Misuse of Regression Adjustment for Additional Confounders Following Insufficient Propensity-Score Balancing.