LAUSR

Differential diagnosis is fundamental to medicine. Using DiffNet, a differential diagnosis generator, as a model, we studied the structure and organization of how collections of diagnoses (i.e., sets of diagnoses) are related in the ICHD3. Furthermore, we explored the clinical and theoretical implication of these answers. Methods: DiffNet is a freely distributed differential diagnosis generator for headaches using graph theoretical properties of ICHD3: (1) DiffNet considers each ICHD3 diagnosis as a node. (2) An edge exists between two ICHD3 diagnoses if they are connected by either classification hierarchy or are cross-referenced in ICHD3 comment section. In the current project, we generated a set of differential diagnoses using DiffNet for each ICHD3 diagnosis. We then determined algorithmically the set/subset relationship between these sets. We also determined the smallest list of ICHD3 diagnosis whose differential diagnoses would encompass the totality of ICHD3 diagnoses. Results: All ICHD3 diagnoses can be represented by a minimum of 92 differential diagnosis sets. Differential diagnosis sets for 10 of the 14 first digit subcategories of ICHD3 are represented by more than one differential diagnosis sets. Fifty-one of the 93 differential diagnosis sets contain multiple subset relationships; the remaining 42 do not enter into any set/subset relationship with other differential diagnosis sets. Finally, we included a hierarchical presentation of differential diagnosis sets in ICHD3 according to DiffNet. Conclusion: We propose a way of interpreting headache differential diagnoses as partial ordered sets (i.e., poset). For clinicians, fluency in the 93 diagnoses and their differential, as put forth here, implies a complete description of ICHD3. On a theoretical level, interpreting ICHD3 differential diagnosis as poset allows for researchers to translate differential diagnoses sets topologically, algebraically, and categorically.