Hello, I am trying to define a cohort which gives me all patients that have a positive blood culture with a common skin commensal (CSC) and 0-2 days have a different blood culture with the same CSC.
I have the blood culture in the specimen table and the result in the measurement table, where the identified organism is in the measurement_concept_id. The specimen and measurement are linked via the measurement event field.
I couldn’t find an option to make sure the second measurement has the same CSC/concept id and the measurement is from a different blood culture than the first.
That’s what I came up with so far:
cohort entry event: specimen blood culture
inclusion criteria:
2 CSC on same day as initial blood culture with 2 blood cultures on the same day
or
1 CSC on same day as initial blood culture and 1 CSC 1–2 days later with a blood culture on the same day
However, I have the impression that patients having the same CSC on the same day with two different blood cultures are not included in the cohort. That’s why I am wondering if the initial event is somehow “reserved” and cannot be found in the inclusion criteria, in other words, the inclusion criteria only find events different from the initial event.
However, if I define the first inclusion criteria to only have 1 blood culture on the same day as the CSC I have too many patients included in the cohort.
Does anybody have additional insights on this or a suggestion on how I could fix my cohort definition?
Thank you very much for your help!
I include screenshots of my cohort definition in ATLAS here:
Hi, don’t think you need deep nesting…if your entry event is the blood culture then you just need the 2 ORs in your inclusion rule, but the second one is an AND::
having any of the following:
At least 2 CSC between 0d before and 0d after entry event
Or:
having all of the following:
At least 1 CSC between 0d before and 0d after entry event
AND at least 1 CSC between 1d after and 1d after entry event
In your atlas cohort, I don’t think you need to nest under each CSC looking for th HOB again. The initial event’s date is carried down below each OR/AND clause and you are seeking around the index event start date where before…and after… index event.
I was going to suggest:
At least 2 CSC between 0d after and 1d after index
But if both CSC occur on the day after index, that would pass this rule, and I think your requirement is that you either find 2 on index or one-after-the-other in subsequent days.
This isn’t possible since we don’t have an option to do something like ‘with a different concept_id as the other’ or ‘with the same concept id as the other’.
Disregarding the blood culture in the inclusion criterion doesn’t meet all cohort criteria. I also want to make sure that the two CSC are from different blood cultures. That’s why I added the blood culture as an additional subcriterion.
I try to recreate “patients having a positive blood culture with a CSC and having a different positive blood culture 0–2 days later with the same CSC”.
I translated this to
having any of the following:
At least 2 CSC between 0d before and 0d after entry event
And at least 2 blood cultures between 0d before and 0d after index event
Or:
At least 1 CSC between 0d before and 0d after entry event
AND at least 1 CSC between 1d after and 2d after entry/index event
And at least 1 blood culture between 0d before and 0d after index event
Changing the second part so that it doesn’t have that deep nesting has no impact on the cohort. I have a lot of patients missing which have a CSC on the same day. They also have two different blood cultures for that day, but somehow they are not included by the cohort definition. That’s why I thought that there might be a problem with reusing the same blood culture event as the entry event in the inclusion criterion.
If I just want 1 blood culture on the same day as the two CSC, I get a lot of wrongly added patients as they have 2 CSC in the same blood culture and I don’t want them to be included.
If I want
“And at least 2 blood cultures between 1d before and 1d after index event”
a lot of patients are correctly included in the cohort but not all missing patients. And some are included incorectly additionally. Could it be that there is a problem with the dates?
