Closed kmkostka closed 4 years ago
Some of these co-variates are disease/RA-related, some are in-patient admission related.
Given that we see a high correlation of admission-related codes, it is likely the index date coincide with in-patient admission date, which means the first occurrence of drug of interest was the first documentation in the system, rather than initiation of the drug. Agree with Kristin that we probably should not include this database for the analysis.
Agreed. We will not use this data source for either design.
Error: In Hospital Charge Datamaster data, there is a large number of "High correlation between covariate(s) and treatment detected."
Some of these covariates appear to be artifacts of being an in-patient centric database evaluating an outcome that is also in-patient centric.
Sample size is very low which may exacerbate this correlation. Will likely exclude this database from the analysis.
Question: Should we update the RCT package covariates to exclude?