The notion of hierarchical diagnoses comes up from time to time, so it is about time that I made some comments about it. Like most criticism of psychiatry, I think the concept is overblown and will illustrate why that is. The idea of hierarchies in diagnosis is basic. It means prioritizing the most clearly defined illness. Clearly defined in this case may mean a clearcut phenotype or disease mechanism. In the psychiatric examples provided that generally means a feature high in the hierarchy that is not replicated at lower levels, but many features at lower levels that overlap.
The overall goal is to not end up with a long list of
conditions or diseases that could fit the presenting problems and end up
treating everything on the list rather than the most likely cause of the
problem. I think of it as second year
med student differential diagnosis. When
you are learning physical diagnosis you take an exhaustive history, review of
systems, and physical exam and try to come up with a differential diagnosis
list. Even with that beginning level of
knowledge in medicine you start to realize that diagnosis number 3 to 8 are
improbable and start to focus just on the top 2. By the time you are an intern
those lengthy differential diagnoses lists are a remote memory.
It also involves the application of the parsimony
principle. Is there a single diagnosis
that incorporates all the features of the observed disorder rather than a list
of conditions? I can illustrate this
with an example right out of my physical diagnosis text from medical school (1):
“In selecting the diagnosis from a list of hypotheses, the
physician is to choose a single disease to explain all of the patient’s
manifestations rather than explain them by the coincidence of several
diseases. This is called the law of
parsimony. For example, when a patient is found to have a dilated heart,
hepatomegaly, ascites, and pedal edema, the single diagnosis of cardiac failure
explains all of the findings. It is more likely to be the true diagnosis than
to suppose a coincidence of heart disease producing cardiac enlargement,
cirrhosis of the liver producing hepatic enlargement, and nephrosis leading to
edema from hypoproteinemia.
Nevertheless. The “law” must be applied cautiously. The experienced clinician realizes that in
the process of aging the patient accumulates the more debilitating conditions
the longer he lives, and his demise is often accompanied by a combination of
several diseases.” (p. 3-4)
And from a comparable section of DSM-II (2):
“The diagnostician, however, should not lose sight of the rule
of parsimony and diagnose more conditions than are necessary to account for
the clinical picture. The opportunity to make multiple diagnoses does not
lessen the physician's responsibility to make a careful differential diagnosis.”
This excerpt on strictly physical diagnoses of
non-psychiatric conditions illustrates a couple of points. First is that disease states in general do
accumulate over time. Some are more
likely to occur during different time frames during a life time. The age at onset is a relevant concept but
even then, there are exceptions. Secondly, transdiagnostic symptoms (covering
many diagnoses) are common in the physical world. In the example edema and
ascites are the obvious example but even the gross organ findings can be
considered transdiagnostic signs. For
example, there are hundreds of possible causes of cardiomegaly and
hepatomegaly, even though from a clinical standpoint the majority are not all
fully investigated. Once heart failure
develops it is treated as a syndrome without a specific pathophysiological
cause. That illustrates a third point in
the diagnostic process and that is the triage aspect. Whatever is acute and life threatening gets
priority and is diagnosed and treated first. I will illustrate how all these
concepts apply directly to psychiatric disorders.
Hierarchies are thought to be more important in psychiatry
because of symptom overlaps as well as certain diagnoses being more important
not to miss. A common problem is the overlap of anxiety and depression. Some
people have long history of both disorders concurrently or one morphing into
another. Ghaemi (3) has suggested that mood disorders should be prioritized
over anxiety disorders in the hierarchy.
He stresses the importance of diagnosis in this process and suggests
that treatment of depression in this case would be the priority. He also
briefly reviews some evidence in diagnostic hierarchies that basically show
that they can be arranged so that diagnoses at the top of the hierarchy would
contain the symptoms of every lower level in the hierarchy. For example, if
bipolar disorder with psychosis is at the top of the hierarchy those patients
would also have the symptoms seen at defined lower levels in the hierarchy like
unipolar depression, schizoaffective disorder, schizophrenia, anxiety
disorders, and personality disorders. He
claims that there is a current emphasis to make every possible diagnosis rather
than a hierarchical approach, but does acknowledge that some psychiatrists
“intuitively practice this way.” (p. 223).
Treatment setting is an obvious factor – with acute care psychiatrists
seeing clear presentations of the more severe forms of schizophrenia, unipolar
depression, and bipolar disorder. They
would be most likely to treat these acute forms and the pharmacotherapies are
very similar.
It is always good to remember the task of the psychiatrist
in all these academic exercises particularly when they are coupled with
criticism of the psychiatrist in the field. What is the task of that
psychiatrist? That task varies with
setting but after 22 years in acute care, the job is to recognize the acute
illness and treat it while keeping the patient and the staff safe. How well does a bipolar disorder->unipolar
disorder->schizophrenia-> obsessive compulsive disorder-> anxiety
disorder-> personality disorder->attention deficit~hyperactivity disorder
(BUSOAPA) hierarchy hold up in that setting?
Probably not very well. That psychiatrist is confronted
with an entirely different hierarchy. The first syndromic level encountered:
“Is this patient medically stable? Are these symptoms I am seeing
manifestations of an intoxication, withdrawal, or secondary medical condition?”
I routinely encountered these cases and
often had to send them directly to the intensive care unit for both acute
conditions threatening the life of the patient and acute conditions creating
the psychiatric symptoms. The second syndromic level: “Is this patient
responsive to me during the interview?
If not, are they delirious, catatonic, or is there another psychiatric
reason? Did I miss an acute medical
condition like a stroke and aphasia at the first syndromic level? The third syndromic level: “Is this person
able to produce an accurate history that I need to make a diagnosis and
formulate a treatment plan? Does the history that they are giving me sound
plausible?” If not why not why - and
what needs to be done? Do I need to
gather a lot of collateral history to get the full picture?
These considerations place the BUSOAPA hierarchy at the
minimum as a hierarchy within a hierarchy that contains about 40 or 50
psychiatric diagnoses and many more if all possible medical diagnoses
masquerading as psychiatric diagnoses are considered. The diagnoses also have
clear implications for Ghaemi’s concern that a hierarchical model would reduce
misdiagnosis and polypharmacy. Patients
with alcohol use disorder (AUD) can experience psychotic symptoms and mood
symptoms. I have seen them misdiagnosed
as having schizophrenia and bipolar disorder. Does that mean AUD should be at
the top of the hierarchy? Placing it
there would skew the data because most people with AUD do not have those severe
symptoms, but on the other hand if they were misdiagnosed with schizophrenia or
bipolar disorder – appropriate treatment with benzodiazepines for alcohol
withdrawal may be held to the detriment of the patient.
A recent exercise was sent to me to see if it suggested a
need to modify DSM criteria for major and mild neurocognitive disorder. The patient was described as having Alzheimer’s
disease, cerebrovascular disease, an HIV infection, and heavy chronic alcohol
use. They had symptoms of
psychosis. They were coded as having
major neurocognitive disorder due to all the listed etiologies. The problem is that the current diagnostic
criteria use a hierarchical approach and state that for probable Alzheimer’s
Disease in both the major and mild neurocognitive disorder categories the mixed
etiologies need to be ruled out or eliminated. That is currently difficult to
do on a clinical basis and will lead to more uncertainty in the Alzheimer’s
Disease diagnosis using these criteria. In terms of the BUSOAPA hierarchy – it
is an argument to put the neurodegenerative disorders diagnoses at the top
since they frequently contain most of the symptoms of psychosis, mood
disorders, and anxiety while having unique cognitive profiles that would not be
seen at lower levels.
Getting back to Ghaemi’s original argument for empirically
studied hierarchies in psychiatry to reduce misdiagnosis and improve treatment
by reducing polypharmacy – is that likely? They might work at the level of
nomenclature only. Hierarchies might
work in highly selected environments with low acuity patients. I am thinking about an outpatient psychiatric
teaching clinic. It might be easy to illustrate how the patient population matches what is happening hierarchically. In other settings looking at higher levels of
acuity – the acuity becomes the hierarchy. Residents in acute care should learn
almost immediately that pharmacotherapy in the inpatient setting needs to be
directed at the likely acute diagnosis.
Even then that clearcut diagnosis can be obscured in the outpatient
setting. The best example I can think of is women with postpartum bipolar
disorder +/- psychosis who are stabilized and eventually readmitted with
diagnoses of schizophrenia or schizoaffective disorder. In that case, the
diagnosis was in the top spot of the BUSOAPA hierarchy based on the direct
observations of the inpatient psychiatrist, but it was modified on an
outpatient basis by staff who did not directly observe the acute manic episode.
A violation of the hierarchical model to be sure, but also poor continuity of
care and ignoring what happens in acute care.
That brings me to the issue of hierarchies in previous versions
of the DSM. I will focus on DSM-II because it is the most clearcut. The explicit hierarchies in DSM-II included both
acuity and severity. Specifically:
“1. The condition which most urgently requires
treatment should be listed first.”
And:
“2. When there is no issue of disposition or
treatment priority, the more serious condition should be listed first.”
And:
“It is recommended that, in addition to
recording multiple disorders in conformity with these principles, the
diagnostician underscore the disorder on the patient's record that he considers
the underlying one.”
These three sentences from DSM-II capture what appears to
be the concern about diagnostic hierarchies in subsequent versions of the
DSM. They easily map on to what happens
clinically as depicted in my diagram at the top of this post. It is also consistent
with the general approach to all medical diagnoses and the training of
psychiatrists.
A long list of diagnoses is a rookie mistake. Experienced
psychiatrists are generally trying to address the main problem and the urgent
problem. There is some evidence for that in the
numbers of diagnoses that are used in clinical practice and those
numbers are considerably lower than the usual DSM-5 count. That does not mean
that polypharmacy will not be involved, but it does mean that any polypharmacy
used is there to address the main problem and not multiple separate diagnoses
all at once. The concepts of comorbidity
and transdiagnostic symptoms are discussed these days like they represent new
ideas. Any physician trained in a medical school and residency program knows
about both starting with medical and surgical diagnoses and progressing to
psychiatric diagnoses. The general concepts are the same. The necessary
evaluation in psychiatry should adequately reflect the complexity of the
situation and that means a detailed longitudinal history and that can include
disorders that are commonly viewed as symptoms – like primary insomnia.
A detailed longitudinal history may not lead to a correct
current diagnosis until the symptom patterns change. The best example in the
case of a pure psychiatric disorder is bipolar disorder. Goodwin and Jamison (4) make this point in
their discussion of false unipolar disorder – people with recurrent unipolar
depressive episodes until the first episode of mania occurs. In their table
they show that according to three different longitudinal studies, patients experience
1 to 5 episodes of unipolar depression without a manic episode and they
constituted a significant portion of the unipolar sample. A hierarchical rule in this case results in
misclassification until a manic episode occurs.
A strictly hierarchical diagnosis will also not capture
clinic reality. It can potentially lead
to catastrophic results. The best
example I can think of is schizophrenia and depression. If you consider
schizophrenia in a hierarchy independent of depression, you will miss the
opportunity to treat serious depression in patients with a schizophrenia
diagnosis. It is also a case of seeing a
new patient treated for both diagnoses and considering what medication to
stop. The features of schizophrenia are
such that the presentation of depression is very subtle and medication changes
of antidepressants and antipsychotics should only be changed with extreme
caution and after adequate collateral history has been obtained.
Criticizing DSMs is a popular American sport. That has resulted in elevating the DSM to levels that are really not consistent with the way these documents are viewed by psychiatrists. The documents are generally limited by the fact that they are not treatment manuals and do not incorporate considerable amounts of research that could provide guidance in these areas. With a document that is limited to nomenclature – any changes including hierarchies will result in a loss of information at another level. To me this seems like an endless exercise in trying to reduce uncertainty to an unachievable level.
That is why
psychiatrists need to be trained and don’t result from reading a manual.
George Dawson, MD, DFAPA
References:
1: DeGowin RL. Bedside Diagnostic Exam. 3rd ed. Macmillan Publishing Company, New York, 1976:
3-4.
2: Diagnostic and
Statistical Manual of Mental Disorders. 2nd ed. Washington, DC. American Psychiatric Association. 1968: 2-3.
3: Ghaemi SN. The concept of a diagnostic hierarchy. In:
Ghaemi SN. Clinical Psychopharmacology:
Principles and Practice. New York. Oxford University Press. 2019: 222-230.
4: Goodwin FK,
Jamison KR. Manic-Depressive Illness.
New York. Oxford University Press. 2009:
66.
