Prototypal diagnosis: will this relic from the past become the wave of the future?


History's ironic sense of humor is perhaps best displayed in its playful recursiveness. Don't throw away that dated old tie or dress; with a little patience you can expect it eventually to make a retro comeback as a new/old fashion fad. And so it is with prototypal diagnosis in psychiatry: made seemingly irrelevant not long ago by the new technology of criteria based diagnosis, but now regaining in popularity and legitimacy.

It is our misfortune that psychiatric diagnosis is stuck in a purely descriptive mode — we still require subjective word descriptions as our only tool for making diagnoses because we lack any objective biological tests. This shortcoming should be corrected for the dementias in the next five to ten years, but laboratory tests for the other psychiatric disorders seem even further off now than they did when we were preparing DSM-IV twenty years ago. The subsequent explosive neuroscience revolution has taught us a great deal about how the normal brain works, but perhaps its most profound lesson is that the ineluctable complexity of brain functioning will offer us no easy answers to the elusive riddles of psychopathology.

Our descriptive categories, however defined, are wildly heterogeneous in their underlying causes. There are likely to be hundreds of different ways to arrive at what we now lump together as schizophrenia. All of our labels, whether criteria based or prototypal, are only very rough first approximations. But they are all we have, so we have to make the best of them.

The defects of prototypal diagnosis are well known. It is completely unreliable in everyday clinical life and works well only in the most hothouse of conditions: with experienced, well trained clinicians having sufficient time to make multiple dimensional ratings on the easiest diagnostic distinctions. In real life, clinicians will not bother reading or rating the prototypes just as they often don't read the criteria sets.

DSM-I and DSM-II were prototypal systems with such low reliability that clinical psychiatry was becoming the laughing stock of medicine and research in psychiatry was virtually impossible. The radical solution was the provision of criteria-based definitions, introduced by Robert Spitzer into DSM-III. He had the vision to take what had been suggested as no more than a research tool to increase diagnostic reliability and to make it the new standard of everyday clinical practice.

The criteria set method had several great advantages. Under the right conditions, reliability was good to excellent. Clinicians and researchers were now reading off the same page, allowing an easier translation from research findings to clinical practice. Researchers around the world now spoke a common diagnostic language. It is no exaggeration to say the criteria-based diagnosis saved not just psychiatric diagnosis, but also psychiatry, from being regarded as a quaint clinical art not amenable to the newly emerging methods of evidence based medical science.

Of course, the reliability of DSM-III criteria-based diagnosis was oversold. Excellent reliability can be achieved in research studies with highly trained interviewers using a structured instrument on highly selected patients. Reliability evaporates to greater or lesser degrees depending on how much the testing conditions depart from the ideal. Reliability is no better now than it ever was for those clinicians who don't know or care about DSM diagnosis, who have only minutes to do the evaluation, and whose practice has unselected patients who are difficult to diagnose. And a criteria-based system with many narrowly defined categories is bound to have artificial “comorbidity”, with each label providing no more than one building block of a complete diagnosis.

So where do we stand now? Drew Westen nicely lays out the advantages of prototypal diagnosis and his proposed method for accomplishing it. As he suggests, the best application of his system would clearly be in the diagnosis of personality disorders. We actually thought of including something similar as an option in DSM-IV, but decided not to because it seemed unlikely that busy clinicians would have the time or interest to carefully study the prototypes of each personality and provide the necessary multiple ratings.

That is certainly still the case today, but Westen's approach is far superior to the impossibly complicated and cumbersome dimensional systems that have been concocted by the DSM-5 personality disorders work group. And an even more simplified prototypal approach (without ratings) has been chosen for ICD-11, a return to the DSM-II fashion.

Complicated criteria-based and simple prototypal diagnosis can be complementary, not necessarily competitive. It is too bad that DSM-5 and ICD-11 were not developed in closer coordination. I would favor a nested, compatible system, with ICD-11 providing the simple, quick prototypal descriptions that are a shorthand for the complicated, time consuming DSM-5 criteria sets. The prototypes would be used in clinical settings where time and ease of use are essential; the criteria sets would be for more specialized clinicians and for all research. This would give us the best of both worlds while we await the likely slow progress toward diagnosis by biological tests.