 Negative control outcome

A negative control outcome cannot plausibly be caused by treatment, so an estimated effect can reveal bias.

Unlike treatment comparison alone, this scientific control checks for an association that should vanish under the intended causal model.