D H Howard1, A Karcz2, J D Roback3. 1. Department of Health Policy and Management, Emory University, Atlanta, GA, USA. 2. Institute for Health Metrics Burlington, MA, USA. 3. Department of Pathology and Laboratory Medicine, Center for Transfusion and Cellular Therapies, Emory University, Atlanta, GA, USA.
Abstract
OBJECTIVE: To determine if billing records accurately report the receipt of red blood cell transfusions. BACKGROUND: Many red blood transfusions are given inappropriately, but efforts to monitor transfusion rates are hampered by a lack of data that facilitate benchmarking. METHODS: Using billing records and electronic medical records from 53 community hospitals, we estimated the sensitivity and specificity of billing records for measuring the receipt of transfusions in patients undergoing surgery for hip fractures. RESULTS: The sample included 12 091 patients, of whom 5215 received red blood cell transfusions according to electronic medical records. The sensitivity of billing data for measuring transfusions was 71·6% (95% CI: 60·4-82·8%). The specificity was 92·6% (95% CI: 88·3-97·0%). CONCLUSIONS: Researchers can use billing records to measure transfusion rates but should consider excluding hospitals with very low transfusion rates, which may indicate that the hospitals do not accurately report transfusions in billing data.
OBJECTIVE: To determine if billing records accurately report the receipt of red blood cell transfusions. BACKGROUND: Many red blood transfusions are given inappropriately, but efforts to monitor transfusion rates are hampered by a lack of data that facilitate benchmarking. METHODS: Using billing records and electronic medical records from 53 community hospitals, we estimated the sensitivity and specificity of billing records for measuring the receipt of transfusions in patients undergoing surgery for hip fractures. RESULTS: The sample included 12 091 patients, of whom 5215 received red blood cell transfusions according to electronic medical records. The sensitivity of billing data for measuring transfusions was 71·6% (95% CI: 60·4-82·8%). The specificity was 92·6% (95% CI: 88·3-97·0%). CONCLUSIONS: Researchers can use billing records to measure transfusion rates but should consider excluding hospitals with very low transfusion rates, which may indicate that the hospitals do not accurately report transfusions in billing data.