Hickam's Dictum: Let's Talk About These Many Damn Diseases

This post is about our article on Hickam's dictum, just published online (open access!) today.

I don't know if I read the 2004 NEJM CPC that mentioned Hickam's Dictum (HD - "A patient can have as many diseases as he damn well pleases") and popularized it, but knowing me, I probably did. My interest in HD piqued over the past 5-10 years because it has been increasingly invoked in complex cases, and whenever this happened, I always thought it was more likely that a unifying diagnosis (according with Ockham's razor) was present, and that the case had not yet been completely solved. So, we set out to investigate HD formally using three lines of evidence that you can read about in the article. We learned much more than we were able to report in the article because of word limitations, so I will report other interesting findings and insights gained along the way here (if you want to skip over the summary info to the "other interesting findings", scroll to the bold "other insights" subheading).

First, a summary of our results. As should be obvious, we confirmed that patients get multiple diagnoses, but case reports alleging to instantiate HD did not document random diagnoses occurring in a patient - there was a pattern to their occurrence. The vast majority of the time, there was a primary diagnosis which explained the patient's chief complaint and acute presentation, as well as one or more of the following: 

1. An incidentaloma (about 30% of cases)
2. A pre-existing, already known condition (about 25% of cases)
3. A component of a unifying diagnosis (about 40% of cases)
4. A symptomatic, coincident, independent disease, unrelated to the primary diagnosis, necessary to fully explain the acute presentation (about 4% of cases)

As we explain in the discussion, finding an incidentaloma during investigation of the chief complaint represents a spurious coincidence. Finding a new disease superimposed upon a chronic one and being surprised suggests that clinicians anchored to the chronic condition and forgot that new diseases can be superimposed on it; e.g., they failed to recognize that having recurrent CHF does not preclude development of CAP this admission. When authors report a primary disease and its complications, epiphenomena, or underlying cause, they appear to have failed to realize that those are all components of a unifying diagnosis! In the three (3!) cases we categorized as #4, we actually were being generous and honestly these probably represent one of the other categories but we didn't have sufficient information to confidently confirm that.

The take home message is that you should always look for a "unifying diagnosis" (see below for a definition of that; previously no attempt has been made to rigorously define unifying diagnosis) and if you can't find one, you should ask if some of the information represents an incidentaloma (or a false positive or red herring), or there are causal connections between observed findings that you are unaware of or haven't made yet.  And from the outset, you should realize that known, pre-existing diseases don't really count, because of course you can get new disease superimposed on old! (See the figure at the top of this post and at the bottom with a caption.)

Heretofore, people took OR and HD and their alleged conflict at face value, and failed to properly analyze and define them, taking for granted that each held some sacred and mysterious truth, leaving it to flummoxed diagnosticians to decide which one applies in a given case. Our framework disambiguates both axioms and shows that neither is sufficient to explain the occurrence of multiple diagnoses in a way that can guide diagnosticians in real time. Our framework, showing that timing, probability, and causation must be brought to bear on the diagnostic process obviates invocation of either axiom.

There is more to the article and discussion than this brief summary, so please check it out - we paid a $5000 open access fee so you can access it with no firewall! 

Other insights, not (completely) described in the paper owing to space limitations:

First is that John Bamber Hickam was very real. At the time of his premature death in 1970, he was chair of medicine at Indiana University, and the next year the Archives of Internal Medicine devoted an entire issue to him (reference 2), republishing seven of his articles that had previously been published elsewhere. I have never before seen that kind of homage paid to a deceased physician, and I probably never will again. Reference 1 of our article gives a first hand account of Hickam having coined his namesake dictum. So this issue is settled: Hickam was a famous physician in his time, and he alone was responsible for the dictum bearing his name.

Next, as alluded to above, is that everybody seems to have taken HD at face value -- "patients get multiple diseases" -- which is so self-evident as to render it meaningless. Of course patients get multiple diseases! Every past medical history of a patient over the age of 50 is testament to this obvious fact. But is there more to it? Is it the "answer to Ockham's razor"? What does Ockham's razor say about diagnosis, anyway?

Ockham said nothing about diagnosis, but at some point in the past 50-100 years, physicians began invoking OR to mean one should seek a single, simple, or "unifying" diagnosis. There is no agreement on this because nobody previously bothered to define exactly what OR means in relation to diagnosis. (The notion of OR guiding you to a single diagnosis seems most compatible with the notion of HD being a "counterargument" to OR; here, we debunked that and laid the groundwork for a better definition of OR in relation to diagnosis.) The use of Ockham's razor in medicine is often attributed to Sir William Osler, but we could find no primary documents from Osler's era supporting this presupposition. (If anybody can find first source material proving Osler invoked Ockham, please contact me.) We even went to the library and got a 100-year-old edition of Osler's textbook of medicine, and could find no mention in any index or relevant section on diagnostic principles, and then we found a digitalized version of Osler's textbook and performed an electronic ctrl-F search: no mention of Ockham or Occam or any razor was made in his textbook, making it unlikely that Osler was an ardent adherent of Ockham.

How should OR be applied to medicine? We dismissed "single diagnosis" because it is apparent that patients get multiple diagnoses. How about "simplest explanation"? This is too vague. There is no clear standardizable way to determine what's the simplest diagnosis in a given case, so I think we can dismiss that possibility too. In a commentary, we posited that Ockham's razor should lead the diagnostician to seek a "unifying diagnosis" and that it should be defined as a single causal pathway leading from an endpoint diagnosis to the chief complaint/acute presentation and the illness causing it, including all the causal associations, epiphenomena, and complications of that illness. Perhaps we need deeper philosophical justification for these choices, but here our task was to empirically evaluate it. Our definition of OR as "unifying diagnosis" is readily practicable: can you link everything together into a causal narrative? If so, and you can explain how A led to B led to C, D, E, you have a unifying diagnosis. This approach is consistent with the NEJM Case Records of the Massachusetts General Hospital and Clinical Problem Solving series, where the final diagnosis was always a unifying diagnosis, even if it had associated diagnoses causally linked to the primary diagnosis (as described in part 2 of our article).

We also recognized along the way that time is a critical consideration when thinking about multiple diagnoses. No simple adage can explain all the diseases a person gets during their entire lifetime -- this broad time frame is too wide. Disease onset is stochastic, with infinite combinations possible during a lifetime. So when thinking about multiple diagnoses and OR, we have to make a stipulation: a unifying diagnosis can only unify the symptomatic presentation of an acute disease; i.e., unity should only be sought for an acute presentation. So, healthcare screening at the PCP's office is off limits - of course you're going to find high cholesterol and HTN if you initiate screening! The formally undiagnosed osteoarthritis of the hip that grandad has been limping around with for years and that you "discover" when he presents for cholecystitis doesn't count, either. The goal of the diagnostic encounter must be centered on an acute, symptomatic presentation, with timing of symptom onset a boundary condition for the applicability of OR.

This analysis also explains away the so-called "Saint's Triad" (no relation to the first author of the 2004 NEJM CPC), which is the combination of hiatus hernia, diverticulosis, and gallstones that Dr. Saint, a South African surgeon commonly found in his patients. In the first published report of the triad, one gets the sense that the three women described presented with symptoms of just one of the triad's components, and the other two were discovered incidentally. That suspicion is reinforced by perusal of the second published report of Saint's triad, where the described patients clearly presented with symptoms owing to only one of the diseases, with the other two being incidental findings during routine diagnostic investigations. Thus, in most/all reported cases, two of the three diseases comprising Saint's triad are incidentalomas.

Three aspects of our classification scheme warrant clarification, some of which is provided in the article. First, we make no distinction between consequential and inconsequential incidentalomas. The essence of an incidentaloma is that it is asymptomatic and its discovery accidental, and our additional insight is that it represents a spurious coincidence (in the sense of co-occurrence) with the primary diagnosis. Even if retrospectively the finding of a consequential incidentaloma (e.g., renal cell carcinoma) was lucky ("I'm glad I got that pneumonia, otherwise they would have never found my kidney tumor!"), prospectively there was no intention or duty to discover it. 

Second, there is some overlap between incidentalomas and pre-existing disease. Consider grandpa's osteoarthritis of the hip that is "discovered" for the first time (by a doctor) when he is admitted for acute cholecystitis. Is this an incidentaloma? Arguably not because it is not asymptomatic, so even though it had not previously been formally diagnosed, we would classify it as pre-existing disease, our category 2. Using this logic, latent or subclinical tuberculosis, so long as it is asymptomatic, is also an incidentaloma if it is discovered accidentally, even though those cases are considered "prevalent TB" for economic calculations of the burden of disease on the population level. In a previous work, we showed that for reasons of diagnosis, a pivotal distinction must be made between incidence and prevalence. 

Third, we must make a necessarily arbitrary determination of what counts as a coincidence of symptoms, for the critical part of our analysis that deals with time. Too restrictive a definition (the symptoms of two diseases occurred on the same day or within the same hour) would exclude all coincident diseases, whereas too liberal a definition (symptoms occurred the same month or year) would include too many diseases as having cooccurred. We stated in the paper that we considered two diseases to be temporally coincident if they both had symptomatic onset "within the same two week period, or thereabouts". Further epidemiological or statistical modelling could shed additional light on what ought to count as a coincidence of two diseases.

In regard to coincidence, there is an awesome section in the discussion about Reichenbach's common cause principle, which is central to our analysis, and which to our knowledge has never before been discussed in the context of diagnostic decision making. The common cause principle states that "if an unusual coincidence between events A and B has occurred, it is likely either A caused B or vice versa (i.e., A&B are probabilistically and causally dependent), or that there is a common cause C of events A&B. I encourage you to read the discussion, as I cannot state it more clearly or succinctly here. Reichenbach's common cause principle, one of the philosophical justifications for Ockham's razor, provides the philosophical and logical backbone for our entire analysis.

Another corollary that we only touched on in the article is that there are some situations where you expect multiple downstream diagnoses, stemming from upstream common causes : common reservoirs (e.g., for respiratory pathogens, or sexually transmitted diseases - it is not surprising to have chlamydia and gonorrhea, right?) Immunosuppression or deficiency is another situation where we expect multiple diagnoses stemming from the immune defect. Consanguinity is another. Not mentioned is that the rarer the combination of diseases, the wider the time window that we may consider to represent a coincidence. So, the combination of MCAS and POTS, diagnosed 2 years apart, may represent an unusual coincidence, and we may expect a causal relationship. In the case of MCAS and POTS, the common cause is often a factitious disorder or other mental illness, leading unwary physicians to pile on diagnosis upon diagnosis in an attempt to explain otherwise inexplicable symptoms.

In one memorable case, I was called to see a young man with immunodeficiency and a poorly healing  wound who was having a rigor. Upon arriving at the patient's room, the resident accompanying me removed the sheet to examine the wound and this revealed a court ordered ankle monitor. Curious, I searched the chart for the "immunodeficiency" and the only documentation I found - some genetic CD-123XYZ deficiency - is poorly described with a single case report from the 1980s. It had not been tested for in this patient. So, we have a peculiar coincidence to explain. In this case, the unifying diagnosis was malingering with a self-inflicted and recurrently reinfected wound, to avoid a prison sentence.

I should also mention some examples that have apparently confused readers of the tweetorial I posted about our article. I think some confusion has arisen because people aren't actually reading our article, so they don't understand our framework. If you are admitted for X diagnosis, the doctors give you heparin, and you get HITS, the diseases are all causally linked this way: X-->admission-->heparin prophylaxis-->HITS.

Also to clarify: discernment among the 4 categories can be difficult. In one case in our series, the patient had acute bacterial meningitis, and imaging showed an incidental colloid cyst. Often, one must make inferences about how findings fit into our framework. Clearly, bacterial meningitis is not an incidentaloma - it is the index illness. This makes it easier to realize that a colloid cyst has been around for a while, so it's the incidentaloma. It was previously asymptomatic and would not have been discovered (unless it became symptomatic one day) sans the meningitis precipitating neuroimaging. We are not saying diagnosis is easy, it clearly is not. We are saying that if you too facilely or glibly invoke Hickam's dictum, you run the risk of missing something important about the "causal narrative" of the patient's presentation.

Figure 3 of the article is below and shows a schematic of our framework. If anybody has a case that cannot be fit within this framework, please contact me with details. I'm also interested in alleged instantiations of our "category 4" diagnoses, where two independent symptomatic coincident diseases are necessary to explain an acute presentation, as I don't think this happens often, if ever.

Figure 3: Framework for Understanding Multiple Diagnoses. Pictured is a 25-year timeline of diagnoses for a hypothetical patient. The onset of labelled diagnoses (e.g., appendicitis, congestive heart failure) is marked by a vertical line on the timeline and for illustrative purposes the duration of disease is indicated by an attached horizontal line pointing into the future. Acute diseases terminate with a vertical line while chronic ones extend into the future, indicated by an arrow. Diagnoses up until 2023 occur more or less at random times. The focal diagnosis is the patient’s presentation in 2023 with pneumococcal pneumonia (shaded). Additional diagnoses noted during that episode can be grouped into four categories as shown: incidentalomas discovered during the pneumonia episode representing a spurious coincidence (category 1); temporally disparate previously diagnosed acute or chronic conditions (category 2); complications or associations of the pneumonia that are causally dependent and therefore probabilistically enhanced (category 3); or, rarely, causally independent diseases with symptoms coincident with the pneumonia (category 4, dashed line). The lines or brackets associating possible secondary diagnoses (categories 1-4) with the primary diagnosis (pneumococcal pneumonia) are qualitatively wider or narrower to represent the frequency of occurrence of that category in part 1 of our analysis.