Estimating the prevalence of anemia and its association with perioperative blood transfusion in patients who have no haemoglobin measurement using multiple imputation and inverse probability of treatment weighting

Abstract

Anemic patients are at increased risk of intraoperative and postoperative transfusion, infection and mortality. Anemia is common in surgical patients, previous studies report a prevalence of 30-40%. While patient blood management programs advise screening for anemia, other perioperative initiatives advocate limiting preoperative laboratory testing. Therefore, a large number of surgical patients may not be screened, missing the opportunity for further investigation and management. In our single-center retrospective cohort study of all adult surgical patients over one year, we sought to estimate the proportion of anemia in patients without a preoperative blood test and their associated intraoperative and postoperative transfusion risk using multiple imputation to estimate missing hemoglobin values and inverse-probability-of-treatment-weighting to balance the study groups. Multiple imputation generated 39 datasets containing plausible hemoglobin values for patients missing a hemoglobin. Patients were classified as known-anemic, known-not-anemic, imputed-anemic and imputed-not-anemic if the hemoglobin was measured or imputed. We then compared red-cell transfusion yes/no by logistic regression, and number of red-cell units transfused by linear regression, pooling the results. Thirty-nine percent of 7879 patients were missing a hemoglobin, of whom 20-35% may be anemic. Only known-anemic patients had an increased intraoperative transfusion risk (odds ratio (OR) 3.13, 95% confidence interval (95%CI) 2.19 to 4.48). Both known-anemic and imputed-anemic patients had increased postoperative transfusion risk (OR 3.93 (95%CI 2.92 to 5.29), OR 2.99 (95%CI 1.61 to 5.56), respectively). Anemia may be common in patients missing a haemoglobin, and may confer an increased risk of postoperative transfusion compared to known-anemic patients. Removing them from observation studies on anemia risks sample bias. Excluding patients from pre-operative testing may miss the opportunity to identify anemia and modify postoperative transfusion risk and other adverse health outcomes. Future research could use point-of-care tests to reduce screening costs, and further investigate adverse outcomes risks in these missing patients.