On Psychiatric Diagnostics
Another defense of an imperfect system + my confidence weighted diagnosis dream
I
There are two reasons why treatments don't work. First, the treatment isn't that good, and it fails. Chemotherapy often helps with cancer, but sometimes it isn't enough. Alternatively, maybe the treatment isn't working because the diagnosis is wrong. The problem is not “treatment resistance” but diagnostic invalidity.
Invalidity comes in two forms. Maybe your diagnosis is wrong (you've selected the wrong option from the list of options). Or maybe the whole system is flawed (none of the available options are correct). Maybe you are using the wrong level of analysis. Maybe nature was incorrectly carved at its joints.1
How should we carve up the landscape of human experience into pathological boxes? Currently, the DSM sits on a throne titled ‘the most commonly used way to describe and cookie-cut apart this landscape’. Some people really don’t like it.2 In their defense, it is politically influenced (and maybe that's not too bad? Or for more formal writing, see here) with disorders voted into or out of existence by committees (e.g., disruptive mood dysregulation disorder and homosexuality).
It’s also self-propagating. Healthcare insurance companies rely on it to determine what they will include in their ‘we care about mental health so choose us’ reimbursement list. If people want to get paid for working in this space, they can either accept cash, which many don’t have. Or, regrettably, insurance. To accept insurance, you have to play by their rules and use the DSM labels. It’s not just insurance. What lexicon will all the research be made under if it wants large institutional funding? I bet you can guess. In this manner, the DSM pulls itself up by its own bootstraps; we use it because it is popular, and it is popular because we use it. Had another system rose to fill this niche earlier, it too would be in power while holding its bootstraps. If there were anti-trust laws in this space (I don't think that works or should be seriously contemplated), they would be blaring alarms.
However, none of this makes it inherently wrong.2 The DSM is a good place to start as far as a psychiatric diagnostic evaluation goes, but it would be a mistake to get enamored by its definitional labels. I’ve never been a fan of labels. Labels are shorthand terms for concepts. We do not derive information from labels. We simply use them to lighten our descriptive load.
II
We are not big fans of the Diagnostic and Statistical Manual now in its fifth edition (DSM-5) or the International Classification of Diseases now in its tenth revision (ICD-10). Both of these classification systems are based on symptoms, not actual neurophysiology. It is all self-reported. Therefore patients who are stoic are likely to underreport symptoms, and hysterical patients are... well, all over the place. The former may not receive a diagnosis, and the latter receives too many. Neither gets a diagnosis that accurately reflects what occurs in the brain.
- The Neuroscience of Clinical Psychiatry, Higgins et. al
We’ve established that psychiatric nosology and diagnostics are nuanced. In fact, the DSM includes a diagnosis of “not otherwise specified” (NOS) that was meant to capture such instances. As long as outward symptoms are used to classify psychopathology, our map will be crude.
Not that long ago in the field of neurology3,it was not uncommon for a woman with ambiguous neurologic symptoms to be diagnosed with conversion disorder or hysteria before it was recognized months or years later that she actually had multiple sclerosis and autoimmune originating lesions on her spinal cord and brain. With modern brain imaging, that mistake can be avoided.
In psychiatry, we don’t have objective measures (lab tests, medical imaging, crystal balls, etc.) like in the case of Multiple Sclerosis to diagnose mental health conditions. Instead, we have a checklist of symptoms and timelines and rely on the patient’s story (history) and observation (behaviors, cognitive status, etc) to figure out if there is a match.
The latter is sort of objective. The former is a shot in the dark. A combination of these limits plus the DSM's ‘checklist system’ of diagnosis make diagnostics challenging.
III
Every job has a part of it that is not highly advertised. A part that those in the career know all too well, but is skipped out on during their Hollywood representation. Computer programmers aren't typing abnormally fast and hacking into things in the command prompt with all the lights off while drinking Red Bull (mostly). They're googling how to fix random bugs and talking to rubber ducks.
If you go to a psychiatrist, they won't be deeply contemplating your relationship with your mother (mostly). They're thinking about how much collateral they have to call that day. In the hospital, a common conversation is as follows:
Doctor: It’s nice to meet you. I know you’ve already spoken to a lot of people today, but would it be ok if I go over some details with you? It helps the team better understand your situation and how to best help.
Patient: Sure.
Doctor: I heard about your suicide attempt and the things going on in your life. You’ve clearly been under a lot of stress. One of the things we screen for is something called mania. It's the opposite of depression in many ways - people will go days with no sleep, talking really fast, have lots of energy and constantly be moving like they have a motor attached to them. They have grandiose ideas and do things out of character that are impulsive. Sometimes they even hear voices talking to them. Has anything like that ever happened to you?
Patient: No.
Doctor: OK. Is it okay if I speak with your family?
Patient: Sure.
...
Family: oh yeah. She was admitted to the hospital for that a few months ago. She thought she had the cure for world hunger in her notebook that she spent all day scribbling in. She paced and paced and never slept for days on end. She maxed out her credit cards buying a 3d printer and supplies that she had no idea how to use. The doctors told us she had something called mania.
...
Patient: oh that? Yeah I was in the hospital for a few days a few months ago, but it was a misunderstanding.
Diagnosis is challenging to begin with. If you have inaccurate or incomplete data to start with (the person has poor contact with reality, fears stigma, harbors embarrassment, or has secondary gain) then it is even more obscure. One might correctly ‘guess’ an accurate diagnosis in these circumstances and pat themselves on the back in retrospect. But doing so would gloss over something important. Confidence in a diagnosis should exist on a spectrum, not as a dichotomy.
Diagnoses should be malleable and slowly become more and more solidified over time. It can be true that the best diagnostic choice at Time 1 is diagnosis A, but at Time 2 the best choice would be diagnosis B. That outcome does not make a diagnosis of A at Time 1 a mistake. More information, which often comes with time, leads to better diagnostic clarity.
In theory, the dichotomous nature that I am lamenting against does not exist. ‘Of course, diagnoses can change over time’ the ivory tower academic retorts. But in the practical world. flexibility and transparency are not reflected well in modern healthcare structures. Instead, diagnoses auto-populate across encounters into note-bloat that people don’t read. Often, even doctors can't read actual notes from other hospitals because of their immense fear of litigation and respect for patient privacy. Instead, they go off of insurance billing claims to figure out what was diagnosed at the time. Other prescribers may have pressure to maintain a diagnosis they are less certain of because ‘maybe that person knew something I don't.’ Medications tend to get trialed, increased, adjuncted (bringing along all their side effects) breathing life into the subtle human bias of addition when problem-solving
My proposition to solve this is to include a confidence rating system with diagnostics. Particularly with things that are phasic and easily miscommunicated (such as the manic phase of bipolar disorder).
For example,
I am not confident Person A, with a stimulant abuse history, who won't provide collateral, and an EHR record of schizophrenia, MDD, and ADHD has bipolar disorder; but if he was given medications by his outpatient provider that people with bipolar disorder would have and says he has it, then… I guess they have bipolar disorder.
I am confident Person B, with no substance use history, who is in front of me displaying the signs of mania, whose mother just called and confirmed a timeline of this episode consistent with mania and previous hospital stays for such… has bipolar disorder.
With a rating system in place, you could track long-term outcomes and answer questions like: How often am I overconfident in a diagnosis? What factors play a role in contributing to diagnostic uncertainty that I should pay extra attention to? How does my judgment compare to my peers? A whole ecosystem of opportunities exists here, but instead, we (informally and crudely) gauge diagnostic uncertainty based on the number of diagnoses that have accumulated in the EHR.
The only surefire way to diagnose bipolar is to witness the manic episode yourself. Second to that, at least give me a confidence-weighted spectrum.
Modern psychopharmacology is fancy. The neuronal pathways, pharmacokinetics, and pharmacodynamics for various substances have been mapped out. As a society, we use some of these intentionally as medicine. What lies below is not a detractment from those advances. But a reminder to not forget the fundamentals. That is, diagnostics come before treatments. There is no point in treating ‘something’ if that something isn't present
Entire careers have been dedicated to this subject. I can’t cover all of its nuances. But I assure you it is nuanced.
I have yet to find a satisfying explanation for having neurology and psychiatry separate other than history.