Comparison of Statistical Approaches for Dealing With Immortal Time Bias in Drug Effectiveness Studies