Simply Adjusting for Schedulers' Bias in Estimated Case Durations Can Accomplish the Same Objectives of Improving Predictions as Use of Machine Learning.