Background: Anemia has been associated with increased postoperative morbidity and mortality. We used the American College of Surgeons National Surgical Quality Improvement Program database to retrospectively assess the relationship between preoperative anemia and 30-day postoperative mortality and morbidity in noncardiac surgical patients, careful to distinguish confounding variables from mediator variables.
Methods: Each patient with preoperative anemia was matched to one without anemia using propensity matching on potentially confounding baseline variables. Logistic regression was used to evaluate the relationship between preoperative anemia and 30-day postoperative mortality and morbidity. The primary hypothesis was evaluated after adjusting for covariables showing residual imbalance after matching.
Results: Within the database, 574,860 of 971,455 surgical cases met our inclusion criteria, and among those 145,218 (25.3%) were anemic at baseline. The unadjusted odds ratio (95% confidence interval) for 30-day mortality comparing anemic patients with nonanemic patients was 4.69 (4.01-5.49). Among the propensity-matched group of 238,596 patients, the total effect (i.e., not adjusting for mediator variables) of preoperative anemia was estimated as an odds ratio of 1.59 (1.42-1.78). After adjusting for suspected mediator variables, preoperative anemia was only weakly associated with an odds ratio of 1.24 (1.10-1.40) for 30-day mortality.
Conclusion: Preoperative anemia appears to be associated with baseline diseases that markedly increase mortality. Anemia per se is a rather weak independent predictor of postoperative mortality. Our analysis also illustrates how analyzing large variable-rich registries challenges investigators to discriminate between confounding variables and mediator variables, i.e., factors that might be considered as "causal pathways" for the effect of the exposure or intervention on outcome.