INTRODUCTION: Every experienced surgeon has a patient whose life was saved by a blood transfusion (the "good"). In contrast, an overwhelming amount of evidence suggests that perioperative blood transfusion can be associated with adverse surgical outcomes (the "bad"). We wondered what patient characteristics, if any, can explain this clinical dichotomy with certain patients benefiting from transfusion, whereas others are harmed by this intervention.

METHODS: We queried the American College of Surgeons National Surgical Quality Improvement Project database containing patient information entered between 2010 and 2012 to identify differences in mortality and morbidity among patients receiving blood transfusion within 72 hours of their operative procedure compared with those who did not receive any blood. We calculated the relative risk of developing a serious complication or of operative mortality in propensity-stratified patients with equivalent predicted risk of developing a serious complication or operative mortality.

RESULTS: There were 470,407 patients in the study group. Of these, 32,953 patients (7.0%) received at least a single blood transfusion within 72 hours of operation. The percent of transfused patients who died or developed serious morbidity was 11.3% and 55.4% compared with 1.3% and 6.1% in nontransfused patients (both P < .001). Operative mortality, rates of failure to rescue, and serious postoperative complications are increased in patients who receive a postoperative transfusion, both in unadjusted comparisons and in propensity-matched comparisons. Dividing patients into regression-stratified deciles with equal numbers of deaths in each group found that patients at the greatest risk for development of death or serious complications had nonsignificant risk of harm from blood transfusion, whereas patients in the least risk deciles had between an 8- and 12-fold increased risk of major adverse events associated with transfusion.

CONCLUSION: We found that high-risk patients do not have a significant risk from blood transfusion, but low-risk patients have between an 8- and 10-fold excess risk of adverse outcomes when they receive a blood transfusion. We speculate that careful preoperative assessment of transfusion risk and intervention based on this assessment could minimize operative morbidity and mortality, especially because the patients at least risk are more likely to undergo elective operations and provide time for therapeutic interventions to improve transfusion risk profiles.