Confounding can be eliminated by all except-
|D|| Multivariate analysis|
- Blinding is a method used most often in experimental studies to remove measurement bias.
- All the other options mentioned in the question are actually methods to remove confounding from the study.
- Matching & multivariate analysis are methods used exclusively for removing the effect of confounding from the study while randomization plays two important roles:
- Purpose of randomization in a Randomized Controlled trial
a. If we randomize properly, we achieve non predictability of the next assignment, we do not have to worry that any subjective biases of the investigators, either overt or covert, may be introduced into the process of selecting patients for one treatment group or the other (i.e. selection bias is eliminated from the study).
b. Also, in long run, randomization will increase the likelihood that the groups will be comparable in regard to characteristics about which we may be concerned (i.e. reduce confounding), such as sex, age, race and severity of disease, and that may affect prognosis. However, randomization is not a guarantee of comparability, because chance may playa role in the process, but over the long term, the groups will tend to be similar.