Potential impact of changes in administrative database coding methodology on research and policy decisions: an example from the Ontario Health Insurance Plan

Big Data for a Big Problem: How Can We Enhance the Implementation of Perioperative Cardiovascular Guidelines?

“…In addition, definition and coding methods are regularly revised and can change during the time frame of a given study. 9,10 Associations between data may be unclear: for example, in the study by Gouda et al, the relationship between BNP testing and the surgical procedure was assumed if the test was done within 30 days before surgery. However, the high positivity rate suggests that in at least some cases testing may have been done for other reasons, such as follow-up of congestive heart failure, and thus inferring a relationship between test ordering and the surgical event is not necessarily valid.…”