Thoughts on Kendell’s The Role of Diagnosis in Psychiatry


Robert Kendell (1935-2002) (not to be confused with Kandell or Kendler) was a Welsh psychiatrists who explored many fundamental questions about psychiatry. He is perhaps most famous among philosophers for his 2003 article with Jablensky (distinguishing between utility and validity of psychiatric syndromes), indeed, my interest in that article has resulted in me exploring his earlier work.

His 1975 book covers a lot of ground, including reliability, validity, categorical, dimensional and disease entities. It gives a qualified defense of psychiatry, partly reacting to anti-psychiatrists whilst discussing contemporary issues which were then influencing the formulation of DSM III. His basic message might be ‘psychiatry is not as bad as people make out but it could be improved and we have lots of options for going about doing so’. The book is primarily a discussions of those options for the future.

What struck me was just how familiar it read, the problems he identifies and the various solutions he discusses look very recognizable to a modern philosopher of psychiatry. This was especially true in the introduction (that introduction would be a candidate for employment on undergraduate teaching for philosophy of psychiatry). I just felt like modern philosophy of psychiatry had told me little new on these topics which Kendell’s book had not already discussed (granted, discussed in limited detail, being a short book which covers a lot of ground). An exception would be statistical approaches to validity, where modern statistical approaches are more sophisticated than what he addresses. The second exception is that, whilst his discussion of reliability is sophisticated, he uses empirical studies of the 1970s and beforehand to highlight his claims, whereas today we have much more empirical information about how psychiatrists employ diagnosis.

One possible reasons for the lack of progress is that issues over validity, reliability and categorical really were not so controversial once the DSM III framework was adopted. If you want a reliable categorical system which easily lends to testing for disease entities then DSM is pretty good at this. The problem, and why lack of progress is so concerning, is that after about 40 years of testing categorical classifications for disease entities we simply have not found them. Existing DSM classifications have not been validated. If 40 years ago people said ‘lets see if there are disease entities out there’ then I think the DSM and validity project have done a reasonable job of testing this. However, since the answer came up as generally no, we really need develop something new, something outside traditional approaches. Kendell does discuss dimensional systems and alternatives to disease entities but concludes these are usually difficult to put into practice. So either we need work out how to put them into practise or we need some new developments, something beyond disease entity non-disease entity dichotomy, beyond valid invalid dichotomy. The same may be true for categorical vs dimensional, and even the notion of reliability may need challenging.

Classifications and medication

Washington Post headline “Most antipsychotic drugs prescribed to teens without mental health diagnosis, study says” (from article). From the article: “60 percent of those ages 1-6, 56.7 percent of those ages 7-12, 62 percent of those ages 13-18, and 67.1 percent of young adults, ages 19-24, taking these drugs had no outpatient or inpatient claim indicating a mental disorder diagnosis”.

The implicit notion here seems to be one of surprise and concern: surely people should have a diagnosis before they are given drugs. Interestingly, many philosophers of psychiatry might disagree. Philosophers of psychiatry often complain that most people do not neatly fit classifications, typically they meet the diagnostic criteria for numerous classifications, they exhibit symptoms below the threshold of criteria some classifications, their behaviour changes over time so they no longer fits previously diagnosis, the cut off points required for diagnosis are arbitrary conventions, etc. In essence, many philosophers reject the current categorical approach (you have either have this diagnosis or you do not) and instead prefer a dimensional model (you lie on a continuum with normality, exhibiting a range of symptoms to varying degrees [this is somewhat of a simplification]).

A consequence of the dimensional approach is that classifications are not so important. On a dimensional approach a doctor needs know what symptoms you exhibit, not which classification you does a half-hearted job of describing your symptoms. Surely then prescribing medication without providing a diagnosis is perfectly acceptable.

I personally prefer a categorical approach and plausibly this issue over drugs shows an advantage of categorical classifications. You need meet a set of fairly stringent criteria to get a diagnosis. On one hand this means some people who have some but not all the symptoms go undiagnosed, on the other it means diagnosis cannot be (should not be) handed out to all who want it. If a diagnosis is generally required to prescribe strong medication then we have an additional safety net above and beyond the judgment of a single doctor prescribing drugs since diagnosis usually requires a specialist. Of course, there are times when someone needs medication immediately and cannot wait for diagnosis, but surely that situation would not cover half of all adolescents?

Making psychology more textual but still statistical

I just read a couple of textbooks on anxiety. I was looking to see if anxiety was ever described as manifesting in a particular way and I found both textbooks to be largely unhelpful. They were severely lacking in qualitative descriptions of patients, very little on the experience of anxiety. Instead it is mainly statistical, describing statistical correlations between anxiety and gender, age, upbringing, other psychological traits etc.

I was particularly concerned by how anxiety was demarcated from other psychological states (depression, worry, fear etc) and different types of anxiety were demarcated (social, agoraphobia, phobia, PTSD, etc) without employing much qualitative descriptions by patients. Without such information are we really sure the boundaries between, say, anxiety and depression have been adequately drawn, similarly so with boundaries between different types of anxiety. Psychologists certainly have criteria, checklists and questionnaires to make such distinctions on but how much trust should we put in them? To be fair, both textbooks were aware of this problem but unfortunately seemingly suggested more studies using those same criteria would solve the problem, rather than trying think about and improve the criteria.

If we want to get clear about something then you might consider asking a philosopher to think about it, to analysis the concepts, to look for assumptions. All well and good. However, I also think this approach is severely limited, since philosophers have a tendency of making distinctions on obscure grounds little connected to everyday experience. Also, it is difficult to put diverse human experience like anxiety into necessary and sufficient without risking turning the concept of anxiety into some abstract notion far removed from human experience. In my experience, sometimes conceptual analysis of psychological states results in something largely unrecognisable. The psychologists ‘anxiety’ I can relate to, this will not always be the case with the philosopher’s ‘anxiety’.

There may be a middle ground. Conceptual analysis is important but can only go so far, at some point concepts should be opperationalised and need applying to the world. Philosphers should ideally give their concepts formalised criteria and then do statistical studies on them. Statistics hopefully should show the concept has some ‘fit’ with the world and the concepts can be modified in light of those statistics to get the fit better, to make it more relevant to patient reports. Unlike the psychologists though, make the opperationalised criteria rich and textual, highlight them with patient case studies, analyse those case studies conceptually to draw boundaries between concepts and draw inter-relations between concepts. More importantly, recognise these are only concepts and recognise that concepts rarely (perhaps never) fully mesh with the world, so employ many different concepts, formulate many different opperationalised criteria for statistical studies rather than continually employing the same old criteria. A psychology which willing drew different boundaries between anxiety and depression in different studies would have greater difficulty comparing results of different experiments but those results themselves would likely be so much more richer empirically, the statistics would reveal so much more.

I’m not sure how closely they can be merged or how exactly how it would be best done (major institutional changes would be ideal!) but the qualitative and the statistical can be brought closer together and this may require both psychologists and philosophers learning from one another.