History of venous thromboembolism (VTE) by type

[helenmcd]

Subcategorisation of codes for venous thromboembolism from issue #33

These are codes for venous thromboembolism, including 'history of' codes. We want to be able to categorise the type of venous thromboembolic event, for a more detailed description of the prevalence of VTE among COVID-19 vaccine recipients.

I have downloaded the codelist generatedin issue #33 and attached flags for type of VTE. venous-thromboembolic-disease-type.xls

Have indicated types as:

• "pe" pulmonary embolism (note as issue #33 this excludes codes which specify a non-thrombotic type such as fat or air embolism but assumes that an unspecified PE is a thromboembolic event)
• "dvt" deep vein thrombosis of the lower limb. Have assumed that if the code says "deep vein thrombosis" but does not further specify a site, this is most likely lower limb, so included these codes here.
• "cvt" cerebral venous thrombosis. I included phlebitis and thrombophlebitis of intracranial sinuses in this, but not internal jugular vein.
• "unspecified" unspecified venous thromboembolism, site not specified
• deep vein thrombosis of another specified site that is not lower limb I think in practice we might just combine as "other" but in case we want more detail in future I have classified as
  • "portal" portal vein thrombosis
  • "hepatic" Bidd-Chiari syndrome
  • "smv" superior mesenteric vein
  • "vc" vena cava (superior, inferior or unspecified)
  • "other" upper limb or internal jugular vein

I would be glad of a second opinion, particularly I wondered:

• is it consistent/reasonable to presume "deep vein thrombosis" is most likely lower limb, but treat "venous thromboembolism" differently on the basis that it might be used for unusual/hard to code cases of VTE?
• given the case series of clots with thrombophilia highlighted the splanchnic circulation (https://www.nejm.org/doi/full/10.1056/NEJMoa2104840?query=featured_home ) we might want to separate these from 'other' - it seems hard to predict what we might want to look at in future so I went for detail, but are these sensible divisions?
• is it best to categorise internal jugular vein with upper limb dvt as 'other'?

[annaschultze]

Hi,

I've done a sense check and this all seems very reasonable to me. In response to your questions I think @CarolineMorton input would be great, my take is:

• Agree with assuming DVT is lower limb but keeping VTE as other; although would be useful to have a coding perspective.
• Agree with existing divisions
• Will leave this for Caroline as think this requires medical knowledge to classify appropriately.

[CarolineMorton]

Hi all,

Sorry for the delay. I have been through the excel document and agree with all the categorisations of the codes included. I don't think any need to be changed. I think keeping VTE as other makes sense as we can't say for sure that it is lower limb, and occasionally you do get them in other places. I suspect given small numbers we will want to split into DVT/PE, other and CVT for the analysis to get the numbers, as I think Budd Chiari for hepatic will be very small for example.

Re: internal jugular vein with upper limb dvt - I agree this should be other.

[helenmcd]

Thanks @CarolineMorton! Agree we will likely in analysis want to categorise as DVT/PE, CVT, and other (with all the rest) - but hopefully useful to have facility to separate out e.g. portal vein thrombosis in future studies if wanted.

[annaschultze]

I've added this to opencodelists here: https://codelists.opensafely.org/codelist/opensafely/vte-classified-codes/08b94f7d/

[helenmcd]

Thanks @annaschultze!

Have reviewed ICD-10 codes for the same grouping - while I was there, took the chance to review (1) codelists and (2) criteria for defining VTE.

---

(1) Codelists

---

Existing codelists in the OpenSAFELY database There are three existing OpenSAFELY ICD-10 codelists for VTE (1) https://codelists.opensafely.org/codelist/opensafely/venous-thromboembolism-icd-10/6e25de26/#full-list (2) https://codelists.opensafely.org/codelist/opensafely/venous-thromboembolic-disease-hospital/2020-10-01/#full-list The second includes all codes in (1) and additionally codes for VTE in unspecified sites, which we'd want to include here. Neither includes any codes relating to pregnancy or postpartum, which we would want to include for a full capture of 'a history of VTE' (3) There's also a codelist specifically for cerebral venous thrombosis which includes CVT from all causes (inc pregnancy, post partum and in the context of sepsis https://codelists.opensafely.org/codelist/opensafely/cerebral-venous-thrombosis-secondary-care-or-mortality/43b72958/ and has been discussed with MHRA and PHE

Previously validated codelist

• Also found a VTE codelist by Sultan et al. for VTE (from a validation study in Swedish databases of VTE in pregnancy and post partum - discussed below).
• supplementary to Abdul Sultan A, West J, Stephansson O, et al Defining venous thromboembolism and measuring its incidence using Swedish health registries: a nationwide pregnancy cohort studyBMJ Open 2015;5:e008864. doi: 10.1136/bmjopen-2015-008864
• Validated list (though validation notes below - nonspecific without also requiring anticoagulant prescription) but does not include cerebral venous thrombosis.

New search Ran a fresh search using WHO browser ICD-10, 2019 version.

• Did not include codes for: varicose veins, superficial thrombophlebitis, haemorrhoids, other varices (e.g. oesophageal, genital), unspecified venous disorders, vein compression or portal hypertension, or unspecified venous insufficiency.
• Did include codes which were specific to pregnancy or postpartum (since we are interested in any history of VTE, and also for an incidence study we might find that whether a VTE in the postpartum period was coded using the specific postpartum code might vary)

Comparison of codelists Proposed codelist marked with 'include==1' in the attached. VTE_ICD-10_220421.xlsx

This combines the codelist by Sultan et al (which includes codes in pregnancy and postpartum) with cerebral venous thrombosis codes. In addition propose to add:

• I870 Post thrombotic syndrome (as we're interested in prevalence of a history of DVT, not incidence)
• K751 Phlebitis of portal vein
• and the code headings I26, I80, and I82 as per previous OpenSAFELY codelists. Categorised by type as CTV3 codelist above, with addition of 'renal'

Points for discussion

• an alternative would be to just use the validated codelist by Sultan et al, and a separate codelist for cerebral venous thrombosis, but in particular I think we should add the header codes I26, I80 and I82 if they may be included in the SUS data we have. Main difference then would be omission on I870 (post thrombotic syndrome) and K751 (phlebitis of portal vein).
• should compare to 'Vaccination monitoring collaboration for Europe' proposed codes - would be good to align if possible
• if we could add more than one 'flag' I would separate pregnancy/postpartum codes, and also incident vs H/O codes (e.g. post thrombotic syndrome) for different uses in future. How best to handle this?

---

(2) Criteria for defining VTE

---

Validation study results and implications for defining VTE: Sultan et al found that in the Swedish database, among pregnant/postpartum women that including VTE codes without treatment confirmation risks the inclusion of false-positive cases, particularly for cases recorded in an outpatient setting.

• They have since conducted a validation of VTE diagnosis in CPRD with HES linkage, comparing different VTE definitions to other estimates of VTE incidence. (Sultan et al The Incidence of First Venous Thromboembolism in and around Pregnancy Using Linked Primary and Secondary Care Data: A Population Based Cohort Study from England and Comparative Meta-Analysis. PLOS ONE 8(7): e70310.
• They concluded: "a highly specific definition of VTE (VTE definition A) using data from both primary and secondary care health care settings is required to accurately estimate the incidence of VTE in and around pregnancy".
• Their specific definition was "Our most stringent definition included only VTE diagnoses supported by prescription or evidence of anticoagulant therapy (with either warfarin or unfractionated heparin or low molecular weight heparin) within 90 days of the event or death within 30 days of the event."
• A definition which simply used any diagnosis in either CPRD or HES (or both), irrespective of supporting info (e.g. anticoagulants, symptoms or scan results) estimated twice as many cases as the specific definition.

Relevance to this codelist: the most likely explanation of the findings by Sultan et al seems likely to be recording of historical cases using a diagnosis code (which may be particularly prompted by events such as pregnancy - could also worry about surgery prompting this in the general population) which would affect ascertainment of the timing of the event, rather than whether the event had occurred. So for a prevalence study of any history of VTE, propose to use any diagnosis code in primary or secondary care records - but for any study of incidence of VTE, would strongly consider requiring confirmatory prescription of anticoagulant (or death).

[helenmcd]

Comparison to Vac4EU draft codelists -

The codelist above does not include: K55... since these are abdominal/intestinal vascular events and cannot discriminate between arterial and venous events D73.5 Infarction of spleen I825 Chronic embolism and thrombosis of deep veins of lower extremity (2021 ICD10-CM version - clearly relevant term but not in ICD10 codelist dictionary used in OpenSAFELY) I826 Acute embolism and thrombosis of veins of upper extremity (2016 ICD10-CM version - clearly relevant term but not in ICD10 codelist dictionary used in OpenSAFELY)

We could consider adding in K55 and D73.5. They’re both gut (K55) or spleen (D73.5) infarcts which could be arterial or venous events. If we were adding them, it might be consistent to also add G95.1 vascular myelopathies (arterial or venous event in the spine).

The codelist above also includes codes not in the Vac4EU codelist - in particular codes that specify pregnancy and postpartum. @annaschultze shall we discuss pros and cons of this list? We might want to use both (Vac4EU for comparison to other studies, above codelist if we think it fits our dataset best) but perhaps depends whether the Vac4EU codelist is currently an early draft that might change, or near final version.

[annaschultze]

Decision following call to use the validated codelist by Sultan et al and adding in:

• Code headings (ie I80, I26)
• I870 Post thrombotic syndrome (as we're interested in prevalence of a history of DVT, not incidence)
• CVT codes

I've reviewed the excel spreadsheet and agree with the include column (minus K751, which we agreed to exclude) as well as the classification. The only code I wasn't sure about was I800, which is not included in Sultan, but has been marked for inclusion in our study. I would think it would be relevant, so wasn't sure if that was a typo. Either way I think I'd be in favour of adding this to our codelist.

We have also discussed the EU4VAC codelists and agreed these are likely preliminary, and therefore we will not adapt our codelists based on these at the moment. We may use finalised versions of their codelists as sensitivity analyses at a later stage.

[helenmcd]

Thanks @annaschultze Checked the Sultan et al codes (link https://bmjopen.bmj.com/content/suppl/2015/11/11/bmjopen-2015-008864.DC1) I80.0 is not included Rechecked the code - it is Phlebitis and thrombophlebitis of superficial vessels of lower extremities so should not be included in DVT (apologies, my error).

So have removed I80.0 from the list. Have retained I80, I82 and I26

THe attached list, is as discussed: (1) Sultan et al's codelist for VTE in pregnancy/postpartum (validated in CPRD/HES though note non-specific for incident VTE unless combined with anticoagulant prescription, when it performed well - reasons could be delayed recording of past events, or recording of suspected diagnoses later not found to be the case). (2) Three short headings (I80, I81, I26) (3) One code for post thrombotic syndrome (I870) since we are looking at prevalence of a history of VTE (4) Codes for cerebral venous thrombosis https://codelists.opensafely.org/codelist/opensafely/cerebral-venous-thrombosis-secondary-care-or-mortality/43b72958/

This differs from the [existing OpenSAFELY codelist incident VTE](https://codelists.opensafely.org/codelist/opensafely/incident-venous-thromboembolic-disease/2020-09-30/

) by including codes that specify the event occured during pregnancy or post partum, and post thrombotic syndrome, and by further clarifying type (site) of VTE.