Gravar-mail: Using clinical variables and drug prescription data to control for confounding in outcome comparisons between hospitals