Tuesday, 14 May 2013

The Dismal and Sloppy Manual of Madness


The Diagnostic and Statistical Manual is neither diagnostic nor statistical, but it is a godsend to any insurance-based health system, because every patient must have a diagnostic number before they can claim their money back, and every diagnosis requires a medicine, which makes the manual attractive to pharmacological companies. The recent controversy about the 5th revision is in some ways a compliment, because it shows that it has established itself as a major reference work.

The problem is that the manual has always been a muddle of conflicting opinions, often ill at ease with itself. It is not based on a statistical approach to mental health problems arising from a proper epidemiological population study, but rather to a committee-based discussion about the common constellations of problems encountered by psychiatrists in clinical practice. Because psychiatrists cannot always agree, the manual follows a pick-and-mix approach to diagnosis, in which all you need to have is 1 or 2 symptoms out of 4 or 5 in each category, such that patients can attain the diagnosis with a very different pattern of symptoms.  In line with American productivity, the list has expanded generously from one edition to the next, (see Icebergs and Onions) as if madness were increasing, a sort of Flynn Effect of lunacy.

It is not all rubbish, because psychiatrists can notice patterns like any other professionals, but it tends to generate many over-lapping descriptions, and a proliferation of categories, sometimes under pressure from patient groups, who see the whole exercise as a matter of obtaining legitimacy, whilst others campaign to have themselves removed from the list, such as homosexuals who do not see that their sexual preferences constitute a disorder.  

Knowing that they would not be able to agree on a common theory, the participant psychiatrists have ended up as butterfly collectors exhibiting an assembly of brightly coloured disorders, each with a proud catalogue number. One particularly gaudy specimen is “disorders not otherwise specified”, a rather common Lepidoptera, it would seem.

How should a psychiatric/psychological disorder be defined? Four tests provide a useful start.

1          The diagnosis should have internal coherence, with the relevant symptoms having positive predictive value one to the other. Each symptom should be more strongly correlated with the other symptoms in the diagnosis than they are usually correlated with other extraneous problems. One positive symptom should predict the other positive symptoms, such that they all occur together more frequently than in other disorders.  (In statistically terms, there must be high correlations between most of the symptoms, such that you could get a high R squared in a multiple regression equation).
2          The diagnosis should be distinct from other groupings, with minimal overlap. Some aspects such as depression or anxiety will be very common, but there must be something distinctive if the disorder is to have a separate category name.  (In statistical terms they should be distinct using cluster analysis, or discriminant function analysis, all of which should be apparent using old-fashioned t tests between the group symptom totals).
3          Ideally, there should be some therapeutic or research advantage to the classification. In the absence of clear biological indicators, diagnostic categories have to be provisional, but they can often be clinically useful.  (In statistical terms the diagnoses must predict different outcomes and responses to treatment, and reduce errors of prediction as regards outcomes).
4          It should not be an arbitrary sub-set which leads to tautological self-confirmations. By cherry picking your cases you can create many apparent syndromes. For example, by studying only those cases of autism in which the parents believe that the condition was brought about by a vaccination, you can quickly gather suggestive data implying that there is a syndrome of vaccine damaged children with resultant autism. (In statistical terms the samples have to be drawn randomly from the population and the symptoms and putative causes must be considered objectively for cases and non-cases).

Here is a list of generic problems with DSM which I produced some years ago:

1.      Neither diagnostic nor statistical, just a stamp collection
2.      No cluster analysis, so disorders are poorly differentiated and poorly supported
3.      Does not give statistics of symptom frequency for each diagnosis
4.      Categorical, not dimensional, which is against the evidence.
5.      Too many disorders, many of them overlapping
6.      Diagnostic categories are over-inclusive and pathologize ordinary life
7.      Not purely scientific, but manipulated by interest groups

We are now awaiting the 5th edition of DSM. The British Psychological Society has got its retaliation in first http://apps.bps.org.uk/_publicationfiles/consultation-responses/DSM-5%202012%20-%20BPS%20response.pdf

My own view is that the proper way to look at psychological problems is to study the mental health of entire populations. When you do that you usually find dimensions rather than categories. Take depression. Most people have low mood sometimes. Some people have low mood more frequently. A minority have episodes of extremely low mood, and at those times are very vulnerable, and at risk of self-harm.  Psychologists (and quite a number of psychiatrists) prefer the dimensional approach. They argue that there are no common disorders, only extremes of quantitative traits.

However, although the dimensional approach makes most sense most of the time, some conditions may well represent a discontinuity. For example, all of us may feel suspicious about how we are being treated from time to time (particularly within large institutions) and this might seem no more than understandable mild paranoia. Not every promotion is achieved by the best candidate. Suspicion is natural, and sometimes justified. Yet the level of suspicion in severe paranoia is of a different order of magnitude. It corrodes the sufferer’s ability to live a productive life. Perhaps there is a real paradigm shift in such cases. The category of paranoid schizophrenia may stand up as a category. It is very different, and different from the shades of depression and anxiety which commonly affect most people.

At a more basic level, it would be good to have some theoretical basis for diagnosing mental disorders. A diagnosis of schizophrenia usually is associated with a sharply reduced level of contribution to society, and a fourfold increase in violence. A diagnosis of anorexia nervosa is associated with increased mortality, with about 10% of sufferers dying young.  Depression and anxiety at the extreme levels of the trait are very handicapping, and severe depression is associated with suicide. Homosexuality is not currently a diagnosis, but it is associated with sharply reduced fertility. Which, if any of these outcomes, makes us feel that something has gone very wrong with the person in some important regard?

In the meantime, it is very likely that your own personal problem has a number, and that the latest edition has been written purely to classify you.  The tablets are in the post.

No comments:

Post a Comment