Closed CarolineMorton closed 4 years ago
DEFINITION: Any history of a stroke, excluding TIA.
Step 1: Read code list stroke_Read_codes_HF.xlsx
Changes to the original list posted above, after discussion with Caroline:
Also added two additional codes: G605.00 Subarachnoid haemorrhage from basilar artery Gyu6C00 [X]Sequelae of stroke,not specfd as h'morrhage or infarction
Step 2: QOF qof-stroke.xlsx
Step 3: SNOMED snomed-stroke.xlsx
Hi, @CarolineMorton (as mentioned, my apologies) I had to update the READ stroke codelist (STEP 1) as TPP's mapped list had a number of missing codes, due to my original READ list missing codes. Can you sense check before re-sending to TPP?
Revised code list here: READ2_stroke_codelist_revised.xlsx
Do file to create code list here (shows all my inclusion and exclusion terms): Do_file_code_create_stroke_codelist.txt
REFINED definition: All incident or historical stroke codes (CVA, sub-arachnoid and intracerebral haemorrhage) unless specifically occurring in newborns. TIA or reversible ischaemia syndrome codes excluded Codes suggesting later sequelae of stroke, monitoring of stroke, stroke reviews included Syndromes included: lacunar, brain stem stroke, cerebellar stroke, lateral medullary, vertebral and wallenburg Occlusions, stenosis and thrombosis codes (without reference to stroke) excluded cerebrovascular insufficiency not included cerebral heamatoma's not included
Why is TIA excluded?
Good question @brianmackenna
@ianjdouglas, sorry, you did tell me why TIAs are excluded, but I have forgotten. What's the rationale for excluding them?
Okay, I have access to the NSAIDs study protocol now. We are creating a stroke code list, because we want to exclude those with a history of stroke as most are expected to be receiving aspirin. I'm not a clinician, but presumably, this is not the case for TIA.
Not necessarily - also clopidogrel used as antiplatelet in a lot of cases.
Hi,
I've reviewed the stroke read 3 codes from TPP, and added a column for include/exclude (1 or 0). A few points to note:
I've excluded intracranial haemorrhages for now, because they don't necessarily lead to stroke (e.g. subdural bleeds, traumatic subarachnoid). I've included them if they specifically mention stroke / CVA / ischaemia.
I've included IC bleeds caused by ruptured aneurysms, because they probably lead to strokes.
I've excluded all vascular dementias
I've included TIAs for now because it makes sense to me (I see there's been some discussion).
Very happy to adjust if consensus differs!
Henry
Hi - thanks All. The original idea for this group (and the MI group), was that they may be on aspirin to prevent future events. It's not too important that they may not all be receiving aspirin, more that we want to create a cleaner group on NAIDS who don't meet these possible criteria for being on aspirin. Liam suggested not including TIA in this group as less likely they would be on antiplatelet therapy. I'm not a clinician, and happy to go with concensus on this - any other thoughts on whether someone with a TIA likely to be put on antiplatelet therapy? thanks Ian
Very helpful Ian. Can @laurietomlinson or @CarolineMorton comment on the inclusion of TIAs (see above)?
@HenryDrysdale
I'm not a clinician, so I can’t comment specifically about the exact inclusion/exclusion criteria. However, it’s very important we are consistent about the inclusion/exclusion criteria for the code list throughout this process (i.e., STEPS 1-3, then reviewing TPP's code list), otherwise we risk missing codes.
I think we should speak to Liam about this again. My feeling is that: 1) A suspected TIA is a less definitive diagnosis on the basis of clinical symptoms and I am not sure that the coding is as good as stroke for example 2) lots of TIA patients get off-license clopidogrel (75mg OD) post diagnosis. it is less clear how long they continue on this as it "should be stopped when both the patient and the doctor think it should be"
@hjforbes @ianjdouglas would you be able to follow up with LIam about this? My overall feeling is that we should exclude to get the definite strokes but i can be persuaded otherwise
@hjforbes thanks. I'm slightly late to the party, sorry if I'm confusing things!
I see your point about step 1 if we're including TIAs. @ianjdouglas I think TIA patients often get at least one antiplatelet long term (aspirin or clopidogrel). I'm a doc but not a neurologist, so v happy to be corrected!
Aneurysms: I included these because they have a high risk of causing haemorrhagic stroke. However, they are less likely to be on antiplatelets (certainly not to treat the stroke), so perhaps we should exclude, given Ian's comments above.
Arterial occlusions: I included because they might actually mean stroke (e.g. anterior cerebral artery, middle cerebral artery), and these patients are likely to be on antiplatelets. (Carotid occlusion is less likely to mean stroke.) However if we're trying to include only definite strokes, we should exclude occlusions.
My view: stroke codes are a mess. In addition, we are already including two very different pathologies - thromboembolic stroke and bleeding -which overlap but may result in different treatments. We have consluded that we think TIA is likely to be very inaccurate which is fair enough in the absence of clear evidence. Stroke is now a rare event so if we are too specific we risk having very low power. I think we have no idea whether using non-specific codes (eg carotid artery dissection) are commmonly used and/or lead to lots of misclassification. What is the key thing here? I think it is to identify people with confirmed cerebral vascular pathology and if so the current list is okish. If we are trying to think about anti-platelet indications we need to try harder as SAH may be a contraindication. It would be easier to identify prescriptions @CarolineMorton @hjforbes @ianjdouglas
@ianjdouglas shall we discuss this on the next drugs call?
Thanks All. We are using this (and the MI list) to mop up any likely aspirin recipients that weren't already identified via prescribing or QoF codes. So the intention is as a safety net, which would make me thing we should be specific with codes where we think aspirin is likely, and ignore the others. If we miss some less specific codes for people who are actually on aspirin, but who aren't identified through prescribing or QoF, then the likelihood is they are not also being co-prescribed another NSAID, and so they won't show up in our group anyway, so we wouldn't lose anything. How about if we select ischaemic specific codes plus non-specific stroke? Will try and get more from Liam too! If not, yes, let's discuss on the next call. Am keen it doesn't take up lots of everyon'es time as it's not a massive bit of the overall study. Thanks v much
discussed with @laurietomlinson: if point is to get cleanest population, i.e. no aspirin use, then we should think about including TIAs as a proportion of those will be on aspirin or similar agent. Thinking of it like this means that it makes sense to include TIA. I still think we should speak to Liam (@LiamSmeeth) about this and confirm so we are all on the same page.
OK - I can see what you mean. I emailed Liam earlier and will follow up tomorrow
Hi agree many/most with a TIA will be getting aspirin (certainly if they had a TIA in the past year) i f they have not progressed to stroke by 1 year I suspect many will drift off aspirin? So established CHD/past MI, ischemic or at least non-haemorrhagic stroke, or TIA in the past year? L
Meeting note: 16/04/2020 - it was decided not to use stroke as a criterion for exclusion, under the assumption that the vast majority of aspirin users with a histor of stroke would either a) be identified as users of aspirin by prescribing or QoF codes, or b) not be prescribed another NSAID simultaneously, and so the number likely included in the NSAID cohort would be very small, if any
Can I close this issue as we are no longer using?
@hjforbes We now need stroke in the other study group. Are you around to discuss?
See #14 (i will now close this issue)
For study on NSAIDs.
Read 2 code list:
stroke_codelist_lshtm.xlsx
We now need snowmed and qof codes