Keywords: observational data, drug safety, bias
Some questions cannot be answered with just one study (randomized or observational) or even meta-analyses of them. Observational databases can help with questions such as rare events, but they vary widely in their original purpose, collection mechanisms, verification, and hence, appropriateness to answer any given drug safety question. We focus here on the potential relations between collection mechanisms and appropriateness for inquiry as well as potential biases.
By taking a granular look at an example, we can see how data may fit the purposes of one query but not another. We also use an example to explore some issues that arise if we compare how data are collected for say a randomized control trial versus a records database such as electronic health records or a spontaneous reporting database.