Published online Nov 26, 2022. doi: 10.4330/wjc.v14.i11.565
Peer-review started: June 15, 2022
First decision: August 1, 2022
Revised: September 18, 2022
Accepted: October 18, 2022
Article in press: October 18, 2022
Published online: November 26, 2022
Risk stratification tools exist for patients presenting with chest pain to the emergency room and have achieved the recommended negative predictive value (NPV) of 99%. However, the current stratification tools result in unnecessary clinical interventions due to a low positive predictive value (PPV).
Healthcare costs are astronomical in the United States, including for patients who present with chest pain. These costs emphasize the need for a better stratification tool to safely identify patients who present with chest pain for early discharge without unnecessary testing.
This study aimed to create a machine learning model (MLM) for risk stratification of chest pain patients with a better PPV while maintaining an NPV of 99%.
This retrospective cohort study used demographics, coronary artery disease history, hypertension, hyperlipidemia, diabetes mellitus, chronic kidney disease, obesity, and smoking as the covariates for the prediction of an abnormal cardiac stress test (CST). Binomial regression (BR), random forest, and XGBoost MLMs were used for prediction. Bootstrapping was used for the internal validation of the prediction models.
The XGBoost MLM had the best PPV of 24.33%, with an NPV of 91.34% for abnormal CST. The BR MLM had a PPV of 21.34% and an NPV of 92.55%. The random forest MLM had a PPV of 20.55% and an NPV of 90.24%.
The XGBoost MLM provided a better PPV than currently used stratification tools (24.33% vs 13.00%-17.50%). Though the NPV from the XGBoost MLM remained lower than the recommended value of 99%, it highlights the potential use of MLMs in clinical decision-making.
Data sharing and external validation of the MLMs will be crucial for their recognition and widespread adoption.