Selection bias found in interpreting analyses with missing data for the prehospital index for trauma.