Infections, co-occurrence, causation, sites and morphology.

One of the topics we briefly discussed on the last call in March was the relationship between 'infections' and the consequences/manifestations of those infections. As illustrated in the GP/FP work, 'infective uveitis' *could* be recast as 'uveitis co-occurrent with eye infection', or perhaps 'uveitis co-occurrent and due to eye infection'. The instinctive (and understandable) reaction to this is that such an approach would be disruptive and unnecessary, plus the editorial guide and current approach to infections (6.1.3.10 and  6.1.2.12) appears to explain how 'infectious process' and 'inflammation' interact.

However, from the perspective of current ECE discussions on co-occurrence and causation it might be worth digging a little deeper. I think we can avoid going right back to 'what does SNOMED CT mean by an infectious disease', but I might be wrong. There's no text definition, but I imagine it means any disease caused by an infectious organism (by initiating an infectious process). The problem/question I have relates to the way that the infectious 'condition' relates to other 'conditions' in each 'clinical life phase' (and thus, to morphologies and body structures).

In the 'infective uveitis' example it's generally understood that the infection is causing the inflammation - and this will be the case for the majority of 'infectious process' + 'inflammation' life phase combinations. However there are a un-ignorable number of cases where the infection is stated to be happening in a morphologic abnormality from another cause (e.g. infected blisters, splinters and abrasions), and in a smaller number of cases infection of an already inflamed site (e.g. infected eczema). There is currently no declared distinction between the way each of these will be dealt with in SNOMED CT, consequently 307311001 | Infected ascites subsumes 4501007 | Tuberculous ascites.

So, does the distinction between 'infective morphology at a site' and 'infected morphology at a site' matter (noting that this language difference is fairly common, but not discriminatory, in the data)? I think there are reasons to consider it worthy  of further discussion:

(1) It may spark a rethink to the reasoning behind why/whether we want to identify pattern 3 (X+Y co-oc and due_to Y) life phases? This could go either way - identify more, or identifying fewer. If it's agreed that infections 'cause and happen at the same time as the inflammation', then either they should all be explicitly modelled this way, or - if not - many other 'co-oc and due to' combinations can simply be implicity treated as 'co-occurrent' because the causal aspect is 'understood and inevitable'.

(2) Improve our understanding of the 'causative agent' relationship. Its semantics are known to be problematic for various substance-related content (allergies, overdoses, abuse), and it might be helpful to revisit 'disorder + material entity' combinations, and identify the variants - in particular the scope and nature of the part played by the material entity.

(3) Inform the remedial modelling of 301810000 | Infection by site (disorder). This is problematic, and establishing what the correct solution is will benefit from an improved shared understading of how 'infections with organisms' relate to 'abnormalities in body structures'. Currently - as can be seen by inspecting the sibling classes of 'Infection by site' (that is, the immediate subclasses of 40733004 | Infectious disease), it can be seen that there are plenty of named infections of sites (e.g. 299990002 | Infection of foot (disorder) and 275498002 | Respiratory tract infection (disorder) that are not subtypes of 'Infection by site' - by my calculations about 900 named site infections are currently misclassified. I think this stems from the RG in the modelling of 'Infection by site'. It may be enough to remove this (leaving the roles ungrouped), but I think it would be better to agree standard modelling patterns (with preferred role group combinations) for several of the following (which are currently inconsistently applied):

i. infection

ii. infection at site

iii. infection by organism

iv. infection by organism at site (as an aside, for iii and iv, including consideration of whether infections with hepatitis B/C/... viruses necessarily happens in the liver.)

v. infection by organism causing morphologic abnormality

vi. infection by organism causing morphologic abnormality at site (e.g. tuberculous ascites, infective corneal ulcer)

vii. infection by organism in morphologic abnormality

viii. infection by organism in morphologic abnormality at site (e.g. infected ascites, infected varicose ulcer)

 and if needed:

ix. infection by organism in morphologic abnormality, causing other morphologic abnormality, at site

...and consider how these might generalise into other areas of representation.

Ed

Ed,

Thanks for your usual detailed discussion.  In your last point (3) your analysis of existing issues with infection by site is correct in that not all of the existing subtypes have been remodeled according to the new guidance; however, the two examples you give are currently subtypes of Infection by site.  As for the 900 or so that are not classified correctly, I would be interested in getting a listing of those so they can be prioritized in the revision process (which has been stalled awaiting a more robust editing environment).

Thanks Jim

Yes, sorry. Almost the moment I sent this I realised my October 2015 analysis was out of date - with several of the previously non-subsumed 'infection of named site' corrected for January 2016 (I'm sure reducing the 900 figure significantly). Nevertheless there are still a number of sibling classes that are not yet included, and there may be others elsewhere. Once I've built my UK April 2016 data I'll have a clear answer. I still think it's worth exploring the semantics of the various modelling combinations in this area - 'pathological process=infective process' seems to have vary its meaning when it's role grouped with a body structure, an organism or a morphology, and a clean approach to each pattern might be useful. Thanks anyway. Ed