201. Patient-Reported Outcome–Based Prediction for Postdischarge Complications after Lung Surgery.
- Author
-
Yang, Ding, Wei, Xing, Hong, Qian, Zhao, Chenguang, and Mu, Juwei
- Subjects
LUNG surgery ,MACHINE learning ,TUMOR surgery ,HOSPITAL admission & discharge ,LOGISTIC regression analysis - Abstract
Background Patients undergoing lung tumor surgery may experience various complications after discharge from the hospital. Using patient-reported outcomes (PROs), this study attempted to identify relevant indicators of postdischarge complications after lung tumor surgery and develop a predictive nomogram model to evaluate the risk for individual patients. Methods Patients who underwent lung tumor surgery between December 2021 and June 2022 were included in this study. PROs were assessed using the Perioperative Symptom Assessment for Lung Surgery scale and were assessed preoperatively at baseline, on postoperative day 1 (POD1) 1 to POD4, and then weekly until the fourth week. A random forest machine learning prediction model was built to rank the importance of each PRO score of patients on POD1 to POD4. We then selected the top 10 variables in terms of importance for the multivariable logistic regression analysis. Finally, a nomogram was developed. Results PROs, including coughing (POD3 and POD4), daily activity (POD1), and pain (POD1 and POD2), were associated with postdischarge complications in patients undergoing lung tumor surgery. The predictive model showed good performance in estimating the risk of postdischarge complications, with an area under the curve of 0.833 (95% confidence interval: 0.753–0.912), while maintaining good calibration and clinical value. Conclusion We found that PRO scores on POD1 to POD4 were associated with postdischarge complications after lung tumor surgery, and we developed a helpful nomogram model to predict the risk of postdischarge complications. [ABSTRACT FROM AUTHOR]
- Published
- 2023
- Full Text
- View/download PDF