assessments in addition to categorical diagnoses—was well accepted by clinicians. Patients and their parents or guardians found the self-rated dimensional assessments useful for describing their symptoms and helping their clinicians understand their experiences.

The cross-cutting and severity measures, which were mostly self-administered by patients, were intended for use at initial evaluation and at follow-up visits to help both clinicians and patients make treatment decisions and track treatment outcomes, consistent with the contemporary U.S. emphasis on measurement-based care and patient-reported outcome measurement. However, the APA Board of Trustees decided that there was not yet sufficient evidence that use of most of the diagnostic-severity and cross-cutting symptom measures would improve clinical care and patient outcomes to warrant approval for general clinical use. Therefore, the Board of Trustees approved only a few simple scales for assessing severity—such as BMI for anorexia nervosa, symptom counts for substance use disorders, and “setting counts” (indicating whether symptoms were seen in only one, only two, or three or more settings) for oppositional defiant disorder—for inclusion in the main Section II of DSM-5. A representative sample of other measures is included in DSM-5’s Section III as “emerging measures” needing further testing, and all measures are available online (https://www.psychiatry.org/psychiatrists/practice/dsm/educational-resources/assessment-measures). Clinicians are encouraged to use the measures and report their experiences to the APA to provide evidence of the measures’ utility for clinical practice.

Finally, as mentioned previously, the proposed ICD-11 system for PD is conceptually quite similar to the DSM-5 AMPD, but there are three notable differences between the ICD and DSM models. First, the DSM-5 model has greater specification for both functional impairment and for traits, but this has resulted in a model that WHO considered to be too complex for implementation except in research or in the most specialized settings in high-resource countries. Second, the models each have five broad trait domains but share only four: negative affectivity, detachment, disinhibition, and a domain that is called antagonism in DSM-5 and dissociality in ICD-11. Then, each has a fifth trait domain that the other does not: anankastia (a focus on the control and regulation of one’s own and others’ behavior to ensure conformity to one’s high standards) in ICD-11 and psychoticism (eccentric perceptions, cognitions, beliefs, experiences, and behaviors) in DSM-5. This domain is not included in the ICD-11 proposal because schizotypal disorder is classified as part of the schizophrenia spectrum in ICD, whereas in DSM-5 schizotypal PD’s primary placement is in the PD chapter, and trait psychoticism is needed to characterize this PD. Third, the proposed ICD-11 model includes a borderline qualifier, in contrast to the six combinations of traits forming specific PDs in DSM-5’s AMPD. This is not intended as a claim that the borderline pattern has unique ontological status. Rather, because this has been the most frequently diagnosed ICD-10 PD, its inclusion is specifically intended to facilitate the transition from a categorical to a dimensional PD model, and to give clinicians tools for documenting the variability in presentation of people who were previously considered to have BPD by also noting the particular trait domain(s) that characterize individuals’ PD presentation.

The RDoC research framework embodies a fully dimensional approach to mental disorder. This approach does not merely entail assessing severity dimensions of currently recognized disorders (e.g., a dimension of mild to moderate to severe MDD) but rather reflects the view that psychopathology should be studied with respect to the full range of operation of its various constructs, from healthy through severely pathological range.

There are multiple reasons for taking this approach. First, given RDoC’s emphasis on etiology, its near-term goal is not to improve current diagnoses or develop an alternative clinical nosology but rather to direct the research community toward a more comprehensive understanding of how a variety of factors intersect over time and across different contexts to yield various types and degrees of psychopathology. It is becoming increasingly clear that most mental and behavioral functions in psychopathology are on continuous dimensions with functioning in the general population rather than qualitatively distinct. The pathological trait criteria of PD provide a clear elaboration of this continuity, as do some criteria of MDD (e.g., diminished interest or pleasure in activities) and GAD (e.g., excessive anxiety and worry; L. A. Clark, 2005) that are related to RDoC constructs. However, similar patterns hold for other mental functions that initially might seem to be more distinct from normality.

For instance, large community studies have shown that hallucinations and delusions are distributed continuously in the population (van Os, Linscott, Myin-Germeys, Delespaul, & Krabbendam, 2009) and that the need for services is more correlated with the extent of these phenomena than with their simple presence versus absence (Kaymaz & van Os, 2010). Research also suggests that genetic loads may be continuous across clinical and nonclinical populations. For instance, parents with a child diagnosed with ASD who had one or more unaffected siblings were asked to report on the psychosocial functioning of all their children, and they reported considerable overlap in the functioning levels of those with and without the disorder. Within the range of overlap, the relation between genetic loading and functioning level was virtually identical in both groups. These data suggest that the same genetic factors are operative in both typically developing and diagnosed children (Robinson et al., 2016).

Second, RDoC’s focus on dimensions of mental and behavioral functioning provides a more quantitative basis for prevention research. Scales that have been validated for various mental functions and behaviors can help identify individuals who are beginning to trend toward dysfunction that, if the trend were to continue, could eventuate in disorder. This approach is directly comparable to approaches in other areas of medicine, such as the measurement of blood sugar, in which progressive divergence from normal levels prompts increasingly aggressive treatment options—from lifestyle changes to medication or other, more intensive interventions. Thus, as with the rest of medicine, assessment of symptom levels across the full normal-to-abnormal spectrum is fundamental to the development of preventive interventions.

An early example of the potential for this approach to mental disorders with respect to cognitive domains is provided by data from the Pennsylvania Neurodevelopmental Cohort study, in which children admitted to a hospital for a variety of reasons were given a large battery of neuroimaging and cognitive tasks and followed for several years. A retrospective analysis of the youths who later developed psychotic symptoms showed that their cognitive functioning fell behind that of typically developing children at about 9 years of age and stayed about 1 year behind normal cognitive development for the rest of the study (Gur et al., 2014). Studies aimed at understanding the nature of such developmental delays, and more precise means of identifying individuals at particular risk for psychosis, could eventually lead to targeted prevention interventions.

Both ICD and DSM, each in its own way, have made modifications to acknowledge the existence of dimensional features that are relevant to mental disorder and to incorporate dimensional features into their diagnostic and classification systems. For example, DSM-5 incorporated the concept of severity consistently across its classification system and added a set of cross-cutting symptom dimensions that can be used to provide a more complete clinical picture without using additional diagnoses that increase spurious comorbidity. Likewise, the developers of ICD-11 have taken steps toward abandoning artificial subtypes by, for example, proposing a severity dimension as the primary basis for classifying PD, with trait dimensions as specifiers. They have also proposed to implement a more dimensional system of symptom expression in schizophrenia and other primary psychotic disorders. And yet both systems remain fundamentally categorical for reasons that are germane to the various purposes for which they were developed and still are primarily used: compilation of health statistics, allocation of mental-health resources, clinical communication, and decision making in regulatory, legal, and health-insurance systems, all of which ultimately serve public mental health care needs. Providing a basis for decisions about what constitutes a case of mental disorder is a fundamental requirement of these systems, and such decisions are inevitably categorical.

In contrast, the RDoC system represents a fundamentally different approach to mental illness, with the goal of identifying the basic brain and behavioral processes that, together with sociocultural forces, give rise to multiple dimensions that may become dysfunctional and constitute psychopathology. It seeks deeper understanding of the scientific basis of psychopathology through the integration of biological and behavioral measurements, while also recognizing that these processes are developmental and embedded in interpersonal, social, and cultural contexts. NIMH’s overarching goal is to serve the public’s mental health care needs; with the RDoC system, it is pursuing this goal with a longer time horizon. Given the complexity of mental illness, the RDoC initiative sets forth a research program that may require many years before its findings materially affect current diagnostic and classification systems.

How mental disorder is defined is critical for establishing the threshold between health and disorder, and several aspects of this issue warrant discussion. First, current guidance for assigning diagnoses in ICD and DSM is based upon symptoms and signs of psychopathology, representing a descriptive approach to classification. This aspect of diagnostic classification is widely accepted. Further, like physical disorders, mental disorders are diagnosed on the basis of various tests (e.g., chest X-rays, standardized assessments of intellectual functioning) as well as observable signs (e.g., swelling, suicidal gestures) and patient-reported symptoms (e.g., pain, hallucinations). However, there is a general perception that psychological assessments are less reliable and less valid than medical tests, even though there is substantial evidence that this is not necessarily the case (e.g., Kraemer, Kupfer, Clarke, Narrow, & Regier, 2012; Meyer et al., 2001). To be sure, assessing the signs and symptoms of mental disorders requires considerable clinical judgment based on the observation of an individual and on his or her self-reported mood, cognitive and sensory experiences, and behaviors; however, identifying and correctly classifying physical signs and symptoms to diagnose medical illnesses can also rely heavily on clinical judgment. Nonetheless, the availability of objective measures for many physical signs (e.g., assessments of liver size using CT scans rather than palpations of the abdomen) creates a general—though not fully justified—sense that the diagnosis of general medical problems is more scientifically accurate than that of mental disorders.

Mental-disorder diagnoses generally rely heavily on patients’ reports of their own experiences, even though patients’ symptoms often affect their ability to report on their experiences. For example, depression affects one’s perspective on life events, so depressed people may view mildly negative experiences and events as being far more distressing than do nondepressed individuals; people with anorexia nervosa typically do not perceive how thin they are objectively; individuals with substance use disorders often deny the extent to which their behavior affects others as well as themselves; and so on. Moreover, as previously discussed, many mental-disorder symptoms are common in the general population (e.g., Nuevo et al., 2012) and exist along dimensions that are continuous with normal variation (e.g., Tyrer et al., 2015). These issues complicate the determination of where the threshold lies between the healthy and psychopathological ranges of specific symptom presentations.

The difficulty and inherent arbitrariness of assigning a diagnostic threshold for mental disorders are widely agreed upon and not conceptually controversial. Nonetheless, there is ongoing debate regarding the definition of mental disorder itself. Wakefield (e.g., 1992, 2007) has been a primary proponent of the view that to be labeled a disorder, an abnormal health condition must include an element of “harm” (e.g., personal distress or sociocultural difficulty, impairment, or disadvantage). Others have criticized this view for a number of reasons, including its limited practical utility in clinical decision making (Jablensky, 2007). Moreover, in general medicine, the degree to which signs and symptoms are distressing or impairing typically is considered extrinsic to the diagnostic process per se, although it may play a key role in determining treatment. Down syndrome and cataracts are examples of health conditions that are diagnosed without regard for the presence or amount of distress or psychosocial dysfunction.

Abstract debates about defining mental disorder aside, ICD-11 and DSM-5 still must meet their users’ need to decide such things as who should be counted as having a particular disorder in national health statistics and who should be eligible to receive and be covered financially for a particular type of heath service. In fact, determining “caseness”—whether or not a person has a particular condition—explicitly involves setting a threshold between disorder and nondisorder, and is perhaps the most fundamental requirement of these systems. Consistent with Wakefield’s view, classifications of mental disorder that are intended to guide determinations of eligibility for services or treatment decisions often have attempted to resolve the difficulties inherent in setting diagnostic thresholds by requiring that symptoms and signs be associated with clinically significant distress or impaired psychosocial functioning, a solution typically referred to, especially in relation to DSM, as the clinical significance criterion (note that this criterion implicitly acknowledges the presence of these symptoms and signs in the general population and their continuity with normal variation). We discuss the somewhat different stances that ICD-11 and DSM-5 have taken regarding this issue in their respective sections below.

As for RDoC, it is just as agnostic to defining mental disorder as it is to current diagnostic categories. By providing a framework for the exploration of functional processes that extend from the healthy to the pathological range, it aims to provide empirical data for establishing caseness in the long term rather than positing thresholds at this time.

There are a few mental disorders that essentially consist of a unidimensional symptom, such as trichotillomania, the primary symptom of which is recurrent pulling out of one’s hair, resulting in hair loss, despite repeated attempts to decrease or stop. However, as discussed, most mental disorders are multidimensional, composed of multiple emotional, cognitive, and behavioral dimensions, many of which are shared across disorders, although some are unique. Further, disorder severity is an overarching dimension common to all disorders, although how it is manifested depends on individuals’ particular problems. To diagnose a mental disorder, therefore, one must determine what kind, what combination(s), and “how much” of different aspects of a patient’s clinical presentation are needed to constitute a particular disorder. Because most individuals with mental illness show a mix of symptoms, multiple thresholds typically need to be considered in the diagnostic process. To diagnose individuals with both mood disturbance and psychotic cognitions, for example, it is necessary to determine the sequence, relative duration, and severity of these two types of symptoms before choosing a diagnosis from among schizophrenia, mood disorder with psychotic features, or schizoaffective disorder. Again, in their respective sections below, we discuss the somewhat (but not radically) different approaches of ICD-11 and DSM-5 to dealing with specific aspects of mental disorders’ multidimensionality.

Important boundaries in everyday life are often somewhat arbitrary—for instance, the boundaries between states, countries, and time zones; between ages at which people are and are not of legal drinking age or eligible for Medicare; between tax brackets, and so forth. This arbitrariness (or semiarbitrariness) is widely recognized, and although we may not like a particular threshold and may even fight to change it, we generally accept the idea that thresholds are pragmatically necessary. In the case of mental disorder also, establishing thresholds between health and disorder is accepted as pragmatically necessary, particularly when a major reason for doing so is to identify those individuals who should receive mental-health services in the context of health systems with limited resources. However, regardless of their necessity, setting thresholds for mental disorder is highly consequential for multiple reasons. First, where thresholds are placed affects clinicians’ and researchers’ conceptualization of the relevant phenomena, clinical care, and knowledge generation. “Misplaced” boundaries may even interfere with optimal clinical care and research efforts.

Second, threshold placement is highly consequential because of the many social ramifications of a mental-disorder diagnosis. Diagnoses in general have consequences that are both positive (e.g., access to and payment for treatment; the right to “reasonable accommodations” under the Americans with Disabilities Act) and negative (e.g., social stigma; loss of ability to perform certain occupations). For example, children and adolescents with ASD are eligible to receive certain educational accommodations in the United States (Carter, Skimkets, & Bornemann, 2014; Corrigan, Druss, & Perlick, 2014), but many individuals who had been diagnosed with Asperger’s disorder and their families were strongly opposed to the DSM-5’s inclusion of that disorder in the autism spectrum, at least partly because of the stigma associated with autism.

Third, diagnostic thresholds may also be highly consequential in forensic settings—for example, insofar as they may be used to establish “diminished capacity,” which affects culpability for crimes; ineligibility for the death penalty as a result of intellectual disability; or need for civil commitment because of the high risk of reoffending associated with certain forms of mental disorder. Diagnostic thresholds also are a foundational aspect of epidemiology, with estimates of the incidence and prevalence of specific mental disorders used as a basis for policy and planning. Thus, from a public-health perspective, major changes in thresholds (e.g., via revision of a diagnostic manual) have considerable implications for policy and resource allocation because they affect prevalence estimates. For example, diagnostic thresholds are used to determine eligibility and reimbursement for health services, as well as for social and educational programs. If changing the diagnostic threshold for a particular disorder were to raise prevalence estimates from 5% to 25%, either considerably more resources would need to be directed toward its treatment or the level of severity required for service eligibility would need to be raised. Public-health agencies are also concerned about the continuity of data. Continuous use of the same thresholds facilitates interpretation of changes in prevalence estimates over time—for instance, assessments of the effectiveness of public-health campaigns to reduce teenage drug use that are based on changes in the prevalence of substance-use disorder among young adults.

Health professionals use either explicit or implicit thresholds to decide whether to assign a particular diagnosis or apply a particular treatment. Some treatments are relatively benign and unlikely to have negative side effects, such as cognitive-behavior therapy for anxiety disorders. Other treatments may have potentially dangerous side effects, necessitating the patient’s full understanding of their risks and benefits, such as second-generation antipsychotic medications administered to adults with schizophrenia. The consequences of false positives (diagnoses assigned when a mental disorder is not actually present, which increase when lower thresholds are used) and false negatives (diagnoses not assigned when a disorder actually is present, which increase when higher thresholds are used) vary widely according to the specific circumstances surrounding particular diagnostic decisions.

We next discuss the different approaches that our three focal institutions take to setting mental-disorder thresholds, particularly with respect to the definition of mental disorder, its multidimensionality, and the consequential nature of mental-disorder diagnoses.

The CDDG for ICD-10 mental and behavioral disorders define a mental disorder as “a clinically recognizable set of symptoms or behaviors associated in most cases with distress and with interference with personal functions” (WHO, 1992b, p. 11). “Clinically recognizable” is a critical phase in this definition because it is not difficult to think of sets of symptoms or behaviors that typically are associated with distress or interfere with personal functioning but are not considered mental disorders (e.g., bereavement following the death of a loved one or anxiety following a job loss). After considerable discussion, the IAG (2011) recommended retaining this definition for ICD-11, favoring its simplicity over the more complex definition used in DSM-IV (discussed in the Threshold Issues in DSM-5 section), which was generally intended to encompass the same range of conditions.

This ICD-10 and ICD-11 definition is conceptually similar to the one subsequently adopted for DSM-5 (APA, 2013), but the DSM-5 definition is more elaborately worded and mentions exemptions for culturally approved responses to a common stressor or loss (e.g., bereavement) and social deviation. In the ICD-11, these exemptions are made clear in the context of diagnostic guidelines for specific relevant disorders (e.g., bereavement reactions should not be mistaken for depression; sexual behaviors should not be diagnosed as paraphilic disorders solely because they are socially stigmatized) but are not mentioned in the overall definition of mental disorder. The difficulty inherent in distinguishing mental disorder from normal variation on the basis of symptoms and behaviors alone was described in the previous section, “Categories and Dimensions.” With most conditions, there simply is no clear line that separates the two, so any threshold is to some extent arbitrary. Moreover, different thresholds may be appropriate for different purposes or in different settings. For example, in primary-care settings in developing countries, the diagnosis of depression may focus on identification of those cases with the most severe symptoms and greatest functional impairment (WHO, 2016c).

The RDoC approach to thresholds follows directly from a consideration of their consequences. As noted, the necessity of organizing research designs around DSM or ICD categories has constrained the kinds of research that can be conducted, particularly with respect to individuals with symptoms that fall below current thresholds for diagnosis. There certainly is clinical utility in setting at least some thresholds, but the problem for researchers is to determine the particular kinds of criteria to use for setting thresholds (e.g., various symptoms or types of functioning) and where to set thresholds to facilitate empirically based decisions (e.g., whether to treat, hospitalize, or prescribe medication). These types of research questions constitute a critical part of the RDoC framework. An important aim is to support research that will provide systematic information about the range of mental functioning and distress from typical levels through various levels of impairment, and about asymptomatic risk states indexed by biomarkers that may precede symptoms, such as cortical thinning (e.g., Cannon et al., 2015). Relevant findings could be used to inform future revisions of the DSM and ICD on how and where to set treatment thresholds, including whether to establish more than one treatment range (e.g., mild, moderate, severe) and offer different treatment recommendations for each range.

This aim of RDoC has two implications for research. First, the emphasis upon particular functional constructs (e.g., fear, cognitive functioning) that can be measured by various means across different RDoC units (e.g., physiological responses, observed behaviors, subjective reports) is intended to lay the groundwork for a strongly quantitative, psychometrically sound approach to assessment. Second, taking a dimensional approach will often necessitate research designs that involve an examination of psychopathology based upon continuous rather than categorical variables. For instance, rather than a design that involves patients with two subcategories of major depression (e.g., mild and severe) and controls, an RDoC design might include research participants with a range of mood or anxiety symptoms (including those with minimal or no symptoms) to explore how changes in neural reward-system activity relate to reward-related behavior in a laboratory setting, or to the correlation between reward-related behavior and cognitive performance or clinical mood states. Thus, at the current time, RDoC does not have a strong position about thresholds as they apply to contemporary clinical practice. However, a major aim of RDoC is to promote a research literature that will help clinicians in the future provide more sensitive assessments that, in turn, will lead to improved ways to determine thresholds (or ranges) for empirically based diagnosis, prevention, and treatment interventions.

Once again, we see that the public-health and clinical-use purposes shared by ICD and DSM have led them to adopt similar—although not identical—approaches to defining and using thresholds in mental-disorder diagnosis, whereas RDoC can take a more flexible approach. That is, RDoC-based research can either set one or more thresholds if doing so is important for a particular research purpose, or not address threshold issues if they are not relevant for research purposes or applications in clinical settings or health policy. A long-term RDoC goal is to provide information that will facilitate setting thresholds in diagnostic systems of mental disorder. This goal includes providing for the possibility that various thresholds will be needed for different purposes. To analogize from general medicine, research has informed the setting of various thresholds for the treatment of blood pressure and obesity (i.e., at lower levels, diet and exercise may be sufficient, whereas higher levels may require more “aggressive” treatment). However, at present, reasoned clinical judgment is still required in most circumstances.

As mentioned earlier, the first version of ICD to include a classification of mental disorders—the ICD-6, approved by the World Health Assembly in 1948—grouped mental disorders into three classes: psychoses, psychoneurotic disorders, and disorders of character, behavior, and intelligence. Psychoses were characterized by gross distortions of emotional and mental functioning, such as delusions and hallucinations, and corresponded to conditions viewed as constituting “insanity” or “madness.” Psychotic conditions included severe mental disorders such as schizophrenia, manic-depressive reaction, and involutional melancholia (a very severe form of depression) and could be caused by organic processes (e.g., as in dementia) or by substances such as alcohol or cocaine. Psychoneurotic disorders were viewed as milder mental disorders characterized by symptoms such as anxiety, depressed mood, or somatization that did not involve major distortions in perceptions and experience of reality. Neurotic symptoms could occur in the context of psychoses, in which case they were considered to represent lesser manifestations of the more severe underlying disorder and not independent phenomena. In contrast, symptoms such as delusions, hallucinations, and other gross distortions of mental functioning were not characteristic of psychoneurotic disorders. In other words, these phenomena were viewed as hierarchical: A person with a psychotic disorder could not also be diagnosed as having a neurotic disorder. The third group of disorders, however, which included alcohol or other drug addiction, “mental deficiency” (disorders of intellectual development in ICD-11), and PD, corresponded to entities that were independent of the psychotic/neurotic axis and could coexist with other forms of mental disorder. By the time ICD-8 was approved in 1965, substance-related disorders and PDs had been moved to an expanded neurotic-disorders grouping, with only “mental retardation” remaining in the third grouping. In ICD-9, approved in 1975, a distinction was made between organic psychotic disorders (those due to dementia, substance use, or delirium) and psychoses due to primary mental disorders, but the broad ICD-8 groupings were otherwise preserved. Throughout this time, the idea that symptoms of milder disorders were entirely attributable to more serious disorders and, therefore, diagnostically irrelevant remained the dominant view. Comorbidity was not a major issue for the field.

A major explosion in diagnostic classes—from four to 10—occurred with ICD-10, approved in 1990. Even so, some of ICD-10’s diagnostic classes represented clinically unhelpful conglomerations of disparate phenomena, made necessary by the limitations imposed by ICD-10’s categorical system. For example, the section “Behavioural syndromes associated with physiological disturbances and physical factors” included eating disorders, sleep disorders, and sexual dysfunctions, which have very little overlap in terms of signs, symptoms, etiology, or treatment. A hierarchical model of mental disorders was much more difficult to conceptualize in the context of this expanded array of options.

To preserve aspects of the hierarchical model, the ICD-10 CDDG contains many rules that were intended to assist clinicians in making choices among diagnoses, to reduce comorbidity rates, and to maximize parsimony in accounting for presenting symptoms. For example, per the ICD-10 CDDG, for patients presenting with symptoms of both depressive and anxiety disorders, depressive diagnoses should be given diagnostic primacy over anxiety diagnoses if the depression preceded the onset of the anxiety. Likewise, GAD should not be diagnosed in the presence of a depressive episode, a phobic anxiety disorder, panic disorder, or obsessive-compulsive disorder. (These hierarchical rules are akin those implemented in DSM-III for similar reasons.)

Other diagnostic formulations conveyed a theoretical position about relations between sets of symptoms. For example, the ICD-10 CDDG indicated that agoraphobia and panic disorder should not be diagnosed together. Rather, agoraphobia included the qualifiers with or without panic disorder. Several new “combination” categories, such as schizoaffective disorder and mixed anxiety and depressive disorders, were introduced to avoid forcing a diagnostic choice for or assigning multiple diagnoses to a single presentation. Sometimes these decision rules or diagnostic conventions were tied to explicit clinical rationales (Maj, 2005), but usually they were theoretically based and lacked specific empirical support. These rules were also sometimes difficult to follow and logically inconsistent (Kogan et al., 2016), and they were increasingly ignored over time. Moreover, they came to be seen as conceptually incompatible with a criterion-based approach to classification. If disorders were determined by the presence of a particular number of specified symptoms, and this was a valid way of defining them, why would they not be considered present because of the contemporaneous presence of another condition? (DSM-III’s hierarchical rules were largely abandoned in DSM-III-R for similar reasons.)

These and other factors (which primarily involved health conditions other than mental disorders) contributed to the use of multiple diagnostic codes to describe complex clinical presentations, which is now the globally dominant ICD (and DSM) convention in diagnostic coding and reporting. Contemporaneously, a model in ICD-10 for identifying primary and related disorders was universally unpopular and largely ignored, and has been abandoned in ICD-11. Therefore, the assignment of multiple diagnoses is often the best way to capture the complexity of clinical presentations and may be viewed in part as a marker of disorders’ severity. We mentioned earlier the problem of “double depression,” the combined diagnosis of recurrent depressive disorder and dysthymic disorder in ICD-10. The ICD-11 Working Group on Mood and Anxiety Disorders considered alternative ways to address this issue but ultimately chose to leave the two codes as they were, concluding that the assignment of two codes was the most accurate way of describing this clinical presentation even though the working group did not consider individuals with both diagnoses to have two independent mood disorders.

Some psychopathology researchers are intellectually offended by the inelegance of a two-code approach to what they view as a single mood disorder, but the fact is that at the level of brain mechanisms, we actually do not know whether the chronic state and the periodic episodes of exacerbated symptoms represent two processes or one. Still, a conceptual and scientific problem is introduced when these two codes are interpreted as representing separate, objective disease entities rather than as capturing a particular presentation, especially when mental-health policies and research policies are structured on this basis (van Os, 2016). At a pragmatic level, an individual whose condition is characterized by both of these diagnoses is likely to require different, more complex, or more intensive interventions than an individual whose condition is described by either diagnosis alone.

The ICD-11 CDDG, therefore, adopted a relatively pragmatic approach to the issue of comorbidity that views the assignment of multiple diagnostic codes as sometimes necessary to describe complex clinical conditions accurately, but does not intend this as a scientific statement about the independence of reified co-occurring disease entities. This partly explains why the term “co-occurrence” rather than “comorbidity” is used in the ICD-11. The proposed ICD-11 CDDG does specify some exclusion rules, but many fewer than were in ICD-10, and those that are included tend to be based on explicit overlap of diagnostic features rather than on theoretical considerations. For example, the proposed diagnostic guidelines for adjustment disorder note that it is common for episodes of other mental disorders to be triggered or exacerbated by stressful life experiences and that a diagnosis of adjustment disorder should not be assigned if the symptoms are sufficiently severe or specific to meet the diagnostic requirements for another mental disorder, such as PTSD or a depressive or anxiety disorder. Likewise, the guidelines state that developmental language disorder with impairment of pragmatic language should not be diagnosed in the presence of ASD. This is because the pragmatic language impairment can be considered part of the characterization of ASD in the context of ASD’s broader pattern of symptoms.

The proposed ICD-11 diagnostic guidelines contain a section called “Boundary With Other Disorders” for each category (First et al., 2015) that describes the basis on which a given condition should be distinguished from common differential diagnoses and the circumstances under which it may be appropriate to assign co-occurring diagnoses. Sometimes, this decision is based on the clinical utility of an additional diagnosis in the context of a particular case and explicitly left to the judgment of the clinician. For example, the diagnostic guidelines for stereotyped movement disorders indicate that although repetitive and stereotyped motor movements can be a characteristic feature of ASD, assignment of both diagnoses may be warranted if the stereotyped motor movements constitute a separate focus of clinical attention (e.g., because of the potential for self-injury). Likewise, the diagnostic guidelines for complex PTSD acknowledge that characteristic symptoms of that disorder related to affect dysregulation, distorted view of the self, and difficulties in sustaining relationships may be similar to those observed in some individuals with PD. When individuals meet the diagnostic requirements for both disorders, decisions about whether to assign both diagnoses should be based on considerations of clinical utility.

At the same time, the developers of the ICD-11 CDDG have taken specific steps to reduce problematic artifactual comorbidity. One example is in the classification of PDs, described previously. Comorbidity among supposedly distinct PDs and, conversely, the high frequency of such diagnoses as “other specific PD” and “PD, unspecified” in previous ICD versions were glaring problems—first, because a clear majority of individuals diagnosed with PD either meet criteria for two or more PDs or do not meet the criteria for any one, specific PD (thus meeting the criteria for one of the above nonspecific alternatives), and second, because individuals have only one personality,¹⁰ so PD comorbidity does not make conceptual sense. In addition, there is strong evidence that there is a general factor in PD (e.g., Sharp et al., 2015; Trull, Vergés, Wood, & Sher, 2013; Wright, Hopwood, Skodol, & Morey, 2016) and that the most important element in PD is its severity (e.g., Tyrer et al., 2011; Tyrer et al., 2015).

As described earlier in the section “Categories and Dimensions in ICD-11,” the ICD-11 proposal for PD reflects this evidence by offering a single PD diagnosis divided into three subcategories marked by different levels of severity, plus a subthreshold category termed “personality difficulty.” This approach is fundamentally different from that of the ICD-10, which listed eight specific PDs, plus “other specific PD” and “PD, unspecified.” As also described previously, the proposed restructuring allows any of the ICD-11 diagnoses to be characterized further as prominently featuring one or more of five personality-trait dimensions, as well as providing a borderline qualifier. This formulation facilitates diagnosis of all individuals with personality pathology and allows clinicians to describe the nature of that pathology either quite simply (e.g., as “moderate PD”) or in considerable detail, using whatever combination of prominent trait dimensions and/or the borderline qualifier best characterizes the person.

Across the ICD-11, additional features are being built into the classification, particularly for use with electronic coding applications. These features are intended to enable conceptual relations between different parts of the classification with the goal of reducing the proliferation of additional codes and artifactual comorbidity. The first of these is multiple parenting, which allows a particular code or set of codes to appear in more than one place. For example, the diagnostic entities that represent the various etiologies for dementia are listed in both the “Diseases of the Nervous System” chapter and the “Mental and Behavioural Disorders” chapter, with the same code in both chapters. This obviates the need to create a partially redundant set of codes to represent the psychiatric or behavioral aspects of dementia while conveying that dementia is a legitimate part of the scope of practice of mental-health professionals. A second feature is cluster coding, which allows codes from different parts of the ICD-11, such as codes representing the multiple manifestations of conditions such as diabetes mellitus in different body systems, to be linked so that they are understood as being different aspects of the same underlying condition in a particular individual. Although we have not progressed to this point in our understanding of mental disorders, eventually it should be possible to link codes that represent the symptomatic manifestations of an underlying causal factor such as substance abuse or of a more general dimension such as externalizing or negative affectivity.

As mentioned previously, around the time DSM-III was published, comorbid mental disorders were typically thought to be “ordered,” such that one was a complication of the other. On the basis of available information and clinical judgment, decisions were made in DSM-III regarding which disorders were more likely to be complications of others. The final criterion for approximately 60% of disorders in DSM–III was “Not due to another mental disorder, such as . . .” (Boyd et al., 1984), followed by a list of disorders that presumably might be the “primary” disorder of which the disorder being considered was a complication. For example, a criterion of GAD was “Not due to another mental disorder, such as a Depressive Disorder or Schizophrenia” (APA, 1980, p. 233). The manual offered neither a rationale to justify these hierarchical exclusion criteria nor guidance as to how to apply them, stating only that the phrase meant that the diagnosis was not to be given “if the characteristic symptoms are caused by [the other] disorder” (APA, 1980, p. 32).

However, clinicians and researchers alike found it difficult to determine whether one disorder was “due to” another, and not long after the release of DSM-III, researchers began to publish the results of studies that ignored the exclusion criteria and revealed widespread comorbidity (e.g., Boyd et al., 1984). Many exclusion criteria were eliminated in DSM-III-R (APA, 1987), thus permitting more comorbid diagnoses, and an explosion of research on comorbidity followed. Whereas from 1980 to 1989, 110 articles and books with “comorbidity” as a keyword were published, in the ensuing two decades, the corresponding figures were 3,808 and 14,057, respectively (values based on a PsycINFO search conducted November 9, 2017).

This research belied the notion that DSM disorders were distinct entities, and the challenge of comorbidity is frankly acknowledged in DSM-5:

At the same time, the DSM-5 Task Force aimed “to better fill the need of clinicians, patients, families, and researchers for a clear and concise description of each mental disorder organized by explicit diagnostic criteria” (APA, 2013, p. 5) and recognized that “it is premature scientifically to propose alternative definitions for most disorders” (p. 13). Nonetheless, the Task Force strongly endorsed efforts to reduce diagnostic comorbidity through several lines of investigation. The first was the recognition of cross-cutting symptoms—frequently occurring, frequently treated symptoms (e.g., sleep problems, anxiety). Their assessment was an alternative to assigning multiple comorbid diagnoses for treatment planning, justification of treatment to payers, and outcome tracking. Cross-cutting symptoms that need treatment but are clearly subordinate to a primary diagnosis can be documented without the need to make a comorbid diagnosis, which previously had often been a NOS diagnosis. For example, a clinician could add insomnia of moderate severity to the problem list for a patient with schizophrenia without needing to justify the treatment of insomnia with a diagnosis of insomnia disorder, other specified insomnia disorder, or unspecified insomnia disorder.

Another major step taken by the DSM-5 Task Force in relation to comorbidity was to re-examine how disorders were grouped into chapters. The chapter structure of earlier versions of the DSM was based primarily on clinical phenomena: Disorders with primary symptoms relating to mood, anxiety, and psychosis each had their own respective chapters. This symptom-based structure made sense when research findings were scarce, and it became ingrained in the provision of clinical services and in research. For example, on the basis of the diagnoses contained in the DSM-III and -IV “Anxiety Disorders” chapter, anxiety clinics might have included patients with obsessive-compulsive disorder, the primary symptoms of which are quite distinct from those of other anxiety disorders, but not patients with sexual aversion disorder, even though they exhibit significant underlying phobic and panic symptoms associated with anxiety disorders (Brotto, 2010).

Although these kinds of groupings made some sense for the provision of targeted treatment services, the DSM-5 Task Force was faced with emerging evidence that the existing symptom-based chapter structure was not entirely valid. The Diagnostic Spectra Study Group was formed to take a closer look at how the DSM-IV chapter structure might be revised to reflect not only symptom similarity but also related etiologies or pathophysiology. Revising the DSM-IV chapter structure in this way was felt to be an initial step in correcting errors in the boundaries between disorders. With disorders grouped together more accurately, artificial comorbidities resulting from erroneous, overly fine-grained splitting of disorders might eventually be corrected with ongoing research. This study group worked closely with developers of ICD-11 to ensure cross-national harmonization of the two systems’ chapter structures.

The Diagnostic Spectra Study Group selected 11 validators to investigate relations between disorders. Following E. Robins and Guze (1970), they reasoned that disorders could be considered closely related to the extent that their validators showed the same patterns. The specific presumptive criteria for grouping disorders were (a) high rates of comorbidity among disorders as currently defined, (b) familiality, (c) shared genetic and/or environmental risk factors, (d) shared temperamental antecedents, (e) shared cognitive and emotional processing abnormalities, (f) shared neural substrates, and (g) shared biomarkers. Symptom similarity, the course of illness, and treatment response were also included as potential validators, but they alone could not be used to determine relatedness.

Some of the potential new diagnostic groups targeted for study were a group for OCRDs (e.g., Tourette syndrome), and two new groups formed largely from the current group of mood disorders: specifically, a group of unipolar depressive disorders and GAD, which would reflect the anxious-misery dimension mentioned previously, and a second group for bipolar disorders. It was thought that if OCD and GAD were assigned to new groups, then remaining anxiety disorders (e.g., panic disorder, phobic disorders, and PTSD and related disorders) might be grouped as fear-related anxiety disorders on the basis of their underlying neurocircuitry. The study group also suggested that schizophrenia and related disorders such as schizotypal PD be examined for similarities with each other, as well as with bipolar disorders. The pervasive developmental disorders were seen as ripe for reexamination, particularly given that autistic disorder and Asperger’s disorder were listed as two distinct disorders despite their many shared characteristics and validators, and that pervasive developmental disorder NOS was used frequently in clinical settings, suggesting the possibility of better specification with this broad domain. Addictive disorders, for which there was evidence of shared neurocircuitry, were another suggested grouping. Pathological gambling, an established DSM disorder with ties to substance use disorders, and various less established conditions, such as internet gaming addiction, were potential new additions to this group.

In the final, published version of DSM-5, several of these options were implemented on the basis of a close examination of the 11 validators. For example, the various pervasive developmental disorders were combined into a single diagnosis, ASD, for which clinicians could specify the severity of symptoms in two domains: social communication and restricted, repetitive behaviors. Gambling disorder was included in the newly conceptualized chapter “Substance-Related and Addictive Disorders,” which encouraged research on other non-substance-related “behavioral addictions,” such as severe overengagement in internet gaming. The OCRDs were given a chapter of their own, separate from anxiety disorders, that included such disorders as trichotillomania and body dysmorphic disorder, which previously had appeared in other chapters on the basis of their symptom presentation rather than their underlying characteristics.

Other possibilities ultimately were rejected: Tourette syndrome and other tic disorders were placed in the “Neurodevelopmental Disorders” chapter rather than with OCRDs, a separate “Trauma and Stress-Related Disorders” chapter (which included, e.g., PTSD) was added, and GAD was kept with the anxiety disorders. The reasons for these decisions varied. For example, it was argued that the data were not sufficiently strong to consider tic disorders as OCRDs (Phillips & Stein, 2015; Phillips et al., 2010); PTSD was not categorized with the anxiety disorders because, as currently defined, it is much more complex; and the decision not to classify GAD with MDD and related disorders was made in large part because the forces of tradition and shared phenomenology were stronger than those of comorbidity and genetics.

A third major effort to reduce artificial comorbidity in DSM diagnoses focused on implementing a dimensional model for PD. As in prior versions of the ICD, PD comorbidity and the high frequency of PD-NOS diagnoses were major problems in DSM-III through DSM-IV. A trait-dimensional PD model was considered briefly for DSM-IV but was deemed premature. Nonetheless, DSM-IV included a brief section, “Dimensional Models for Personality Disorder,” that listed a number of existing trait models with varying numbers of dimensions, acknowledging that they “had much in common and together appear to cover the important areas of personality dysfunction” (APA, 2000, p. 690). Importantly, between the publication of DSM-IV and the beginning of the DSM-5 revision, research had shown that the large majority of what had been considered competing models of normal- and maladaptive-range personality could be integrated into a single hierarchical model, in which the dimensions of the five-factor model of personality (e.g., Benet-Martínez & John, 1998; McCrae & Costa, 1987) formed the most differentiated level (Markon et al., 2005). (Even more differentiated levels—usually called “facets”—of the five major trait domains exist, but their integration across various personality models is incomplete.) Because all personality configurations can be described using this hierarchical trait model, its implementation in DSM-5 was considered an important step in solving the artificial comorbidity problem in the PD domain. This integrated five-factor model is the starting point for the trait aspect of both the AMPD and the ICD-11 PD proposal so, as described earlier, they are largely similar.

Arguably, a trait model alone is insufficient for PD diagnosis because even extreme traits may be adaptive in certain environments (Livesley & Jang, 2005). This consideration was the basis for the DSM-5 Personality and PD Work Group’s proposal that impairment in personality functioning be the first criterion for a PD diagnosis—specifically, individuals met this criterion if they were rated at or above a threshold of 2 on a 0-to-4 severity dimension of personality functioning, with 0 representing healthy functioning, 1 denoting subthreshold personality problems, and 2 through 4 representing moderate, severe, and extreme impairment, respectively. Note that these severity levels are conceptually quite similar to ICD-11’s no PD, personality difficulty, and mild, moderate, and severe PD, despite differences in their labels and in the particulars of how they are specified.

Finally, as described earlier in “Alternative Model for Personality Disorders,” the AMPD includes six specific configurations of personality impairment plus pathological trait dimensions, which has the unfortunate consequence of retaining the problem of comorbidity. For example, a person could have severe personality impairment and exhibit a combination of pathological traits that warrants diagnosis of both antisocial and narcissistic PD. Comorbid PD diagnoses could be avoided entirely by expanding the definition of PD–trait specified (L. A. Clark et al., 2015), but the DSM-5 Task Force felt that specific PD configurations would help clinicians transition from traditional categories to a more fully dimensional system in the future. As noted earlier, the APA Board of Trustees went even further and rejected the AMPD for placement in DSM-5’s main Section II, placing it instead in Section III, “Emerging Models and Measures.”

One of the findings of the voluminous research into mental-disorder comorbidity is that there is as much overlap between clinical syndromes and PDs as there is among clinical syndromes or among PDs (L. A. Clark, 2005). Moreover, evidence indicates that common personality traits, which range from healthy to pathological, underlie both types of disorders and thus provide at least a partial explanation for both types of overlap. We mentioned earlier that comorbidity of MDD and GAD can be explained entirely by shared genes; importantly, the trait of negative affectivity (also known as neuroticism) also shares a significant portion of these common genes (Kendler, Gardner, Gatz, & Pedersen, 2007). Other research indicates that common personality traits link social anxiety disorder and avoidant PD (Watson, Clark, & Carey, 1988; Wilberg, Urnes, Friis, Pedersen, & Karterud, 1999), schizophrenia and schizotypal PD (e.g., Picchioni et al., 2010), and various externalizing disorders (e.g., substance-use disorders and antisocial PD; Ruiz, Pincus, & Schinka, 2008). The extent of overlap between personality and both PD and clinical syndromes is such that personality is most likely an underlying cause of comorbidity (L. A. Clark, 2005). That is, different personality traits are risk factors for—or confer differential susceptibility to—specific subsets of mental disorders (Hettema, Neale, Myers, Prescott, & Kendler, 2006; Khan, Jacobson, Gardner, Prescott, & Kendler, 2005; Tackett, Waldman, Van Hulle, & Lahey, 2011).

The more specified criterion-based approach of DSM and the less specified diagnostic-guidelines approach of ICD each have strengths and weaknesses with regard to reflecting comorbidity data. If DSM-5 criteria are strictly applied (e.g., by using a structured interview), comorbidity is higher than when a less structured approach is used. This is because practicing clinicians tend to focus on the most prominent or problematic aspects of the symptom picture and tend not to document all possible comorbid disorders in an individual (Wilk et al., 2006). In this sense, diagnostic practice is more consistent with the less highly specified ICD-11. Such an approach has the obvious advantage of addressing patients’ most prominent problems first, but it is disadvantageous if clinically important comorbidities are overlooked (e.g., a pattern of unexpected panic attacks in the context of social anxiety disorder). The DSM-5 cross-cutting dimensions were developed in part to help draw attention to such overlooked conditions in diagnostic evaluations. Similar psychological and behavioral symptoms are also available for coding in the ICD-11 chapter on symptoms, signs, and clinical findings not elsewhere classified.

Both DSM-5 and ICD-11 have hierarchical categorical structures that place more specific disorders in broader groupings—for instance, placing specific phobia, social anxiety disorder, and panic disorder in a grouping of anxiety and fear-related disorders. The structure of ICD–11 can be utilized for particular purposes (e.g., in primary-care settings) with more general diagnostic guidelines for the higher order categories. Specifically, ICD-11 has “parent” categories higher in the structure that subsume “child” categories lower in the structure. For example, the ICD-11 primary-care version facilitates use of less specific, higher order codes, whereas the version for specialist mental-health settings provides codes at a greater level of specification. DSM-5 is organized such that only what corresponds to the lower level of ICD-11 diagnoses can be used; if revisions to the DSM-5 were to follow ICD-11, they would require a more explicitly articulated hierarchical structure than DSM-5 now has.

Overall, there is substantial conceptual similarity in the ICD-11 and DSM-5 approaches to comorbidity. Nonetheless, in practical terms, the more highly specified criteria in DSM-5 can exacerbate the tendency for individuals to meet the requirements of multiple diagnoses.

As might be anticipated, RDoC is much more aligned with structural approaches to psychopathology than with the approaches of the two clinical manuals. Observed patterns of covariation in structural analyses of DSM disorders (e.g., Krueger, 1999; Watson, 2005) contributed to the outlines of RDoC domains, such as Negative Valence and Positive Valence. At a more detailed level, the RDoC construct of Acute Threat (“Fear”) is well aligned with the fear-disorders factor of structural psychopathology models, whereas the construct of Potential Threat (“Anxiety”) bears similarity to the distress and anxious-misery disorders. If an individual has dysfunction in, for example, the fear circuit—perhaps as a result of hyperreactivity or deficient emotion regulation—it is not unlikely for that individual to present with a range of symptoms that meet criteria for multiple ICD/DSM anxiety disorders, such as specific phobia, social phobia, or panic disorder.

Another aspect of psychopathology concerns the externalizing construct, which has largely been related to the ability to exert control over behavior. Deficient inhibitory control, commonly called impulsivity, is linked to substance use disorders, ADHD, borderline PD, antisocial PD/psychopathy, childhood behavior disorders (e.g., conduct disorder), binge eating, risky sexual behaviors, excessive gambling, and criminal behavior. Thus, if individuals are pathologically low in their ability to inhibit behavior, that tendency will likely be observed in multiple spheres of behavior and may lead to a number of different “comorbid” ICD or DSM diagnoses.

An example of this kind of transdiagnostic, construct-based approach is provided by a study of children diagnosed with ADHD that included measurement of temperament factors (Karalunas et al., 2014). A clustering statistic was used to divide the children into three groups on the basis of parents’ reports of their child’s temperament. One group had only mild symptoms of ADHD and essentially normal temperament. The other two groups had equal scores on measures of attention problems, reflecting greater impairment, but they differed in that one group was characterized as having an extraverted temperament, characterized by high activity level and high levels of positive emotions such as elation and excitement, whereas the other group was characterized as having an irritable temperament, high levels of negative emotions such as anger, and difficulty returning to baseline when upset, agitated, or sad.

Although the authors did not report comorbid diagnoses, the irritable group would be expected to include children who met the criteria for such diagnoses as conduct disorder and oppositional-defiant disorder (Evans et al., 2017), whereas extreme cases in the extraverted group might have met the criteria for a diagnosis of bipolar disorder (e.g., Masi et al., 2006). Clearly, these data suggest that observed symptomatic comorbidities may emerge as a result of relations among multiple temperamental factors, which can be studied systematically. On the basis of their findings, the authors suggested that “a biologically informed temperament-based typology . . . provides a superior description of heterogeneity in the ADHD population than does any current clinical nosologic criteria” (Karalunas et al., 2014, p. E1). Results such as these are promising for pursuing a nosology based on the underlying dimensions of psychopathology.

As we have seen with the other key issues discussed, the approaches that the three institutions take to addressing comorbidity are aligned with their basic purposes—WHO and APA to provide information for public-health statistics and clinical care, and NIMH to increase understanding of the mechanisms and processes through which mental disorders arise. Because the ICD and DSM classification systems are necessarily based on extant knowledge, they must acknowledge the comorbidity implied by this knowledge, even if that means giving a patient multiple diagnoses to provide a complete clinical picture. To their credit, the developers of each classification have taken steps to reduce artifactual comorbidity that unnecessarily complicates rather than clarifies patients’ mental-health conditions. Insofar as ICD and DSM differ, they do so in ways that are consistent with their different histories and responsive to their different constituents and purposes. In particular, ICD must have utility in a much wider range of socioeconomic and cultural environments than DSM. In contrast, inherent in RDoC’s goal to extend current knowledge is the challenge of discovering the complex causes that create the phenomenon known as comorbidity. It is hoped that the knowledge gained from RDoC-based research eventually will provide information that will be useful in improving future diagnostic and classification systems.

In response to this body of research, DSM-5 and ICD-11

• were restructured so that the classification systems’ organizations would align better with the empirical structure of psychopathology;
• introduced diagnoses that combined disorders that formerly had been considered distinct into single spectra (e.g., PD in ICD-11, ASD in both systems);
• provided various “extra-diagnostic” ways to capture patients’ complex symptom presentations (e.g., identification of specific functional problems as well as associated symptoms in ICD-11 and cross-cutting symptom dimensions in DSM-5);
• addressed the underlying continuity of symptom and disorder dimensions, from healthy to severely pathological, in various ways (e.g., ICD-11’s category of personality difficulty, which captures subthreshold conditions that nevertheless affect functioning; DSM-5’s diagnostic severity measures); and
• described in more detail the ways in which developmental, gender-related, and sociocultural processes affect the onset and form of mental illness (e.g., ICD-11’s and DSM-5’s sections on culture-related issues and DSM-5’s Cultural Formulation Interview in Section III).

Effort also was expended to harmonize the two systems. This was most successfully achieved in relation to the overall structure of both systems, although some substantive differences remain. The level of responsiveness to research-based findings in the development of both the ICD-11 and the DSM-5 clearly indicates that both systems are intended to be research-based and iterative in terms of their processes for defining psychopathological conditions and the diagnoses that result from applying those definitions. It remains to be seen whether these processes will produce greater convergence or greater divergence in the future.

Of course, both classification systems also continue to provide a defined set of traditional mental disorders that together form a vitally important framework for clinical use in diagnosing individuals with mental disorder; that help patients and their families make sense of their distressing experiences by providing both names for—and, more importantly, an organized corpus of information about—their condition; and that are intended to provide the most useful information available for essential clinical, public-health, and statistical record-keeping purposes.

In contrast, the NIMH developed the RDoC as a new framework for psychopathology research. This framework is fundamentally multidimensional, focused on discovering causal mechanisms, and centered on brain circuitry and brain/cognitive functions. From this center, it extends both “down” to genes/molecules and “up” to macro-level behavior and self-reports, and is contextualized in terms of both individuals’ development and their sociocultural environments. Finally, it is agnostic as to the nature and form of the resultant psychopathology, neither rejecting nor embracing the diagnostic entities of current categorical systems. The hope for this initiative is that it will lead to a better understanding of “basic dimensions of functioning underlying the full range of human behavior from normal to abnormal” (https://www.nimh.nih.gov/research-priorities/rdoc/index.shtml) and that from this understanding will emerge a more valid system of mental-disorder classification.

The RDoC framework is preliminary—an initial set of hypotheses that need to be tested and refined. Considerable research is needed both to test the validity of specific implications of RDoC as it currently is formulated and to develop and test alternative hypotheses that no doubt will lead to changes in RDoC. Neuroscience certainly holds great promise for increasing our understanding of psychopathology, but the embeddedness of brain processes in sociocultural contexts must not be minimized (Gallagher, Hutto, Slaby, & Cole, 2013). Mental health and illness are inherently multilevel, multicausal phenomena that require integrated frameworks linking different levels of research—for example, psychophysiological, sociophysiological, and psychosocial (Paris & Kirmayer, 2016). Through epigenetic and other mechanisms, brain function is remodeled to adapt to particular social circumstances; emergent properties of neurobiological structures can lead to healthy adaptation or pathology across different environments; and self-reflection, narrative interpretation of experience, and patterns of interaction with others (called the interactome after the connectome, which maps brain interconnections) can affect the form, onset, and course of disorder (Berrios & Marková, 2015; Kirmayer & Crafa, 2014; Lilienfeld, 2014; Meloni, 2014; Paris & Kirmayer, 2016; van Os, Lataster, Delespaul, Wichers, & Myin-Germeys, 2014). As work on RDoC proceeds, it will necessarily focus on these consequences of multicausality. Thus, we need basic researchers studying both sociocultural processes and brain circuitry, as well as their interaction. Acquisition of basic knowledge about genomics, environmental effects, neurodevelopmental trajectories, neurocircuitry and its processes is critical; however, for RDoC to achieve its purpose, we also need to devise new research strategies to examine how multiple factors—“from neurons to neighborhoods”—simultaneously and interactively affect the development of psychological health and psychopathology.

In particular, phylogenetically older brain circuits that mediate basic emotional and cognitive processes (e.g., in the midbrain and limbic system) interact with newer brain systems that mediate interpretive and context-dependent processes (e.g., in the cerebral cortices). This fundamental interactivity means that the symptoms that constitute multiple mental disorders (e.g., alterations in sleep, appetite, attention, and perception) are determined in part by individuals’ interpretations of their experiences. It is possible, for example, to develop an illness from the meaning ascribed to emotional experiences. PTSD, for example, may emerge as a result of witnessing the accidental death of a loved one, and major depression from the loss of a parent in childhood (Bifulco, Brown, & Harris, 1987). In such cases, it is typically the closeness of the patient’s interpersonal relationship with the person that modulates the pathological effect. Because interpretation is always mediated by the sociocultural contexts and meaning traditions in which the person is immersed, the study of brain processes needs to take into account the interpretive contexts and cognitive-emotional assumptions in which research assessments are embedded.

Most comorbidity is likely due to shared, “transdiagnostic” mechanisms, so we need to identify those mechanisms and determine both the extent to which they account for similarities in the relevant syndromes and what variance remains unexplained. Phrased differently, we need studies that cut across current diagnostic categories within a multidimensional perspective, as well as research to identify both the general and specific mechanisms of current treatments, asking not only what works or whether Treatment X works but, even more importantly, how treatments work. Clearly, fulfilling the promise of RDoC is a complex endeavor that will take time as well as considerable resources and effort.

In the relative short term—that is, largely on the basis of current knowledge—a set of clinically relevant dimensions is needed to supplement current classification systems. Attempts to introduce dimensions into the DSM-5 diagnostic framework met political and professional opposition. Such resistance has also begun to emerge in relation to ICD-11 and may be stimulated by initial attempts at implementation. Reasons for this resistance, such as clinicians’ unfamiliarity with or lack of knowledge about specific dimensional approaches, can be addressed largely through education, but the lack of established thresholds that would allow for specific decisions based on dimensional information needs to be rectified by clinical research. Beneficial directions would include (a) identifying the most useful dimensions and appropriate cut points for aspects of care that require categorical decisions—from “purely practical” issues, such as defining caseness for administrative purposes (e.g., billing, insurance coverage, and disability determination), to treatment decisions, such as whether to hospitalize patients or treat them on an outpatient basis—and (b) identifying the most useful dimensions for aspects of care that are more variable, such as types of treatment and their respective durations. These dimensions may be fairly directly tied to current diagnoses (e.g., subthreshold vs. mild, moderate, and severe levels of syndromes such as depressive illnesses), may be symptom based (e.g., subclinical vs. mild, moderate, and severe levels of specific symptoms, such as anhedonia or behavioral avoidance), or may reflect functional processes (e.g., dimensions of different types of cognitive processes that cut across current disorders).

There has been considerable progress in the development of dimensional self- and other- (e.g., parent-) report instruments with strong convergent and discriminant validity (Kotov et al., 2017, provide a broad, albeit not comprehensive, list), but few of these have been disseminated into wide clinical use. Most were developed for research and lack validated cut points for clinical use. The DSM-5 dimensional assessment measures themselves have not undergone rigorous research scrutiny beyond the DSM-5 field trials to determine how they may be most useful clinically. The research literature has been accruing since the publication of these measures in 2013, but slowly, because it is difficult to obtain funding for this type of practical, applied clinical research.

Measures also are needed that do not depend on self- or other-reports for patients, from those based in behavioral tasks to biological assessments across many levels—from physiology to molecular genetics. Successful examples of behavioral measures do exist (e.g., measures of intelligence and neuropsychological assessments), but most other behavioral measures have been developed for research use, and their clinical utility has not been established. Finally, how best to measure “the environment” and “development” for integrated, multilevel research has proved quite challenging and remains largely unresolved. It is important to note that the brevity of this paragraph belies the complexity and difficulty of these tasks.

Looking toward the more distant future, it will not be sufficient simply to accrue more fundamental knowledge about genomics, environmental effects, neurodevelopmental trajectories, and processes related to neurocircuitry, as discussed above. In addition, we will need to determine how to use this knowledge in complex, real-world situations—that is, how to translate basic research findings into practical methods for assessment and treatment that can be implemented in clinical settings. Even if we came to understand a great deal more than we currently do about psychopathological processes and mechanisms of change, to be most helpful in determining, for example, which of several treatments would be most beneficial for a particular patient, we would need treatment trials to test hypotheses about how these processes and mechanisms are linked to each other and to manifestations of mental disorder.

We also would need practically feasible, reliable, and valid assessment methods for patients’ psychopathology that are based on research findings but do not necessarily require clinicians to understand the underlying mechanisms in detail. (Analogously, physicians who treat cancer do not necessarily understand the complex cellular processes that give rise to malignant tumors, but as long as the nature of the tumor is diagnosed reliably and validly, they know what treatment options are most likely to be beneficial.) Finally, at the clinic or population level, we also would need to determine which implementation strategies best facilitate the uptake, fidelity, and sustainability of the range of interventions developed for particular disorders. This seemingly late stage is in fact crucially informative at the onset of the intervention-development process (Carroll & Rounsaville, 2003). From a public-health perspective, novel interventions or high-resource assessment methods (e.g., neuroimaging) that can be used with only a minority of the population are no more than minimally effective. Thus, as our basic knowledge increases, there will be considerable applied work yet to do.

It is difficult to overstate that strong forces are operating to maintain current diagnostic practices and classification systems, or to overemphasize how difficult changing them will be. The current categorical diagnostic system is entrenched not only in research but also in drug-approval, legal, insurance-reimbursement, and health-service delivery systems, such that clinicians cannot avoid using it. However, on a practical level, most clinicians implicitly recognize the limitations of diagnostic categories in treating their patients, so after making diagnoses for administrative purposes, clinicians often select treatments on the basis of their patients’ most prominent symptoms, including symptoms that are not listed in the criteria of the recorded diagnoses.

It is important that clinicians and the organizations that represent and serve them—APA and WHO, along with local psychiatric and psychological associations worldwide—continue to evolve in the way they approach their classification efforts. This includes acknowledging the limitations of current diagnostic and classification systems, supporting RDoC and other research efforts that will clarify the diagnostic issues we have described, and advocating for appropriate steps to “de-reify” mental disorders in regulatory, legislative, and legal processes. Organizations representing clinicians also play a key role in disseminating new clinical directions to their members, and this dissemination must include education about the use of dimensional assessments of psychopathology in clinical practice and the value of thinking transdiagnostically when considering treatments. To their credit, many clinicians already have embraced these principles.

As systemic changes are made to incorporate the dimensionality of psychopathology into classification systems, scholarly organizations should be in the vanguard of reflecting these changes, and universities should be in the vanguard of training professionals to use dimensional systems. Currently, the field is organized in ways that reflect the extant diagnostic systems. For example, many scholarly organizations reflect the current categorical system of disorders (e.g., the Eating Disorders Research Society, the International Society for the Study of Personality Disorders), as do a panoply of journals covering specific categories of disorders (e.g., the Journal of Depression and Therapy, Alcoholism and Drug Addiction; Autism Research) and many academic medical centers in which clinical services are organized around diagnostic categories (e.g., clinics for mood disorders, substance use disorders, and eating disorders). Concerning the last case, clinical research is typically carried out only with the patients in these clinics, with little crossover. Such a narrow approach to research inhibits study of the points of overlap and distinction among these diagnostic groupings. The landscape is changing, however. A PsycINFO search found two articles with the term “transdiagnostic” anywhere in the record before 2000; it found 89 in the first decade of the 21st century and, in the current decade, 1,200 already have been published. Moreover, the electronic marketplace, being nimbler than print-based publishers, is already leading the way: Journals such Ecological Psychology show up in our browsers when we go online to learn about the most recent scientific work.

As we worked together over many months to develop this article, we found ourselves coming back time and again to the four key issues that all three systems must and do address in various ways. In the first part of the article, therefore, we described the three systems to provide the historical and contemporary contexts within which they address these issues: etiology, categories and dimensions, thresholds, and comorbidity. We then provided in-depth discussion of each key issue, first from a general perspective and then from the perspective of each institution and its corresponding system of describing mental disorder.

As we discussed and wrote about these key issues, we gradually arrived at what we agreed was our most fundamental “take-home message”: The way one approaches the entire topic of mental disorder and, consequently, the way one grapples with, and tentatively resolves, these key issues depend largely, if not entirely, upon the purposes for which one is doing so. Put simply, the three approaches to mental illness that are this article’s focus each have both strengths and limitations and should be considered complementary rather than competing.

We further thought that the media and, more generally, the public may have viewed the purpose of diagnosis and classification of mental disorder to be only, or at least primarily, to provide descriptions of a set of well-defined disorders to guide clinicians in determining which of these disorders each patient “has.” We wish to broaden this view to encompass awareness that, first, although we know far more about mental disorders that we did a half century ago, they still are not fully understood; second, even describing mental disorder is a complex task, to say nothing of understanding how various factors combine to cause its many manifestations; and third, again, the various ways to think about mental disorder described in this article are each important for their own purposes. These purposes include describing mental disorders on the basis of the best available evidence for use by patients, their families, and their clinicians and for various governmental functions, as well as furthering the understanding of the causes of mental illness and ways to ameliorate it, and, ultimately, prevent its development or minimize its severity.

At the very beginning of this article, we provide two quotes, each of which is an analogy for aspects of our message. One, by Edmund Burke, reminds us that although both processes and dimensions are continuous and without clear boundaries, we nonetheless can distinguish and categorize them. The other, by T. S. Eliot, suggests that as we deepen our understanding of mental disorder, we will not gain knowledge that is altogether unrecognizable; rather, we will gain a new and clear perspective on what we now only dimly understand. It is our hope that the approaches to mental disorder taken by the three systems described in this article eventually will arrive at the same clear point of understanding. That day may yet be a long way off, but it is the goal toward which we strive.