OHDSI Home | Forums | Wiki | Github

Visit type on index date

Hi all,

I have been struggling with how I should define a feature analysis in Atlas so that I get the correct percentages for visit types on index date (I am interested in inpatient, emergency room, outpatient categories).

I have been playing around with different approaches, but they return different results and I am not sure which one is the correct one, you might be able to help me out here:).

Some examples what I tried:

  1. I created concept sets (inpatient, ER, outpatient) where I included only those standard concepts that were relevant for my databases based on database dashboard results.
    for inpatient: concept ID 9201 Inpatient visit, 262 Emergency room and inpatient visit
    for ER: 9203 Emergency room visit
    for outpatient: 9202 Outpatient visit, 42898160 Non-hospital insitution visit, 581476 Home visit, 581478 Ambulance visit, 5083 Telehealth
    then I created the feature analysis with the different categories like this:

  2. I created a feature analysis where I added a condition occurence of any condition and a visit occurence of (the relevant codes from above) and I added a time restriction of where event starts on the same day as the index date

  3. the same as 2. but I added the condition source concept as well instead of working with any condition (I am doing the study with source codes)

My goal would be to create mutually exclusive groups of inpatient-outpatient-ER so that I know in which of these 3 categories the patient was diagnosed (the index date).

Now I not only get different percentages , but they are not mutually exclusive either, but way above 100%, assuming that the system associates different visits with the index date for patients and I am not sure where it goes wrong.

I did additional modifications hoping that I would get mutually exclusive groups, but no success. some examples:

  1. like No3., but I didn´t include the concept ID 262 Emergency room and inpatient visit in the visit type for inpatient category:

  2. like No3, but I thought maybe the issue is that some visit types overlap or are mapped to each other in some way that´s why they are double counted. so I added a nested criteria stating that 0 occurence of visit type 9203 ER and 262 ER and inpatient (for category inpatient and for outpatient) assuming that ER could have caused this issue.

sooo my questions are:

  1. what is the correct way to define a feature analysis for visit types inpatient, outpatient and ER if I wanna see where patients were diagnosed on index date?
  2. How can mutually exclusive groups be created for these visit types in a feature analysis?

Thank you so much, I hope someone understands the system better than me in this regard.