After DSM-5: A Critical Mental Health Research Agenda for the 21st Century

While DSM-5 contains a new slimmed down definition of mental disorder (p. 21), it is not an improvement. It is vague and provides no clarity regarding the boundaries between what is normal and what is mentally disordered. In fact, no definition of “normal” that would allow the differentiation of DSM mental disorders has ever been provided (Kirk et al., 2013) Thus, the DSM is a medicalized dictionary defining the criteria for various mental disorders but without ever specifying exactly what a mental disorder is. Obviously, from a scientific standpoint, this is troubling (see Boyle, 2005, pp. 222–231; Greenberg, 2011; Kirk et al., 2013; Phillips et al., 2012; Wakefield, 1992, 2005).

The question of whether behaviors labeled as mental disorders should be considered disease entities is hardly new and has been addressed in some detail in the scholarly literature (Gomory, 1998; Kirk et al., 2013; Szasz, 1997). All problems defined in the DSM-5, from adjustment disorder to sexual problems to shyness, are just claimed to be medical (see Lane, 2008). This factual assertion takes place in the context of the well-documented fact that there are no biological tests, markers, or well-controlled studies identifying a biological lesion for any mental disorder classification (Frances, 2009; Ross & Pam, 1995; U.S. Department of Health and Human Services, 1999; see also Albert et al., 2011; Whitehouse, 2008).

The DSM-5 (2013) uses confused wording on this issue, stating that a mental disorder is a “syndrome” (p. 21) and elsewhere arguing disease by stating that the “DSM, like other medical disease classifications…” (p. 5, emphasis added). Thus, according to the developers of DSM-5, mental disorders are both syndromes and diseases—conflating two terms with importantly different meanings in medicine (syndromes being the minimum, nonrandom consistent grouping of observations necessary to hypothesize possible medical disease; to validate a syndrome as a disease, objective physiological markers must be found; see Boyle, 2005; Gomory, 1998).

However, the process of the DSM developers indicates that they do recognize a difference between confirmed medical disease and mental disorder. Rett’s disorder was included in the DSM-IV-TR (2000, pp. 76–77). In between the release of DSM-IV-TR and DSM-5, the etiology of Rett’s disorder was discovered (Lasalle & Yasui, 2009). The APA addressed this issue by writing “Like other disorders in the DSM, Autism Spectrum Disorder (ASD) is defined by specific sets of behavior and not by etiology (at present) so inclusion of a specific etiologic entity, such as Rett's Disorder is inappropriate” (American Psychiatric Association, n.d., as cited in Deacon, 2013, p. 851). Although Rett’s syndrome can now be used as a descriptor under other diagnostic categories, the disorder has been deleted from DSM-5. Deacon (2013) points out that this amounts to “The removal of a psychiatric diagnosis from the DSM upon discovery of its biological cause” (p. 851). Thus, the DSM-5 only includes mental disorders for which we lack information regarding etiology, which may or may not turn out to be diseases in the long run; but if their etiology is discovered, confirming that they are diseases, apparently they cease to be mental disorders—although the DSM also claims that mental disorders are diseases. This confusion and discordance obviously casts doubt on the idea that all DSM-5 disorders represent medical diseases.

Comparing the DSM-5 with previous editions suggests that the current DSM contains “spin” perhaps intended to manage the many ongoing controversies and public debates concerning psychiatry. The DSM-5 field trials (see below) are not published in the DSM itself (as in DSM-III) or in associated sourcebooks (as in DSM-IV). Instead, they have only been published in the peer-reviewed literature, where most clinicians are unlikely to ever see them. The DSM-IV (American Psychiatric Association, 1994) and DSM-IV-TR (American Psychiatric Association, 2000, p. xxxi) both included a section on “Limitations of the Categorical Approach” which was transparent about the weaknesses of categorical diagnosis (see Jacobs & Cohen, 2003). Although DSM-5 is also a fundamentally categorical diagnostic system, this section has now been deleted. The DSM-IV-TR mentions that brain changes in schizophrenia may be related to treatment with antipsychotic medication and notes that antidepressants may cause akathisia (a dangerous adverse effect associated with suicidal behavior; Healy, 2004). Despite the increasing body of literature demonstrating the clinical importance of these issues (e.g., Ho, Andreasen, Ziebell, Pierson, & Magnotta, 2011; Stahl & Lonnen, 2011), references to both were deleted from DSM-5.

The reasons for these changes is unknown, but they are noteworthy, especially since they result in providing practicing clinicians with less information about the iatrogenic effects of psychiatric treatment. Space constraints prevent a comprehensive list of how DSM-5 has changed in terms of framing, omitting, and shaping how psychiatric diagnosis is presented to the user of DSM-5. An in-depth analysis of DSM-5 regarding these issues would be a contribution to the literature.

From DSM-III (1980) forward, the DSM was marketed as having high interrater reliability—that two independent clinicians seeing the same client will reach the same DSM diagnosis a high proportion of the time. Mental health textbooks commonly claim that this is the case (Lacasse & Gomory, 2003). However, the meticulous analysis by Kirk and Kutchins (1992) points out the problems with such claims. Past reports of high interrater reliability owe more to the biased research design and interpretation of data than to the reliability of the DSM, and there is no evidence that clinicians in routine practice are able to attain such results (Kirk et al., 2013; Kutchins & Kirk, 1997).

The DSM-5 developers realize how crucial reliability is to the scientific credibility of the DSM, writing, “Reliable diagnoses are essential for guiding treatment recommendations … [and] … for clinical and basic research” (APA, 2013, p. 5). The more recent DSM-5 field trials sought to examine interrater reliability under realistic conditions. However, their design did contain some acknowledged biases, such as prescreening clients and performing the field trials at sites that had very high prevalence rates for the disorders under study (Regier et al., 2013). All the same, the resulting κ values for many diagnoses were quite poor. The pooled kappa (κ) values for major depressive disorder was 0.28, for generalized anxiety disorder, 0.20, and for schizophrenia, 0.46 (Regier et al., 2013). Given the design of the field trials, these results probably represent higher interrater reliability than that which would be attained in routine clinical settings.

Over 30 years ago, DSM-III claimed to solve the reliability problem in psychiatric diagnosis (Kirk & Kutchins, 1992). A generation of clinicians and academics has largely operated under this assumption. The DSM-5 field trial data demonstrate that DSM-5 categories are unreliable (Kirk & Kutchins, 1992; Kirk et al., 2013), a crucially important issue that should be addressed in both research and practice.

The DSM is often said to have utility in terms of clinical communication (e.g., APA, 2013). However, it is unclear how unreliable diagnostic labels can be helpful in clinical communication. A lack of reliability also impacts the enterprise of evidence-based or evidence-informed practice. If a client’s problem cannot be assessed reliably (e.g., different clinicians reach discordant diagnoses with the same client), the problem definition phase of the evidence-based practice model may be undermined. That is, the “answerable question” (Mullen, Bledsoe & Bellamy, 2008) sought in the evidence-based practice model may be wrong. Similarly, the unreliability of psychiatric diagnosis could impact research on evidence-based treatments. If diagnoses are unreliable within psychiatric research, the groups studied within randomized controlled trials will represent a heterogeneous group and the outcomes from such studies cannot be reliable and valid. Thus, the unreliability of psychiatric diagnosis can confound both the defining of client problems and the application of research evidence to clients. Research on clinical communication and the use of evidence in practice should integrate the known unreliability of psychiatric diagnoses.

Following diagnosis, clinical social workers may describe the putative cause of the diagnosed DSM-5 mental disorder in terms that are incongruent with the neuroscience data. For instance, they sometimes explain that depression is caused by serotonin deficiency (Acker, 2013). Clients are likely to absorb such messages as scientific facts (Cohen & Hughes, 2011). This should disturb those who think clinical practice and informed consent should be based on evidence, as serotonin deficiency as a cause of depression is known to be a myth (Lacasse, 2005; Lacasse & Leo, 2006). Similarly, along with a DSM-5 diagnosis, clients may receive a pessimistic prognosis discordant with the actual data on mental health recovery (Harrow & Jobe, 2007; MindFreedom International, 2013). Research efforts that examine what clients are told about their DSM-5 diagnosis will be valuable. Interventions that provide rigorous data to clinicians may have the potential to help clients (Cohen, Lacasse, Duan, & Sengelmann, 2013).

The utility of the DSM for financial reimbursement of services rendered is well documented (Greenberg, 2010; Gomory, Wong, Cohen, & Lacasse, 2011; Kutchins & Kirk, 1987, 1997). In both agency and private practice settings, a DSM diagnosis is often required to receive payment for clinical work. A national survey of social workers found that over 90% rated reimbursement as a common reason for using the DSM-IV, but that only 50% would use the DSM if not required (Frazer, Westhuis, Daley, & Phillips, 2009). This illustrates the difference between bureaucratic/financial and scientific/clinical utility. In an era of evidence-based practice, the question for researchers is, what scientific or clinical utility does the DSM-5 have? This is even more important in light of the research data showing that DSM-5 is unreliable.

By defining the problems of social work clients in biomedical terms, DSM-5 diagnoses may cause harm to clients. Using the language of “mental disorder” and presumed underlying brain disease or defect could have important effects on how clients view their problems and negatively impact their capacity to recover from them (Kinderman, Read, Moncrieff, & Bentall, 2013). Feminist psychologist Paula Caplan (1995, 2011) has written extensively about the potential harm inherent in DSM diagnoses (see also Caplan & Cosgrove, 2004; Kutchins & Kirk, 1997). Furthermore, a DSM diagnosis is often a pathway to treatment with psychiatric drugs. Robert Whitaker (2010) has hypothesized that psychiatric drugs often result in short-term benefit but worsen long-term outcomes, a provocative hypothesis to be sure, but one that deserves close consideration (see Littrell & Lacasse, 2012b). These underresearched topics deserve more attention from the research community.

In the wake of these criticisms of DSM-5, it is natural to ask what might be done to move the field forward. Frances (2013b) has argued for a process of stepped diagnosis, where a conservative diagnostic process takes place over an extended period of time, starting with the least impacting diagnosis possible. This is an effort to minimize stigma and invasive treatment when it can be avoided. In many practice settings, this would represent a positive step forward. Garland and Howard (2014) argue for a transdiagnostic approach to human distress. Others argue for the rejection of psychiatric diagnosis (e.g., Anthony, 2004; Boyle, 2005) and the clinical impact of such approaches needs to be further tested. The potential delinking of diagnosis and reimbursement, as well as the delinking of diagnosis and drug treatment, also offer intriguing possibilities (e.g., Kirk et al., 2013). Finally, the use of DSM-5 Z-codes (e.g., “Phase of life problem,” “Relationship distress with spouse or intimate partner”; APA, 2013, pp. 895–896) in lieu of psychiatric diagnoses should be investigated. There are macrolevel barriers to implementing such ideas in real-world practice settings, but research delving into these alternative approaches would be extremely valuable to the field.

In time, empirical testing may demonstrate that some of these conjectures are in fact wrong. But for now, they appear to be some of the most compelling issues facing academic and clinical social work following the release of DSM-5. Allen Frances described the DSM-5 as potentially taking psychiatry “off a cliff” (Greenberg, 2011). Will the field of social work simply be a helpless passenger along for the ride (Gomory et al., 2011) over this metaphorical cliff? Or can the scientific furor over DSM-5 be harnessed for the purposes of innovative independent research, rigorous problem solving, and critical testing, for the ultimate benefit of our clients?