Pre-randomization and de-randomization in emergency medical research: new names and rigorous criteria for old methods.