Just as hybrid cars with engines that use both electricity and gasoline become more common, do not be surprised to hear more about hybrid models of psychiatric diagnoses included in DSM-5. Psychiatric diagnoses are conceptualized using a top-down approach and a bottom-up approach. Using the top-down approach (ie, a categorical or disease model), expert clinicians use their own experience, the existing literature, and new data in order to define diagnostic criteria for a specific mental disorder. The DSM system is based on the categorical model approach with the main assumption that psychiatric disorders are separate disease entities.
In contrast, a bottom-up approach (ie, a dimensional model) is driven primarily by symptoms collected from patients’ reports, and statistical analyses are used to determine which symptoms cluster together into syndromes. In reality, the categorical and dimensional model approaches are 2 sides of the same coin as you look at the same patient from 2 different angles.
Both categorical and dimensional approaches have benefits and drawbacks. Categorical models have been widely used in psychiatry since the publication of DSM-III in 1980 and have improved the reliability of psychiatric diagnoses. The categorical approach in DSM has fostered progress in basic, epidemiological, and clinical research, as well as treatment.
Categorical models are commonly used in other specialties and put psychiatry in line with the rest of medicine. Generally, categorical models are preferred to dimensional models by clinicians because they are parsimonious and useful. One diagnostic label can convey a considerable amount of useful clinical information in a vivid, succinct manner. Categorical models facilitate decisions making (eg, whether a patient has a disorder, whether a patient needs a certain medication, whether a patient needs to be hospitalized). Overall, communication among clinicians is easier using categorical models.
However, categorical models in psychiatry have several problems, such as the questionable validity of DSM categories. Using categorical models can result in loss of important information. Categorical models evaluate the presence versus the absence of a disorder and do not allow for evaluation of the severity of the disorder in question.
By themselves, categorical models do not necessarily tell the clinician how severe the condition is. For example, a patient diagnosed with schizophrenia, paranoid type, may be stable on one antipsychotic medication and live in the community with outpatient follow-up. Another patient diagnosed with schizophrenia, paranoid type, may have delusions and hallucinations in spite of treatment with 1 or 2 antipsychotic medications and may spend a significant amount of time in the hospital. Because of the artificial boundaries of disorders, many patients do not fit any one category. Consequently, diagnoses of Not Otherwise Specified (NOS) are popular in many clinical settings. Many patients present with symptoms that meet criteria for several conditions. Finally, some researchers posit that the slavish adoption of DSM criteria may have hindered research on the etiology of mental disorders.
In comparison to categorical approaches, dimensional models start and build on the symptoms and therefore reflect clinical reality. In psychiatry, we treat the totality of symptoms and signs. A patient with depressive symptoms receives antidepressant medication, a patient with delusions receives antipsychotic medication, and so on. Dimensional models provide a greater amount of information and allow measuring of the severity of psychiatric disorders. Furthermore, dimensional models display a higher degree of validity, reliability, and stability over time. Finally, dimensional models are preferred for hypothesis generation and testing, and they provide greater utility in research contexts.
However, dimensional models are more complex and more difficult to use than categorical models. Providing too much information can be a hindrance rather than help to clinicians, especially if the information is not clinically relevant. For example, the Positive and Negative Symptom Scale (PANSS) uses a 7-point scale (absent, minimal, mild, moderate, moderate severe, severe and extreme).1
Most of the time, minimal and mild symptoms are not clinically significant and do not affect the clinician’s decisions. Similarly, the difference between moderate and moderate-severe or severe and extreme do not affect clinicians’ decisions in general. That explains why psychiatrists do not use the PANSS in clinical settings. In general, clinicians do not have the time, training, or inclination to use dimensional ratings.
Findings from a survey of Current Psychiatry readers show that most respondents did not use any of the 4 clinical rating scales routinely used in clinical trials, which are required for FDA approval of psychiatric medications.2 Lack of time was the most common reason cited for not using these tools.
Moreover, no dimensional model has achieved wide acceptance. The clinical utility of dimensional models is yet to be determined. Therefore, efforts to empirically demonstrate the clinical utility of dimensional models should be a prerequisite for their future implementation, in order to establish that their advantages outweigh the disadvantages.3 Finally, communication among clinicians using the dimensional model is more difficult than with the categorical model.
Historically speaking, after the introduction of DSM-III in 1980, the categorical approach gained ground and DSM diagnoses have been used worldwide in clinical and research contexts. Advocates of the categorical models hoped and assumed that it was only a matter of time until scientists discovered the etiology of psychiatric disorders as categorized by DSM. To the disappointment of categorical model advocates, more than 30 years after the publication of DSM-III, the goal of validating psychiatric disorders by discovering common etiologies remains elusive.The case for schizophrenia is especially alarming; after decades and millions of dollars spent on research, biological markers or specifics for schizophrenia as a distinct disease have yet to be identified.
As the DSM-5 Work Groups began to work on the DSM-5 revision, a decision was made to incorporate the dimensional models into DSM-5. All researchers agreed to keep the disorders categorical (with new changes) and add dimensional measures along with DSM categories, hence, the concept of hybrid models was born. Common sense indicates that if you have a hybrid model, then you have the best of both categorical and dimensional approaches. The challenge for the DSM-5 is to “find ways to create dimensional measures that are compatible with categorical definitions and not overly disruptive to clinical practice.”4
The DSM-5 Task Force decision to include a dimensional component to DSM categories seems reasonable but ambitious. In a recent publication, the DSM-5 Task Force declared “one of the major—if not the major—differences between the DSM-IV and DSM-5 will be more prominent use of the dimensional measures in DSM-5.”5 Because the DSM-5 Task Force put all or most of their eggs into the basket of dimensional measures, you would expect the DSM-5 Task Force to allocate all available resources to make sure that the major part of DSM-5 changes are successful.
The DSM-5 Task Force recommends a group of dimensional scales to be used with DSM-5 and expects clinicians to use them. The proposed measures are vague, complicated, not designed for clinical applications, not useful to clinicians’ work, and not researched adequately to justify their use. It would require another article’s worth of space to explain in detail why this is the case. The most that the DSM-5 Task Force can hope for with the proposed measures is to find researchers who can solicit funding and use the proposed measures in the funded project. When the project is over and the funds dry out, the measures will be placed on the shelves, along with many other devised and unused measures.
This criticism echoes what Frances has said about the DSM-5 process: big ambitions and weak methodology.6 It takes many years to develop, test, refine, and retest new scales and structured interviews. It took more than 10 years for the developers of SCAN (Schedule for Clinical Assessment in Neuropsychiatry), SCID (Structured Clinical Interview for DSM Disorders), MINI (Mini International Neuropsychiatric Interview), and other instruments to develop and test these tools.7-10
After I finished my psychiatry residency, my main research interest was the use of psychiatric scales and structured interviews in clinical settings. Even with the plethora of structured interviews and psychiatric scales available, I could not find a single tool that was applicable to the daily work of psychiatrists in real clinical settings. None of the existing instruments was really designed for working psychiatrists in real clinical settings. Therefore, I took upon myself to develop a new instrument—the Standard for Clinicians’ Interview in Psychiatry (SCIP).11-16
I envisioned the SCIP as an instrument that would capture the dimensional component of psychopathology, in addition to the categorical component of psychiatric disorders—a hybrid model. More than 50 colleagues in 4 countries (USA, Canada, Egypt, and England) took part in the SCIP project for 10 years and data were collected on 1010 subjects—making this the largest validity and reliability study of a structured instrument. The SCIP was designed to yield dimensional measures and scores for obsessions, compulsions, depression, mania, suicide, delusions, hallucination, agitation, disorganized behavior, negative symptoms, catatonia, attention, hyperactivity and drug addiction, in addition to psychiatric diagnoses according to DSM and ICD criteria.
Any classification system has to be useful to clinicians to entice them to use it. If the classification system is not used by clinicians, no matter how good the system is, the system is a waste and will end up filed away in the bottom drawer or on the shelves of history. At the end of the day, a disease classification has to be useful and used by clinicians to serve its major purpose.