Archived Collabnet Discussions

Latest allergy file 08/12/2011

This is the file we reviewed during today's meeting.

bgoldberg

Wed Aug 17 23:48:18 Z 2011

post5148

topc3631

Five more complex (>2) combined disorders

As promised here are 5 more examples to work on independently

bgoldberg

Fri Jul 11 03:52:20 Z 2014

post10286

topc6921

Notes on allergy construction phase

Here are today's notes from my meeting with Monique

bgoldberg

Thu Sep 06 00:10:51 Z 2012

post7237

topc4962

Revised allergy ontology

Here are some revisions based on the discusions of the last meeting.

bgoldberg

Wed Jun 29 21:36:26 Z 2011

post4582

topc3185

Re: Revised allergy ontology

Thanks Bruce. I've posted a modest revision of this file [doc3937] to experiment with whether it's possible to generate *something like* the Visio Class diagram. Using Protege 4.1 and the OntoGraf plugin I was able to generate the diagram shown in [doc3938]. The static screenshot doesn't show that the role/attribute names can be shown by selecting each relationship. It was a bit fiddly to do, but not impossible, so we might want to try maintaining our working models in this environment. p.s. - I think the hasAgent relationship in the visio diagram should be reversed. I seem to recall we are yet to determine the correct flat/nested models to associate the disposition and the allergic reaction classes with the allergen. Kind regards Ed

edcheetham

Thu Jun 30 14:10:02 Z 2011

post4592

topc3185

Discussion of causality for complications/sequelae

I provided an excerpt from Art6301_Sequela (finding) and sequela of disorders-elaboration phase v1.1 in which I discuss capturing causality and temporal sequence for complications and sequelae.

bgoldberg

Sat Dec 15 19:16:19 Z 2012

post7778

topc5333

Conditions vs. Situation

Dear ECE committee, as I told you in earlier sessions, we have had intensive discussions in the IHTSDO - WHO Joint Advisory Group whether SNOMED CE finding / disorder concepts denote 1. clinical conditions (pathological entities as being subject to the ECE work) or 2. clinical situations referring to clinical conditions We did an evaluation of a random sample of 400 SNOMED CT concepts from the disorder hierarchy, the results of which are summarized in a manuscript we submitted to AMIA 2012. I attach the paper together with a spreadsheet containing the raw data including comments of the raters on "difficult" SNOMED CT concepts. Best regards, Stefan

sschulz

Thu Mar 29 09:32:40 Z 2012

post6424

topc4412

Re: Conditions vs. Situation

Thanks Stefan The paper concludes/recommends that we should "...declare all disorder codes to have a situation interpretation...". I wonder, therefore, if we should revise the working name of our SDP disjunction class to 'situation'? It would seem unhelpful to make too much progress evolving an approach which depends on the 'condition' class only for us to re-interpret it in the near future. I'm also wondering (again) about how to accommodate 'normal' findings (any statements of 'normality' or explicit findings within expected parameters - e.g. skin color or states of mucous membranes). I think we have previously concluded that 'condition' (or, actually, situation) can encompass normality - by definition neither require abnormality or presence of pathology. It feels as though the SDP approach (with a 'condition/situation' class) could include such statements ('my cat's coat is in good condition'). I can see how statments of normality could often commit to being 'structural' (soft fontanelle) or processual (knee reflex on right normal). I'm now wondering whether statements of normality can also commit to being dispostions ('fit and well' or 'tends to plan ahead'). Ed

edcheetham

Fri Mar 30 14:06:57 Z 2012

post6443

topc4412

Re: Conditions vs. Situation

I've added some examples to Bruce's allergy OWL file to try to illustrate using actual logic axioms. Some things to note: (1) the "condition" items don't have to be pre-coordinated. They could be referenced as nested definitions. (2) the use of "only" seems to require the adoption of a classifier that goes beyond EL++ but we might find a workaround. (3) I think this model is _very_ good at elucidating what we really mean, and at throwing out the old "finding vs disorder" conundrum and replacing it with clean distinctions, as between a situation that includes a condition, versus a structure, process or disposition that is included. The examples of cough, pain, seizure, and sickle cell anemia illustrate how it works (in addition to the allergy/allergic reaction examples already included).

kspackman

Wed Apr 04 05:27:32 Z 2012

post6459

topc4412

Re: Conditions vs. Situation

Further explanations of my view of the SDP disjunction, conditions, situations, and how to solve the "finding-disorder conundrum": If we re-interpret the entire "clinical finding" hierarchy as "situation present", this means that a concept in SNOMED CT with a (finding) or (disorder) tag currently could retain the same name, but its meaning would be interpreted as follows: A clinical finding is a situation, i.e. a period of life of the subject, during which there exists (the situation "includes") some condition. Some conditions are only structures. Some conditions are only dispositions. Some conditions are only processes (events). And some conditions are disjunctions of 2 or 3: Some are either events or dispositions (see seizure and cough, below). Some are either structures or dispositions (see anemia below). And I could possibly come up with an example that is either an event or structure - can't think of one at the moment. A seizure (finding) is a period of life of the subject during which there exists some seizure condition. A seizure condition could be defined as the disjunction: seizure event OR seizure disposition. A cough (finding) is a period of life of the subject during which there exists some cough condition. A cough condition is the disjunction: coughing event OR coughing disposition. An anemia (disorder) is a period of life of the subject during which there exists some anemia condition. An anemia condition is the disjunction: circulating blood with a low level of hemoglobin (a structure) OR disposition to low hemoglobin. Abilities are dispositions. This means that "able to tie shoes (finding)" would be a period of life of the subject during which there exists the ability to tie shoes. Note that lack of ability, e.g. "unable to tie shoes (finding)" would have to be defined a a period of life of the subject during which there does NOT exist the ability to tie shoes. i.e. this would invoke a model that says something like: situation and NOT some includes "ability to tie shoes (disposition)".

kspackman

Wed Apr 04 17:36:36 Z 2012

post6470

topc4412

Re: Conditions vs. Situation

Thanks for this. Looking through in detail in preparation for tomorrow and I must confess I am struggling a bit! Recasting 'seizure (finding)' and 'cough (finding)' as described seems to me to add unwelcome complexity, not least to the age old 'which code should I use?' question. The proposal appears to require four coded representations with seizure in their name: - Seizure event (*) - Seizure condition - Seizure disposition (*) - Seizure situation ...to help us clarify a problem that began with not being sure which of the existing two (starred) to use! If this proliferation is not really required then I apologise, but if it is needed then this does not look like an attractive proposal. With the exception of the 'condition' disjunction, this looks like the 'make everything a situation' proposal from 2005 which was resisted due to the inherent disruption. On a more positive note...and perhaps reaching across to the observables project (which we occasionally do!)... Is there any value in thinking about the process/event and structure classes as 'observable', and the dispositions as 'non-observable'? I admit this line of thinking is slipping back to our 2009 'discriminatory questions' approach to dealing with existing content, but it might be of value in determining which content should be classified as each of the S,D & P disjunctions (I have in mind the standard 'allergic rhinitis' since this would seem to be a label we wish to attach to the observable phenomenon and the 'propensity' (which isn't directly observable). It might also help identify (in a kind of 'OntoClean' way) any cases where a potentially non-observable phenomenon is currently classified as the subtype of a necessarily 'observable' one . Close to Kent's post above, I wonder if this approach might finesse the proposal for 'abilities'. I agree that the 'able to...' and 'unable to would fit as dispositions, but current content includes 'does...' and 'does not...' constructs (e.g. 'does tie shoe laces'). These latter categories - as worded - would appear to be 'observable' and therefore might require different treatment. Kind regards Ed

edcheetham

Tue Apr 10 15:58:31 Z 2012

post6487

topc4412

Re: Conditions vs. Situation

The OWL examples show a model of meaning. This underlies models of use, but doesn't need to get fully exposed as codes for people to choose from. For example, I agree with you - I don't think we necessarily need a separate code for seizure condition. It is logically just an "or". Event or disposition. The point I will emphasize again is that not everything in the model needs to be pre-coordinated into the terminology - only the things that we think people need. Your objections are well-taken when it comes to models of use, and we need to avoid disruption as much as possible. But something has to change, and it needs to change in the direction of having our model(s) of use founded on a model of meaning that is more faithful to reality. And I think this model of meaning avoids the confusion of the old "finding-disorder" conundrum, and helps to illustrate the reasons for lack of reproducibility. People would read a code and implicitly assume a process, or a structure, or a disposition, or a situation in which there exists one of those. And based on those implicit assumptions, they would make conflicting assessments of whether a "finding" label or a "disorder" label was appropriate. (Retinal hemorrhage -> finding or disorder? That's tough. Retinal hemorrhage -> structure, process, disposition, or situation? At least now I can tell what you are assuming when you answer).l

kspackman

Tue Apr 10 16:25:56 Z 2012

post6490

topc4412

Elaboration phase document for X with Y project

Here it is, finally. I have borrowed heavily from Ed's excellent documents and diagrams. - thank you Ed. I have not included the current work on >3 entities. I plan to include this as an iteration in the future. Please provide feedback after which I will forward to the editors group and Kent. Bruce

bgoldberg

Thu Sep 04 00:25:02 Z 2014

post10523

topc7110

Disease Model article published

Schulz S, Spackman K, James A, Cocos C, Boeker M. Scalable representations of diseases in biomedical ontologies. J Biomed Semantics. 2011 May 17;2 Suppl 2:S6. Open access publication, available at: http://www.jbiomedsem.com/content/2/S2/S6/?mkt=

sschulz

Wed Jun 08 18:46:56 Z 2011

post4405

topc3076

Re: Disease Model article published

Hi Stefan Many thanks for posting this. Reading it again leaves me with my usual mixed emotions! I remain very interested in the paper's motivation (reconciling the messy needs of terminology developers with the rigour of formal ontologist), and the condition/SDP structure gives me a glimpse of something that might work, but I am still currently unable to align the proposals fully with (a) the patterns of data we have under consideration in the ECE project and (b) a duty to leave some of the prevailing and familiar patterns in SNOMED CT data unchanged. Regarding (a), I'm sure we will need to abstract away from a 'pathological' model to one which can accommodate 'all clinically-relevant entities'. Perhaps this isn't a big step, but nevertheless this is needed. By example, if we take terminology content as a proxy for recording requirements, a notion such as "248166007 | Smartly dressed (finding)" is in scope, and indeed could be conceived of as structural (well-pressed shirt), processual (performs the act of dressing neatly) and dispositional (displays a tendency to dress smartly). Nevertheless I *think* it would be wrong to classify this as a 'condition'. Regarding (b), I honestly think there will be a riot if we disrupt the age-old convention of defining much of the current 'finding/disorder' content in terms of the structure involved (via finding-site=body_structure, or T-code) and what is morphologically wrong with the structure (via assoc_morph=morph_abno, or M-code). I personally think the concern that 'pleural fibrosis' isn't a kind_of 'fibrosis' is a false argument. I can easily emulate 'fibrosis (disorder)' as an appropriately constructed query, which returns pleural fibrosis without any problem. Preserving an 'axial' separation between SNOMED's 'clinical concepts' (let's call them situations with or without explicit context', or 'the stuff that goes in the record') and the body structures to which they refer is becoming familiar to many (at a critical time in SNOMED's widespread uptake) and has significant practical appeal. I am sure that an approach can be identified which preserves this familiar separation but allows recovery of/transformation to the configuration you propose. Nevertheless I will continue to argue against the general case of 'disordered structures' being kinds_of 'canonical structures' (but perhaps we need to leave room for valid 'exceptions' such as undescended testes and accessory nipples). Also: (1) I am still not sure how we would scalably implement this model (or some modification). As your paper suggests, there will be some content which necessarily declares its SDP semantics, and other content which is less clear (the 'ambiguous' concepts). Of the 'ambiguous' group, some *could* reasonably be all three (S, D or P), but others could only really be two. In the latter case, publishing an uncommitted 'condition' representation would be unsuitable. Perhaps we need a couple more disjunct classes to cover 'two-way' ambiguity. (2) I remain unconvinced that most valid 'dispositions' have discretely identifiable bearers (certainly following the pattern of Huntington's and trisomy 21). We are then reliant on either modelling with bare fiat assertions ('this notion is a disposition because we say it is') or defining dispositons in terms of value restrictions on their realisation. Without disjunction we can only really do this for dispositions with monotonic realisation, and even with disjunction this seems to challenge conventional boundaries of 'terminology knowledge'. Take, for example, rhematoid arthritis: one perspective on this is surely as a disposition, but (my) best current knowledge draws a blank as to how this would be defined - either in terms of its bearer or realisation. In an SDP world, would Rheumatoid arthritis have any direct association with 'joints'? Kind regards and thankyou again - perhaps we can explore this on the call 22nd June? Ed

edcheetham

Fri Jun 10 09:58:06 Z 2011

post4422

topc3076

Re: Disease Model article published

Below are Stefan's initial responses to my comments (using the convention of commenting between indented (>) original sections. We agreed that if these discussions resulted in any substantial feedback on the paper then we would use the Journal of Biomedical Semantics commenting mechanism. Kind regards Ed ######### > > Many thanks for posting this. > > Reading it again leaves me with my usual mixed emotions! > I remain very interested in the paper's motivation (reconciling the > messy needs of terminology developers with the rigour of formal > ontologist), and the condition/SDP structure gives me a glimpse of > something that might work, but I am still currently unable to align > the proposals fully with (a) the patterns of data we have under > consideration in the ECE project and (b) a duty to leave some of the > prevailing and familiar patterns in SNOMED CT data unchanged. > > Regarding (a), I'm sure we will need to abstract away from a > 'pathological' model to one which can accommodate 'all clinically-relevant entities'. I agree, and therefore "condition" seems for me a sufficiently general "umbrella" concept > Perhaps this isn't a big step, but nevertheless this is needed. By > example, if we take terminology content as a proxy for recording > requirements, a notion such as "248166007 | Smartly dressed (finding)" > is in scope, and indeed could be conceived of as structural > (well-pressed shirt), processual (performs the act of dressing neatly) > and dispositional (displays a tendency to dress smartly). Nevertheless > I *think* it would be wrong to classify this as a 'condition'. The boundary between strictly medicine and human life in a broad sense is vague. Whether something qualifies as a condition in our sense, certainly depends on the need to document it and to base diagnostic or therapeutic decisions on it. For a dentist, "Smartly dressed (finding)" is probably irrelevant, for a psychiatrist or geriatrist is may be a reportable fact. I don't have a problem to subsume it under "condition", but maybe this word has additional connotations I am not aware of as a non-English speaker. We need, nevertheless, define "condition" and "health-related" or "clinical condition" a bit more precise. A condition is part of or located in a subject of care, it inheres in it, or it has it as a participant. This is clearly not the case of the well-pressed shirt (as such), but of the process of a human dressing a shirt, or the process of a human keeping his clothes clean. > Regarding (b), I honestly think there will be a riot if we disrupt the > age-old convention of defining much of the current 'finding/disorder' > content in terms of the structure involved (via > finding-site=body_structure, or T-code) and what is morphologically wrong with the structure (via assoc_morph=morph_abno, or M-code). I never thought about it in a disruptive way. The good thing of a multiple hierarchy is that it allows for multiple views. > I personally think the concern that 'pleural fibrosis' isn't a kind_of 'fibrosis' > is a false argument. I can easily emulate 'fibrosis (disorder)' as an > appropriately constructed query, which returns pleural fibrosis without any problem. > Preserving an 'axial' separation between SNOMED's 'clinical concepts' > (let's call them situations with or without explicit context', or 'the > stuff that goes in the record') and the body structures to which they > refer is becoming familiar to many (at a critical time in SNOMED's > widespread uptake) and has significant practical appeal. I would have liked to discuss these issues more before submitting the final version of the paper (and having you as a co-author, too), but there were strict time constraints as you remember. Then the current version of the paper would reflect more the position of all ECE members. But a paper is a snapshot anyway, and we can (and should) publish results of the follow-up discussion later. The fibrosis example is probably not problematic for very experienced SNOMED CT users, but it is puzzling at a first glance, because nearly everybody would intuitively assert isA between 'pleural fibrosis' and 'fibrosis'. Certainly, with appropriately constructed queries you can go a long way even with lots of idiosyncratic features in SNOMED CT. But for this you need no know it very well. What we could discuss further is whether the all clinical concepts should really be seen as "situations" (there is an ongoing discussion in the IHTSDO/WHO JAG on this). There is some evidence that shows that SNOMED CT modelers are thinking that way (otherwise it would be wrong to have, e.g. 'Ventricular septal defect (disorder)' as a parent of 'Tetralogy of Fallot (disorder)'. There are other disorder concepts where there is no evidence for this assumption (e.g. 'Diabetic foot' is not a child of 'Diabetes mellitus'). > I am sure that an approach can be identified which preserves this > familiar separation but allows recovery of/transformation to the > configuration you propose. Nevertheless I will continue to argue > against the general case of 'disordered structures' being kinds_of 'canonical structures' (but perhaps we need to leave room for valid 'exceptions' > such as undescended testes and accessory nipples). Disordered structures are never kinds of canonical structures. For instance: 'diabetic foot' isA 'canonic foot' would be simply wrong The question is how 'diabetic foot' isA 'foot' should be considered true. As I interpret SNOMED CT, the anatomy does not commit to whatsoever canonicity, because otherwise most disease definitions would be highly questionable: The fact that some disease (e.g. Hallux valgus) has finding site some anatomical structure (here: foot) , implies in most cases that that structure is not canonical any more. > (1) I am still not sure how we would scalably implement this model (or > some modification). As your paper suggests, there will be some content > which necessarily declares its SDP semantics, and other content which > is less clear (the 'ambiguous' concepts). Of the 'ambiguous' > group, some *could* reasonably be all three (S, D or P), but others > could only really be two. In the latter case, publishing an uncommitted 'condition' > representation would be unsuitable. Perhaps we need a couple more > disjunct classes to cover 'two-way' ambiguity. A priori, all concepts would be ambiguous with regard to SDP. The insertion of the 'Condition' node would have no impact. If there is a subconcept which can only be D or P but not S would not contradict this (in the same way as "organic compound" is under "chemical" does not contradict that the overarching node "substance" includes also inorganic compounds, which are disjoint from organic compounds). And we can't express disjointness or negation anyway in SNOMED CT. But even if logically indifferent, the insertion of additional disjunct classes is a good idea. It won't have any impact on reasoning so far, but it would make things clearer. In case SNOMED CT will be split into logically disjoint partitions in the future, it will also have computational consequences (e.g. an "S" interpretation of e.g. "Cough" would created an inconsistency. ) > (2) I remain unconvinced that most valid 'dispositions' have > discretely identifiable bearers (certainly following the pattern of Huntington's and trisomy 21). > We are then reliant on either modelling with bare fiat assertions > ('this notion is a disposition because we say it is') or defining > dispositons in terms of value restrictions on their realisation. If we accept that the property of being bearer of a disposition propagates from parts to wholes AllergicRhinitisDisposition inheresIn some NasalMucosa NasalMucosa partOf some Nose inheresIn o partOf -> inheresIn ---------------------------------------------------------- AllergicRhinitisDisposition disposition inheresIn some Nose then we can always assert the whole body being bearer of a disposition, and refine according to the advance of science. I think we are on the safe side here. > Without disjunction we can only really do this for dispositions with > monotonic realisation, and even with disjunction this seems to > challenge conventional boundaries of 'terminology knowledge'. > > Take, for example, rhematoid arthritis: one perspective on this is > surely as a disposition, but (my) best current knowledge draws a blank > as to how this would be defined - either in terms of its bearer or > realisation. In an SDP world, would Rheumatoid arthritis have any > direct association with 'joints'? > I would model it similarly: RheumatoidArthritisDisposition inheresIn some SynovialTissue SynovialTissue partOf some Joint inheresIn o partOf -> inheresIn ---------------------------------------------------------- AllergicRhinitis disposition inheresIn some Joint > Kind regards and thankyou again - perhaps we can explore this on the call 22nd June? Of course... looking forward to the meeting. What about posting your and my message on the Collabnet? Best regards, Stefan #########

edcheetham

Fri Jun 24 16:37:23 Z 2011

post4527

topc3076

eMeasure issues group meeting

Requested presentation on allergy value sets.

bgoldberg

Thu Feb 21 01:53:51 Z 2013

post8168

topc5647

Top 2500 KP diagnoses for SDP analysis

This file contains KP's top2500 problem list diagnoses with SNOMED, ICD-9 and ICD-10 maps. The file appears to eb arranged in roughly alphabetical order and does not contain usage data. I will try to get usage data and post a new version if I can.

bgoldberg

Thu Feb 12 17:28:21 Z 2015

post10985

topc7391

Additional file for Wed. meeting

I inadvertently attached my elaboration phase draft for allergy. I am sending now just the revised class diagram.

bgoldberg

Sun Jan 22 08:26:08 Z 2012

post6094

topc4224

My completed assignment

Here is my take on the 5 new terms to model. I made an error in choosing the 4th item, it really is just an X with Y

bgoldberg

Tue Aug 12 05:17:58 Z 2014

post10408

topc7011

Re: My completed assignment

Please find my homework.

ygao

Tue Aug 12 15:33:43 Z 2014

post10409

topc7011

Re: My completed assignment

Please also find the slides about my thoughts on patterns for representation of multiple conditions as a situation (scope in due to, co-occurrent with).

ygao

Tue Aug 12 15:36:45 Z 2014

post10410

topc7011

Re: My completed assignment

and mine...

edcheetham

Tue Aug 12 16:08:02 Z 2014

post10411

topc7011

Pseudoallergies

I have attached my latest thoughts on modeling pseudoallergies and pseudoallergic reactions. After some thought, I have resurrected our old nemesis "associated with" as a means of representing the ambiguity between the underlying substance and the pseudoallerigc disposition and reaction. I have attached an owl file and 2 screenshots illustrating how the disjoint class, allergy or pseudo allergy to substance alonfg with associated with some particular drug can retrieve the appropriate term even if one does not know if the drug is related to an allergic or a pseudoallergic state.

bgoldberg

Wed Feb 13 07:03:01 Z 2013

post8096

topc5580

Meeting this week II

Ed, in preparation for meeting this week, I have attached 3 files. The first is a revised version of the owl file containing the updated made to the class diagram last week. The second is a slide contrasting our proposed class model with the current SNOMED model with suggestions for reconciling the two The third is a draft agenda for the Sydney meeting.

bgoldberg

Tue Sep 27 02:45:33 Z 2011

post5355

topc3756

Updated SDP allergy owl file

This incorporates the latest updates to the model

bgoldberg

Sun Apr 01 17:45:06 Z 2012

post6449

topc4427

XwithYSyndrome20130710

Here is my work on this this

bgoldberg

Wed Jul 31 22:20:54 Z 2013

post8780

topc6019

Re: XwithYSyndrome20130710

Thanks Bruce Here's mine (independent of your work). Kind regards Ed

edcheetham

Thu Aug 08 13:38:41 Z 2013

post8795

topc6019

Re: XwithYSyndrome20130710

Here is my spreadsheet. Best regards, Stefan

sschulz

Fri Aug 09 10:31:28 Z 2013

post8798

topc6019

Files for today's meeting

2 owl allergy files and the x with y spreadsheet (I was only able to get to the first 100).

bgoldberg

Wed Jun 13 19:51:41 Z 2012

post6818

topc4684

Most recent diagram (R4) of allergy classes

This is the diagram we were working on during the last meeting

bgoldberg

Wed May 25 21:43:25 Z 2011

post4335

topc3054