New Statistical Methods For Using Confounders Measured In A Validation Sample To Enhance Time-To-Event Analyses Of Large Databases