Critical considerations, including overfitting in regression models and confounding in study designs for delirium follow-up.