A preoperative risk score to predict red blood cell transfusion in patients undergoing hysterectomy for ovarian cancer

Background: Patients with ovarian cancer experience a high rate of anemia throughout their treatment course, with rates that range from 19–95%. Blood transfusions offer symptom relief but may be costly, are limited in supply, and have been associated with worse 30-day surgical morbidity and mortality rates. Objective: The purpose of this study was to identify risk factors for blood transfusion with packed red blood cell and to develop a transfusion risk score to identify patients who undergo surgery for ovarian cancer and who are at lowest risk for a blood transfusion. Our aim was to help clinicians identify those patients who may not require a crossmatch to encourage resource use and cost-savings. Study Design: This is a retrospective database cohort study of 3470 patients who underwent hysterectomy for ovarian cancer with the use the National Surgical Quality Improvement Program database from 2014–2016. The association between risk factors with respect to 30-day postoperative blood transfusion was modeled with the use of logistic regression. A risk score to predict blood transfusion was created. Results: Eight hundred ninety-one (25.7%) patients received a blood transfusion. In multivariate analysis, blood transfusion was associated independently with age (odds ratio, 1.90, P<.01), African American race (odds ratio, 2.30; P<.01), ascites (odds ratio, 1.89; P=.02), preoperative hematocrit level <30% (odds ratio, 10.70; P<.01), preoperative platelet count >400×10⁹/L (odds ratio, 1.75; P<.01), occurrence of disseminated cancer (odds ratio, 1.71; P<.01), open surgical approach (odds ratio, 7.88; P<.01), operative time >3 hours (odds ratio, 2.19; P<.01), and additional surgical procedures that included large bowel resection (odds ratio, 4.23; P<.01), bladder/ureter resection (odds ratio, 1.69; P=.02), and pelvic exenteration (P=.02). A preoperative risk score that used age, race, ascites, preoperative hematocrit level, platelets, presence of disseminated cancer, planned hysterectomy approach, and procedures accurately predicted blood transfusion with good discriminatory ability (C-statistic=0.80 [P<.001]; C-statistic=0.69 [P<.001] for derivation and validation datasets, respectively) and calibration (Hosmer-Lemeshow goodness-of-fit, P=.081; P=.56 for derivation and validation datasets, respectively). Conclusion: Patients who undergo hysterectomy for ovarian cancer experience a high incidence of blood transfusions in the perioperative period. Preoperative risk factors and planned surgical procedures can be used in our transfusion risk score to help predict anticipated blood requirements.