Using propensity scores to adjust for selection bias when assessing the effectiveness of Alcoholics Anonymous in observational studies