Immortal Time Bias in Observational Studies.