search

ihmeuw / SmartVA-Analyze

SmartVA-Analyze Desktop Application
MIT License
6 stars 11 forks source link

Including unused WHOVA 2016 Medical History variables leads to many fewer "Undetermined" CoDs #22

Open KlausBlass opened 1 week ago

KlausBlass commented 1 week ago

Variable adult_1_1 allows entering only 9 medical conditions for which a health professional told the deceased that he has suffered them. WHOVA 2016 contains 10 more conditions in the Medical History section which may have been diagnosed or positively tested for.

To make these 10 conditions available to SmartVA I added them in the Open Narrative mentioned keyword questions if there exists a keyword for the condition. My rationale is: If the respondent stated, the deceased had a recent positive malaria test, then "malaria" was definitely mentioned during the interview.

This reduces the number of "Undetermined" cases by 15% in our survey. (Most of these are now evaluated as being caused by Malaria).

I suggest to:

  1. either use the 10 discarded conditions from WHOVA 2016 by generating tariffs for them and include them in the shortened instrument,
  2. or at least set the available ON keywords (as I did during input preparation).
aflaxman commented 1 week ago

Thanks for flagging this! If I recall correctly, these additional conditions were not included in the questionnaire used in PHMRC validation study, so we were not able to calculate tariff scores for them. Your strategy of instead including them in the Open Narrative is very clever!

Would you be willing to share code on how to achieved your input preparation? I'm sure other researchers would appreciate the ability to also try your approach (especially if they are working in areas with substantial malaria burden!).