Daniel Hall, Brigham Young University
Life Sciences
OBJECTIVE(S): Predictive analytics (PA) is increasingly being used in the delivery of healthcare. Whether PA can improve patient handoffs on a busy surgical service is unknown. This study aims to determine if predictive models for acute cellular rejection (ACR) episodes and biliary complications after orthotropic liver transplantation (OLT) can be built in order to improve patient care.
METHODS: Recipient, donor, procurement, and operative data; induction immunosuppressive therapy; and laboratory values for 90 days were collected as predictor variables from 386 OLT patient records transplanted during a 5 year period. From multiple predictive algorithms the best model was chosen to give risk scores. Patient records were assigned to separate risk groups for biliary complications and ACR episodes. Sensitivity and specificity analysis was performed on the risk groups.
RESULTS: Biliary complications were detected in 51.6% of records in the high risk group, 18.9% in the intermediate risk group, and 0% in the low risk group. An ACR episode was detected in 28.1% of records in the high risk group, 6.7% in the intermediate risk group, and 0% in the low risk group. Sensitivity and specificity analysis with the intermediate and low risk groups combined revealed sensitivities of 60% and 69% and specificities of 87% and 77% for biliary complications and ACR episodes, respectively.
CONCLUSIONS: PA can effectively risk stratify liver transplant patients for biliary complications and/or ACR episodes. Such predictive models can aid in clinical decision making by communicating in real time a patient’s probable risks in patient handoffs.