Venous Thromboembolism (VTE)

Results from conversations with David Heit, Sean Murphy, Pathak Jyotishman

• Who are controls for the algorithm?
  • [John] The Mayo-derived VTE NLP software had a PPV=100% and NPV=84%.  We improved the NPV to ~95% by excluding any patient that was "negative" by Mayo VTE NLP and negative for any of the ICD-9 codes from the Vanderbilt VTE ICD-9 code algorithm. Thus, patients who are negative by Mayo NLP and negative by Vanderbilt VTE ICD-9 codes will be "controls".  Sean should be able to provide all of this.  I think we also should exclude patients younger than 45 years as controls.  I don't think we need to exclude any patient as a control based on "other known risk factors".
  • [Sean] The Mayo implementation of the algorithm uses a combination of a mention from each of the first two columns.
• How are the three different columns of NLP to be implemented?
  • [Sean] The Mayo implementation of the algorithm uses a combination of a mention from each of the first two columns [in adjacent sentences].  Therefore, there isn’t any correlation of terms listed in the third column.    The terms listed below behave as stand-alone terms, which means any non-negated term in a section of interest (Impression/Report/Plan or History of Current Illness in EHR provided my Mayo) will be considered as evidence for a case. (EAB Note: these are different than the third column: Vanderbilt terms for VTE NLP discovery - will someone please clarify which terms are 'stand-alone')
    • inferior vena caval thrombosis
    • ivct
    • pulmonary embolism
    • pulmonary embolus
    • deep vein thrombosis
    • deep venous thrombosis
    • dvt
    • pe
    • pulmonary emboli
    • pulmonary thromboemboli
    • vein thrombosis
    • venous thrombosis
    • venous thromboembolism
    • vte
• Are there any age/enrollment/gender/race/ethnicity/coverage restrictions? [Jyoti] My understanding is that VTE is not common within a pediatric population.So, perhaps the only restriction we might apply here will be Age >= 18 years.

Hi, Have you decided whether patients under 18 years are excluded? We had run the old algorithm, (i.e. pre Aug 24) and came back with ~280 participants BEFORE exclusion criteria. We intended to start on the new one, but obviously no point if this age criterion is applied. Many thanks, John (CHOP)

Thanks Jyoti, will proceed!

Best,
John

Hi,

We assume Pregnancy_All is all the pregancy icd9 codes in the ICD9 pregnancy worksheet in the data dictionary.

Which of those icd9 codes are considered Pregnancy_Complete and Pregnancy_Livebirth?

Thank you

Hi, For our site, case in progress notes is irrelevant. That makes some of the values in Table A Column 1 and

Table C Column 1 to be duplicates. Since Table A Column 1 mentions require an accompanying Table A Column 2

mention and Table C mentions are stand alone, I'm not sure which method to pick for these duplicates.

Thanks, Ken

We have a subgroup of subjects who have suffered from strokes.  If there are no other ‘related mentions’ but the term ‘stroke’ is in the progress notes, should these subjects be considered cases or controls?

Similarly, we have a number of subjects who have been diagnosed with cerebral palsy.  Sometimes the only ‘causative’ description given is ‘periventricular leukomalacia’.  Should these subjects be considered cases or controls?

Please advise.

Can someone please clarify what is meant by 'at enrollment' in the data dictionary?

Thanks,

Ken

Hi Jyoti,

Would you like us to provide matched controls? If so, matched on what and at what ratio?

Thank you!

~~~~ sarah