Good day
Background: I am trying to conduct a retrospective epidemiological study looking at the incidence of a certain disease (on admission) in one year (2020) compared to another (2019).
In this case the year would be the exposure variable (2019 vs. 2020), the outcome being the incidence of disease.
Data: Retrospective admissions data for a pre-defined population between the period 23/03/2020 to 01/08/2020 was collected. This totals 182 admissions with 7 incidences of the disease on admission (1 readmission).
Data on the time period of 23/03/2019 to 01/08/2019 was also collected. This amounts to 218 admissions with 17 incidences of the disease on admission (1 readmission)
Analysis plan: Use stset command for dataset and perform Cox regression, taking into account the reoccurrence (readmissions) of the disease and non-reoccurrence to perform a time-to-event comparison between 2020 and 2019 (confounders to be adjusted for).
Problem: These are the same time-periods (start 23/03 and end 01/08) but in different years. Can I compare datasets using stset for different date start points (23/03/2020 vs 2019/03/2020).
My (rather crude) solution was to simply use 23/03/2020 as the start date for both years (since the time between 23/03/2020-01/08/2020 and 23/03/2019-01/08/2020 are the same: 131 days),
and to create a new variable for year as exposure using 2020 and 2019 respectively, and compare time-to-event this way.
Thank you kindly for your help.
0 Response to How to use a Cox regression model for time-to-event analysis using different years as control
Post a Comment