There are a number of limitations of this study.
This is a retrospective analysis with lack of randomization, potential
severity bias, and likely population mismatch. Analysis of administrative data depends on identification of relevant
records using ICD-9 codes rather than more accurate clinical data