Development of a machine learning model to predict intraoperative transfusion and guide type and screen ordering.

STUDY OBJECTIVE: To develop an algorithm to predict intraoperative Red Blood Cell (RBC) transfusion from preoperative variables contained in the electronic medical record of our institution, with the goal of guiding type and screen ordering. DESIGN: Machine Learning model development on retrospective single-center hospital data. SETTING: Preoperative period and operating room. PATIENTS: The study included patients ≥18 years old who underwent surgery during 2019-2022 and excluded those who refused transfusion, underwent emergency surgery, or surgery for organ donation after cardiac or brain death. INTERVENTION: Prediction of intraoperative transfusion vs. no intraoperative transfusion. MEASUREMENTS: The outcome variable was intraoperative transfusion of RBCs. Predictive variables were surgery, surgeon, anesthesiologist, age, sex, body mass index, race or ethnicity, preoperative hemoglobin (g/dL), partial thromboplastin time (s), platelet count x 109 per liter, and prothrombin time. We compared the performances of seven machine learning algorithms. After training and optimization on the 2019-2021 dataset, model thresholds were set to the current institutional performance level of sensitivity (93%). To qualify for comparison, models had to maintain clinically relevant sensitivity (>90%) when predicting on 2022 data; overall accuracy was the comparative metric. MAIN RESULTS: Out of 100,813 cases that met study criteria from 2019 to 2021, intraoperative transfusion occurred in 5488 (5.4%) of cases. The LightGBM model was the highest performing algorithm in external temporal validity experiments, with overall accuracy of (76.1%) [95% confidence interval (CI), 75.6-76.5], while maintaining clinically relevant sensitivity of (91.2%) [95% CI, 89.8-92.5]. If type and screens were ordered based upon the LightGBM model, the predicted type and screen to transfusion ratio would improve from 8.4 to 5.1. CONCLUSIONS: Machine learning approaches are feasible in predicting intraoperative transfusion from preoperative variables and may improve preoperative type and screen ordering practices when incorporated into the electronic health record.

Artificial intelligence modelling to assess the risk of cardiovascular disease in oncology patients.

Aims: There are no comprehensive machine learning (ML) tools used by oncologists to assist with risk identification and referrals to cardio-oncology. This study applies ML algorithms to identify oncology patients at risk for cardiovascular disease for referrals to cardio-oncology and to generate risk scores to support quality of care. Methods and results: De-identified patient data were obtained from Vanderbilt University Medical Center. Patients with breast, kidney, and B-cell lymphoma cancers were targeted. Additionally, the study included patients who received immunotherapy drugs for treatment of melanoma, lung cancer, or kidney cancer. Random forest (RF) and artificial neural network (ANN) ML models were applied to analyse each cohort: A total of 20 023 records were analysed (breast cancer, 6299; B-cell lymphoma, 9227; kidney cancer, 2047; and immunotherapy for three covered cancers, 2450). Data were divided randomly into training (80%) and test (20%) data sets. Random forest and ANN performed over 90% for accuracy and area under the curve (AUC). All ANN models performed better than RF models and produced accurate referrals. Conclusion: Predictive models are ready for translation into oncology practice to identify and care for patients who are at risk of cardiovascular disease. The models are being integrated with electronic health record application as a report of patients who should be referred to cardio-oncology for monitoring and/or tailored treatments. Models operationally support cardio-oncology practice. Limited validation identified 86% of the lymphoma and 58% of the kidney cancer patients with major risk for cardiotoxicity who were not referred to cardio-oncology.