Generalization in Clinical Prediction Models: The Blessing and Curse of Measurement Indicator Variables