I have a target cohort, created in ATLAS Cohort Definitions, that starts with 9,518 patients at cohort entry, and goes to 6,569 patients after all exclusions. When I go to Incidence Rates and pair it with an outcome cohort, the target cohort has 6,409 patients (and of those, 1,134 are cases that are also in the outcome cohort).
I don’t fully understand why the cohort goes from 6,569 to 6,409.
In the SQL code that Incidence Rates generates, I found a line that requires the outcome cohort’s start date to be after the adjusted start date of the target cohort – and this accounted for the difference between 6,569 and 6,409 in my cohorts.
I need a plan-English explanation for the drop from 6,569 to 6,409 so I can use it in a consort diagram for a manuscript.
The Incidence Rate function drops prior outcomes, so anyone who has an outcome that starts before their Target cohort start date, they are excluded.
In addition, people must have at least 1 day at risk, which would not occur if the outcome happened on the same day as your time-at-risk start, or if the time at risk start was the last day of their observation period (hence contributing zero days at risk, which would lead to a divide by zero)
-Chris
Hi Chris:
Thanks for the explanation. I understand what you’re saying, but I still don’t understand why the prior outcomes are in effect dropped from the original Target cohort. In my example, the Target cohort had 6,569 people. The Outcome cohort had 30,654 people. ATLAS Incidence Rates indicated that there were 6,409 persons (which I take to be an adjusted Target cohort) and 1,134 cases. Am I correct in thinking that, of the original 6,569, 1,134 are “temporally correct” cases, and 160 (6,569 - 6,409) are “temporally incorrect” cases? And that, for this calculation in Incidence Rates, the temporally incorrect cases are dropped from the original Target cohort? To me, it seems that the 160 were retroactively dropped from a valid Target cohort.
Thanks,
John
Well, I don’t like the idea of introducing new terminology like ‘temporally correct’…because some of those 160 people may simply have no time at risk and therefore do not contribute to the analysis…is that “temporally incorrect”? I just would say there were 160 people that were “excluded from analysis”…the reasons are they had a prior outcome or did not contribute time at risk.
I’d just simply say that 160 people were excluded from the analysis for the reasons stated above. I’m not sure I’d describe it as ‘retroactively dropped from a valid target cohort’…just because they are in the cohort doesn’t make them valid for analysis.