Confounding by indication refers to a certain type of confounding event that can bias observational studies.
Patient selection: patient undergoing a treatment.
Mechanism of bias:
(1) the treatment is given for one or more indications
(2) the indication is predictive of an outcome
In an observational study analysis may show suggest that the treatment is predictive of the outcome. There may be a relationship, but it demonstrating this depends on a careful control of many factors.
This form of bias can be avoided by:
(1) randomized clinical trial
(2) carefully control all other factors
The problem with randomized clinical trials is that denying a patient a potentially beneficial therapy may be unethical. On the other hand, it may be difficult to control for all variables.
To read more or access our algorithms and calculators, please log in or register.
