13 February 2012

Will DSM-V Overpathologize Children?

The chair of the DSM-IV task force (Allen Frances, M.D.) argues that his successors in crafting DSM-5, the latest edition of the standard American psychiatric diagnosis handbook (with its publication planned for May 2013) are making it too easy to classify ordinary childhood behavior as mental conditions. This is a charge he believes that his own draft of the DSM (published in 1994) was itself guilty of in the case of ADD, autism and (indrectly) childhood bipolar disorder, despite their best intentions. He argues that this could lead to overmedication of children and new "epidemics" of childhood mental illness driven by weaker diagnostic criteria. But, he claims that these decisions are driven not by corruption from drug companies, but from insufficient cynicism about what is at stake in these decisions in the people making them.

Of particular concern, in his view, is the new set of diagnostic criteria for Disruptive Mood Dysregulation Disorder (i.e. temper tantrums). He writes:

A prudent DSM 5 would tighten its criteria for ADD and put in a black box warning against the blatant current off-the-DSM-label diagnosis of childhood bipolar. DSM 5 instead does everything wrong it possibly could with ADD and then remarkably takes the mischievous further step of adding yet another new candidate for diagnostic fad (Disruptive Mood Dysregulation Disorder) likely that will increase the already scandalous overprescription of dangerous antipsychotic medication to children.


I am ambivalent. On one hand, there seems to be considerable evidence that there are many people who exhibit a variety of mental conditions at subclinical levels (i.e. without the level of impairment in major life activities necessary to qualify for a diagnosis) who would might benefit from access to low levels of the medications provided under a medical doctor's supervision, that those who meet current clinical criteria for diagnosis receive. On the other hand, many of the diagnostic criteria of both DSM-IV and DSM-V are mushy at best, are not applied consistently in practice, and sometimes seem to have a rather attenunated connection to clinical evidence.

Some of the language is weird. For example, the current proposal includes a "Attention Deficit/Hyperactivity Disorder" subtype for cases without hyperactivity, which would be a bit like calling major depression, "bipolar disorder without manic symptoms." But, psychiatrists can be weird and it doesn't really matter when one gets down to brass tacks.

Conditions like antisocial personality disorder (and its twin, juvenile "conduct disorder") seem overbroad pathologizing anyone who was a persistent juvenile delinquent without distinguishing between behavior rooted in relatively well articulated and specific constructs hypothesized to be inherent in a person's psychology, like psychopathy, and persistent juvenile delinquency with other causes. The draft DSM 5 appears to recognize that this broad classification is really capturing at least a couple of separate dimensions of symptoms but still has rather doubtful rigor and specificity. Most members of the general public aren't comfortable with concluding that everyone who acted criminally as a juvenile and didn't go straight as an adult has a mental health condition, and the evidence in support of this position isn't overwhelming when you look at it.

Similarly, there has been considerable doubt about the notion of treating grief resulting from the death of a loved one as no different from any other kind of major depressive disorder.

The degree to which the DSM-5 is mushy is illustrated by the weak thresholds the body proposing it suggests for diagnostic reliability.

A January 12, 2012 press release on Reliability and Prevalence states that:

DSM-IV diagnostic prevalence in comparison with DSM-5 prevalence for all disorders, information on the rates of DSM-IV diagnoses in each of the clinical settings [used for field trials of new DSM-5 criteria] was collected, as well as prevalence rates for the newly assessed DSM-5 diagnoses. There does not appear that there is a substantial difference in rates between DSM-IV and DSM-5, and in fact rates of DSM-5 diagnoses appear to be lower on average. . . .

Reliability is measured with the kappa statistic, which indicates the level of agreement between the clinicians making the diagnosis. Kappa ranges from 0 (no agreement between diagnosticians) to 1 (perfect agreement). There is no single agreed-upon standard for evaluating scores between 0 and 1. Studies in the rest of medicine were instructive in developing reasonable interpretations for the DSM-5 field trials before the results were available.

Kramer et al. state, “for interrater reliability, in which two independent clinicians view, for example, the same X-ray or interview, one occasionally sees kappa values between 0.6 and 0.8, but the more common range is between 0.4 and 0.6. For instance, in evaluating coronary angiograms, Detre et al. reported that ‘the level of observer agreement for most angiographic items (of 15 evaluated) [was] found to be approximately midway between chance expectation and 100% agreement’ (i.e., kappa around 0.5).

Test-retest studies are less frequent: the diagnosis of anemia based on conjunctival inspection was associated with kappa values between 0.36 and 0.60, and the diagnosis of skin and soft-tissue infections was associated with kappa values between 0.39 and 0.43. The test-retest reliability of various findings of bimanual pelvic examinations was associated with kappa values from 0.07 to 0.26 (6).

From these results, to see a kappa for a DSM-5 diagnosis above 0.8 would be almost miraculous; to see κ between 0.6 and 0.8 would be cause for celebration. A realistic goal is kappa between 0.4 and 0.6, while κ between 0.2 and 0.4 would be acceptable.”

No one is really satisfied with a kappa between .2 and .4, even though standard medical diagnoses do fall in this range. When a DSM-5 categorical diagnosis falls in this range, it will be seriously be reconsidered by the appropriate Work Group. However, when a disorder is rare, or the signs or symptoms are inconsistently expressed by patients with the disorder, it may be that in absence of a biological test or repeated measures over time, a kappa between .2 and .4 may be as good as can be done using only a single clinical interview.


In order words, the authors of the DSM-5 accept a situation where to psychiatrists chosen at random will agree on the diagnosis of a patient under the criteria only 40% of the time, and that in some circumstances will tolerate a situation where a diagnosis of two psychiatrists chosen at random that someone has a condition will agree only 20% of the time. This is troubling enough in cases of physical ailments where there is some kind of objective underlying physical state of a body that is being evaluated with a fairly well understood cause and effect relationship to set of symptoms and consequences, but is more troubling in the case of mental health conditions, where the underlying mechanisms of the conditions are conjectural and the range of diagnoses which two diagnostic professional could reach from the same data is much wider.

There may be a kappa of only 50% in evaluating fifteen specific aspects of coronary angiograms, but nobody with the proper qualifiations is going to conclude from a coronary angiogram that somebody has a broken leg, and the overall diagnosis from a coronary angiogram is likely to have a much higher kappa when presented with the big picture question of whether or not there is something wrong with someone's heart. It isn't at all obvious that the kappa figure for psychiatrists applying either DSM-IV or DSM-5 reflects nearly that level of overall diagnostic coherence.

Also, the DSM-5 will almost surely not disclose, in the body text, the level of reliability for diagnosis of particular conditions so that casual practioners are forewarned about the extent to which a diagnosis is mushy, nor will it have any data in the body text entries for conditions indicating the prevalence impacts of particular conditions relative to their prior diagnostic criteria. The authors of DSM-5 claim that past, much higher estimates of reliability using DSM-III and DSM-IV were a product of bad methodology in making those estimates rather than better defined criteria.

Major recent decisions in the drafting of the DSM-5 are summarized here.

No comments: