A retrospective evaluation of the risk of bias in perioperative temperature metrics