How to mitigate confounding factors in observational studies