Predicting Readmission Risk in Elderly Patients with Coronary Heart Disease
Author Information
Author(s): Luo Hanyu, Wang Benlong, Cao Rui, Feng Jun
Primary Institution: Department of Cardiology of Lu'an People's Hospital, Lu'an Hospital of Anhui Medical University, Lu'an, China
Hypothesis
To investigate the risk factors for readmission of elderly patients with coronary artery disease and to construct and validate a predictive model for readmission risk within 3 years using machine learning methods.
Conclusion
The XGBoost model effectively predicts readmission risk in elderly patients with coronary heart disease based on factors like TyG-BMI, RDW, and diabetes.
Supporting Evidence
- The XGBoost model achieved an AUC of 0.903 in predicting readmission risk.
- Diabetes, RDW, and TyG-BMI were identified as significant predictors of readmission.
- The model maintained strong calibration performance across training and testing datasets.
- External validation yielded an AUC of 0.891, confirming the model's predictive accuracy.
- Decision curve analysis showed higher net benefits for the XGBoost model in clinical decision-making.
Takeaway
Doctors can use a special computer model to figure out which older heart patients are more likely to go back to the hospital, helping them get the right care sooner.
Methodology
The study used machine learning algorithms, including Lasso regression and XGBoost, to analyze data from 718 elderly patients with coronary heart disease.
Potential Biases
Potential selection bias due to the single-center design and exclusion of certain patient data.
Limitations
The study was retrospective, conducted at a single center, and may not generalize to other populations; it also did not include certain clinical assessments that could affect readmission risk.
Participant Demographics
Elderly patients (≥60 years) with coronary heart disease, including 215 readmitted and 360 non-readmitted patients.
Statistical Information
P-Value
p<0.05
Statistical Significance
p<0.05
Digital Object Identifier (DOI)
Want to read the original?
Access the complete publication on the publisher's website