FRIDAY, Sept. 22, 2023 (HealthDay News) — Machine learning models based on top biomarkers can predict mortality in patients with femoral neck fractures, study published online Sept. 20 it was done. Orthopedic Research Journal.
George Asrian, MD, PhD, of the University of Pennsylvania in Philadelphia, and his colleagues show that a machine learning model trained on basic blood and laboratory data, as well as basic demographic data, predicts mortality rates after femoral neck fractures. We investigated whether it could be predicted. The key variables most associated with 1-, 5-, and 10-year mortality were identified and their clinical significance investigated. Data from 3,751 hip fracture patients were included.
The researchers found that the one-year mortality rate was 21 percent for all patients studied and 29 percent for patients over 80 years old. Ten different machine learning classification models were evaluated. LightGBM showed the strongest predictive performance for his 1-year mortality, with an accuracy of 91% in the test set, area under the receiver operating characteristic curve of 0.79, and sensitivity and specificity of 0.34 and 0.98, respectively. For the 1-year model, the most strongly weighted features include age, glucose, red blood cell distribution width, mean corpuscular hemoglobin concentration, white blood cells, urea nitrogen, prothrombin time, platelet count, calcium level, and partial thromboplastin time . The top 10 features in LightGBM’s 5-year and 10-year predictive models included most of these features.
“LightGBM is a robust and powerful tool for predicting mortality over short- and long-term time frames, allowing easy analysis of the most important input variables,” the authors wrote.