Machine Learning Model may help Identify High-Risk Surgical Patients Using electronic health records
A recent study by Aman Mahajan and team at the University of Pittsburgh Medical Center (UPMC) has shown that an automated machine-learning model, utilizing data solely from electronic health records (EHRs), can accurately predict which patients are at a heightened risk of adverse outcomes from surgery. The findings were published in Journal of American Medical Association.
The study, which involved 1,477,561 patients undergoing surgery at 20 community and tertiary care hospitals within the UPMC health network, comprised three phases. Firstly, the model was built and validated using retrospective data, then its accuracy was tested on another set of retrospective data. Finally, the model was prospectively validated in clinical care.
The researchers employed a gradient-boosted decision tree machine learning method to develop the preoperative surgical risk prediction tool. To ensure interpretability and further validation of the model, the Shapley additive explanations method was used. The performance of the UPMC model was compared with the National Surgical Quality Improvement Program (NSQIP) surgical risk calculator in predicting mortality.
The results demonstrated the efficacy of the machine-learning model in identifying high-risk patients. The area under the receiver operating characteristic curve (AUROC) for mortality was 0.972 for the training set and 0.946 for the test set. In terms of major adverse cardiac and cerebrovascular events (MACCEs) and mortality, the AUROC was 0.923 for the training set and 0.899 for the test set. In the prospective evaluation, the AUROC for mortality was 0.956.
Furthermore, the model outperformed the NSQIP tool in terms of AUROC, specificity, and accuracy. The AUROC for the UPMC model was 0.945 compared to 0.897 for the NSQIP tool. The specificity and accuracy of the UPMC model were also superior, with values of 0.87 and 0.85, respectively, compared to 0.68 and 0.69 for the NSQIP tool.
The findings of this study hold significant implications for improving surgical outcomes. By accurately identifying patients at high risk of adverse events using only preoperative variables from EHRs, clinicians can provide individualized perioperative care and interventions. This tailored approach has the potential to enhance patient outcomes and reduce the incidence of adverse events associated with surgery.
Source:
Mahajan, A., Esper, S., Oo, T. H., McKibben, J., Garver, M., Artman, J., Klahre, C., Ryan, J., Sadhasivam, S., Holder-Murray, J., & Marroquin, O. C. (2023). Development and Validation of a Machine Learning Model to Identify Patients Before Surgery at High Risk for Postoperative Adverse Events. In JAMA Network Open. American Medical Association (AMA). https://doi.org/10.1001/jamanetworkopen.2023.22285
Disclaimer: This website is primarily for healthcare professionals. The content here does not replace medical advice and should not be used as medical, diagnostic, endorsement, treatment, or prescription advice. Medical science evolves rapidly, and we strive to keep our information current. If you find any discrepancies, please contact us at corrections@medicaldialogues.in. Read our Correction Policy here. Nothing here should be used as a substitute for medical advice, diagnosis, or treatment. We do not endorse any healthcare advice that contradicts a physician's guidance. Use of this site is subject to our Terms of Use, Privacy Policy, and Advertisement Policy. For more details, read our Full Disclaimer here.
NOTE: Join us in combating medical misinformation. If you encounter a questionable health, medical, or medical education claim, email us at factcheck@medicaldialogues.in for evaluation.