Observational studies - should we simply ignore them in assessing transfusion outcomes?