Bias that occurs when past treatments affect future confounders, making it hard to isolate treatment effects in sequential decisions.