BACKGROUND AND OBJECTIVES: Continuous monitoring of blood use and feedback on transfusions are effective in decreasing inappropriate blood transfusions. However, traditional methods of monitoring have practical challenges, such as the limited availability of experts and funding. Administrative data including a patient classification system may be employed for risk-adjusted assessment of hospital-wide blood use. MATERIALS AND METHODS: We conducted an audit of blood use at two hospitals and determined proportions of appropriate blood use at each hospital. We then used administrative data of 587,045 cases provided by 73 hospitals to develop two mathematical models to calculate risk-adjusted use of blood products. The first model is a logistic regression model to predict the percentage of transfused patients. Patient demographics, surgery and diagnostic groups were utilized as predictors of transfusion. The second model is a case-mix adjusted model which predicts hospital-wide use of units of blood products from the distribution of diagnosis-related groups. For each model, the observed to expected (O/E) ratio of blood use in each hospital was calculated. We compared resultant ratios with proportions of appropriate blood use in two of the hospitals studied. RESULTS: Both models showed good prediction abilities. O/E ratios calculated using the two models were relevant to proportions of appropriate transfusions. CONCLUSIONS: Risk-adjusted assessments of blood product use based on administrative data allow hospital-wide evaluation of transfusion use. Comparing blood use between different hospitals contributes toward establishing appropriate transfusion practices.
BACKGROUND AND OBJECTIVES: Continuous monitoring of blood use and feedback on transfusions are effective in decreasing inappropriate blood transfusions. However, traditional methods of monitoring have practical challenges, such as the limited availability of experts and funding. Administrative data including a patient classification system may be employed for risk-adjusted assessment of hospital-wide blood use. MATERIALS AND METHODS: We conducted an audit of blood use at two hospitals and determined proportions of appropriate blood use at each hospital. We then used administrative data of 587,045 cases provided by 73 hospitals to develop two mathematical models to calculate risk-adjusted use of blood products. The first model is a logistic regression model to predict the percentage of transfused patients. Patient demographics, surgery and diagnostic groups were utilized as predictors of transfusion. The second model is a case-mix adjusted model which predicts hospital-wide use of units of blood products from the distribution of diagnosis-related groups. For each model, the observed to expected (O/E) ratio of blood use in each hospital was calculated. We compared resultant ratios with proportions of appropriate blood use in two of the hospitals studied. RESULTS: Both models showed good prediction abilities. O/E ratios calculated using the two models were relevant to proportions of appropriate transfusions. CONCLUSIONS: Risk-adjusted assessments of blood product use based on administrative data allow hospital-wide evaluation of transfusion use. Comparing blood use between different hospitals contributes toward establishing appropriate transfusion practices.