The DSM is based on extensive observations of patients, with suggestions on categories going back over 100 years. The originators commented that the models were not entirely adequate and needed further modifications1. Current models, too, have been called “a first approximation” needing additional features to achieve better utility and validity2. Specific issues identified as needing improvement include reliability, validity, completeness and utility3, 4.

While standard clinical practice does employ DSM diagnoses in making treatment decisions, it often emphasizes additional information from patient assessment. That is, physicians often use a broad problem solving rather than a diagnosis specific approach5.

Explicitly targeting utility and completeness, we asked a sample of clinicians, by an online RedCap survey, how they use DSM diagnoses in the context of other clinical information in assessing and treating psychotic disorders (i.e, schizophrenia spectrum and bipolar and major depressive disorder with psychotic features). Psychiatrists surveyed were at 27 academic centers in the US, as they are the greatest users of DSM and are most engaged in ongoing consideration of how to choose and use DSM criteria. Answers were anonymous and physicians did not receive any compensation for completing the survey. The study was approved and classified as exempt by the Partners Healthcare institutional review board.

Respondents ranked the importance in their practice of nine clinical assessment considerations (DSM-5 diagnosis, specific presenting signs and symptoms, severity of signs and symptoms, history of signs and symptoms, comorbidities, treatment history, social assessment, family history, and medication history), rated for each of four clinical decision and intervention domains (prognosis, recommended level of care, recommended medications, and recommended psychosocial therapies), using a five-choice Likert-type scale ranging from not important (assigned a value of 1) to extremely important (assigned a value of 5).

Of 566 psychiatrists who were invited to participate in the survey, 129 (22.8%) responded. They represented both sexes, and many ages, regions, sites and types of practice. Results indicated that all of the nine assessment considerations were considered at least moderately important for at least one clinical purpose. Primary hypothesis testing found highly significant evidence of a greater mean rating for current signs and symptoms than other clinical assessment considerations (X2=667, p<0.001). Using a secondary intersection-union approach, we found strong evidence that psychiatrists rate current signs and symptoms as more important than every other assessment consideration included in the survey (mean importance rating=4.46, t=5.86, p<0.001). DSM-5 diagnosis had the lowest observed mean importance rating (mean=2.77).

Post-hoc t-tests found evidence that the mean for DSM-5 diagnosis was significantly lower than the mean for every other ­assessment consideration (mean>3.58, t121-123<–9.65, p<0.001) except family history (mean=2.84, t123=–0.77, p=0.44). Post-hoc tests using linear regression found no association of the difference in mean importance ratings between current signs and symptoms and DSM-5 diagnosis with age (t122=–0.43, p=0.67); sex (t120=1.04, p=0.30); US region (X(4)2=1.21, p=0.88); site (categorized as hospital only, hospital and other, private practice only, and clinic only, X(3)2=2.37, p=0.50); and number of patients seen (X(4)2=0.97, p=0.91).

We did not sample all possible elements that clinicians use in assessments, but had an open question where psychiatrists could note factors not surveyed. Factors suggested included: previous diagnoses, age, cognitive function, risk or history of suicide or violence, forensic history, legal status, cultural background, social networks, work history, family involvement, insight, acceptance of illness and treatment, preferences among treatments, rapport between doctor and patient, and financial resources.

While our survey was being completed, a worldwide screen of expert opinion from mental health clinicians, assessing the value of ICD-11, which is similar to DSM in its categorical approach and content, was published6. This global survey addressed all the categories in the ICD and DSM, exploring the relative use of ICD/DSM for administrative purposes, managing treatment, communicating with other treaters, and teaching. Our survey targeted only US psychiatrists, focused on psychotic disorders, and obtained relative rankings of the use of DSM diagnoses versus other clinical findings in choosing and guiding treatment. Thus, the two studies were partially overlapping. Consistent with our project, the authors of the global survey concluded that the ICD and DSM categories are most useful for administrative and billing purposes and for communicating with other clinicians. They are least used and substantially less useful for choosing individual treatments or advising on prognosis.

Our results suggest that, among patients with psychotic disorders, the DSM-5 diagnosis is less important than identifying other individual features of illness, especially type and severity of symptoms, but also comorbidities and some aspects of personal history. Relevant factors noted by other investigators include suicidality, recreational drug use, obstetric complications, early or recent adverse events, social cognition and neurocognition5. The use of these factors allows more flexibility in description than categories alone. Course can be included as well.

Notable for interpreting the responses, we only contacted clinicians at well-known academic centers. The majority (70.5%) of respondents had hospital-based practices, but this might be expected for those who treat many patients with psychotic disorders. The results represent opinions of clinicians who teach and perform research, in addition to their clinical practices. Most psychiatrists did not respond. Nonetheless, the response rate (22.8%) was typical of online surveys7. Possibly, those who did respond were interested in the subject and might have thought about the matters raised. We are not suggesting that responders were representative of US psychiatrists, but it might be noted that the suggestions, made a century ago, on which ICD and DSM are based, were also from clinical observations, largely from clinicians in select sites. They were not made or since have been confirmed on the basis of other validators1.

Lastly, an argument has been made that changes in DSM and ICD should strive to improve utility and accuracy8. Accuracy in choosing treatments and predicting outcome might be enhanced by incorporating factors that clinicians cite as most important into formal diagnostic systems. That these factors are already in use for making clinical decisions shows that they are practical and suggests that they may be valid. An improved system might incorporate both categorical entities and additional features, such as those provided by recognizing individual symptoms and severity of those symptoms, in new models9. Such models can be tried and tested, then implemented if they show advantages compared to existing systems.