Gravar-mail: Evidence of bias and variation in diagnostic accuracy studies