Andrew W J Flint1,2,3,4, Michael Bailey2, Christopher M Reid5,6,7, Julian A Smith8,9,10, Lavinia Tran5,7, Erica M Wood1,8, Zoe K McQuilten1,2,8, Michael C Reade2,11,12. 1. Transfusion Research Unit, Department of Epidemiology and Preventive Medicine, Monash University, Melbourne, Australia. 2. The Australian and New Zealand Intensive Care Research Centre (ANZIC-RC), School of Public Health and Preventive Medicine, Monash University, Melbourne, Australia. 3. Royal Australian Navy, Australia. 4. Intensive Care Unit, Royal Darwin Hospital, Tiwi, Northern Territory, Australia. 5. School of Public Health and Preventive Medicine, Monash University, Melbourne, Australia. 6. School of Public Health, Curtin University, Perth, Australia. 7. Centre of Cardiovascular Research and Education in Therapeutics (CCRET), School of Public Health and Preventive Medicine, Monash University, Melbourne, Australia. 8. Monash Health, Clayton, Victoria, Australia. 9. Department of Surgery (School of Clinical Sciences at Monash Health), Monash University and Department of Cardiothoracic Surgery, Monash Health, Clayton, Victoria, Australia. 10. Chairman, Research Committee, Australian and New Zealand Society of Cardiac and Thoracic Surgeons (ANZSCTS); Cardiac Surgery Database. 11. Joint Health Command, Australian Defence Force, Canberra, Australia. 12. Faculty of Medicine, University of Queensland, Brisbane, Australia.
Abstract
Platelet (PLT) transfusions are limited and costly resources. Accurately predicting clinical demand while limiting product wastage remains difficult. A PLT transfusion prediction score was developed for use in cardiac surgery patients who commonly require PLT transfusions. STUDY DESIGN AND METHODS: Using the Australian and New Zealand Society of Cardiac and Thoracic Surgeons National Cardiac Surgery Database, significant predictors for PLT transfusion were identified by multivariate logistic regression. Using a development data set containing 2005 to 2016 data, the Australian Cardiac Surgery Platelet Transfusion (ACSePT) risk prediction tool was developed by assigning weights to each significant predictor that corresponded to a probability of PLT transfusion. The predicted probability for each score was compared to actual PLT transfusion occurrence in a validation (2017) data set. RESULTS: The development data set contained 38 independent variables and 91 521 observations. The validation data set contained 12 529 observations. The optimal model contained 23 variables significant at P < .001 and an area under the receiver operating characteristic (ROC) curve of 0.69 (95% confidence interval [CI], 0.68-0.69). ACSePT contained nine variables and had an area under the ROC curve of 0.66 (95% CI, 0.65-0.66) and overall predicted probability of PLT transfusion of 19.8% for the validation data set compared to an observed risk of 20.3%. CONCLUSION: The ACSePT risk prediction tool is the first scoring system to predict a cardiac surgery patient's risk of receiving a PLT transfusion. It can be used to identify patients at higher risk of receiving PLT transfusions for inclusion in clinical trials and by PLT inventory managers to predict PLT demand.
Platelet (PLT) transfusions are limited and costly resources. Accurately predicting clinical demand while limiting product wastage remains difficult. A PLT transfusion prediction score was developed for use in cardiac surgery patients who commonly require PLT transfusions. STUDY DESIGN AND METHODS: Using the Australian and New Zealand Society of Cardiac and Thoracic Surgeons National Cardiac Surgery Database, significant predictors for PLT transfusion were identified by multivariate logistic regression. Using a development data set containing 2005 to 2016 data, the Australian Cardiac Surgery Platelet Transfusion (ACSePT) risk prediction tool was developed by assigning weights to each significant predictor that corresponded to a probability of PLT transfusion. The predicted probability for each score was compared to actual PLT transfusion occurrence in a validation (2017) data set. RESULTS: The development data set contained 38 independent variables and 91 521 observations. The validation data set contained 12 529 observations. The optimal model contained 23 variables significant at P < .001 and an area under the receiver operating characteristic (ROC) curve of 0.69 (95% confidence interval [CI], 0.68-0.69). ACSePT contained nine variables and had an area under the ROC curve of 0.66 (95% CI, 0.65-0.66) and overall predicted probability of PLT transfusion of 19.8% for the validation data set compared to an observed risk of 20.3%. CONCLUSION: The ACSePT risk prediction tool is the first scoring system to predict a cardiac surgery patient's risk of receiving a PLT transfusion. It can be used to identify patients at higher risk of receiving PLT transfusions for inclusion in clinical trials and by PLT inventory managers to predict PLT demand.