Here is our article that defines and disambiguates both #ockhamsrazor and #hickamsdictum for #differentialdiagnosis https://link.springer.com/arti... And a blog post with additional information: https://www.medicalevidenceblo... And tweetorial as promised
We set out to determine which of two supposedly conflicting axioms - Hickam's dictum (HD) and Ochkam's razor (OR) - is better suited for diagnostic decision making. HD is the alleged "counterargument" to OR. First, we had to try to define the axioms
We found that HD is too non-specific - of course patients get multiple diagnoses! Furthermore, nobody has every operationalized OR for diagnosis. Does OR guide you to select a single, a simple, or a unified diagnosis? We settled on a #unifyingdiagnosis as representing OR
We defined a #unifyingdiagnosis as a diagnosis that represents a single #causalpathway from underlying pathophysiological perturbations to the acute presentation, including all underlying causes, #endpointdiagnosis, associations, epiphenomena, & complications....
A "causal narrative" of the illness, so to speak. Such a causal narrative can represent a logical encapsulation of the acute illness episode and can serve as a check on the diagnostic process, and an anchor for how to explain the many abnormalities of the case to patient & family
With this definition, we searched for case reports alleging instantiations of HD (or violations of OR) and found 83. It quickly because apparent that these HD case reports all had a primary diagnosis that explained the acute presentation, as well as one or more of the following:
1. An #incidentaloma (~30%) 2. A known, pre-existing disease (~25%) 3. A component of a unifying diagnosis (~40%) 4. A symptomatic, coincident, independent disease contributing to the acute presentation (~4%)
This framework incorporates time (of onset), probability, and causation, and is a more complete structure for understanding multiple diagnoses than both HD and OR. With it, we can understand the patterns with which multiple diagnoses occur. Figure 3 shows a schematic:
In this framework, incidentalomas represent a spurious coincidence with the acute illness; pre-existing diseases upon which new diseases superimpose are unsurprising, but serve as diagnostic "anchors" for the unwary; and probabilistically dependent diseases comprise #unifyingdx
Only extremely rarely is a completely unrelated (independent) disease symptomatically coincident with another, requiring two diseases to explain the acute presentation, and even withing this minority (4%) we suspect there was incomplete understanding of the case....
And those diseases were actually one of the other categories. Next we searched 225 NEJM cases and found that they all - ALL! - had a unifying diagnosis Finally, we surveyed 265 providers on twitter and found that the vast majority - 79% +/- 5% - thought that our 4th category....
....two independent diseases striking simultaneously, would be least likely/most surprising. This shows that clinicians' intuition comports with our framework It may be tacit knowledge. Our framework makes it explicit and teachable....
Take home message: we have operationalized the approach to multiple diagnoses. When there is more than one, you first try to connect everything into a single causal pathway, representing a #unifyingdiagnosis Then you ignore pre-existing diseases Then you consider #incidentalomas
(Incidentalomas can also be thought of as #falsepositives just like lab tests, or as #redherrings, false clues.) Only if you cannot put it together with this framework that considers time, causation, probability, should you consider an independent second disease
As these will be quite rare (our 4th category.) This analysis comports with Reichenbach's common cause principle which states that "If an unusual coincidence between events A&B has occurred, then it is likely that one caused the other or they both share a common cause"
The entire discussion elucidates the probabilistic and causal relations between diseases in light of the common cause principle and I think it the major contribution of our work. We paid $5000 open access fee so you can read/access w/o firewall. I hope you do.
Please let me know if you have comments or questions. Tagging @AdamRodmanMD because he has been following this work and providing insights for pretty much the duration of the work (2 years). Special thanks to my co-authors @doc_BLocke & @brianpooleMD & our survey respondents!
@Gurpreet2015 @dr_dmorgan @CPSolvers @andrewparsonsMD @UK_CReME @MedEdFlamingo @DrDanRestrepo @Sharminzi @interconsulta @neil_thakrar @CaseyMcQuadeMD @drlessing @dmottacalderon @MohitHarshMD @AnandJag1 @rav7ks I should emphasize that apparently this framework is not obvious or self evident. You need to read the article. I am seeing that people are proposing alleged exemplars of Hickam dictum that clearly fit within our framework, but they don’t seem to realize it. Medicine is hard.