Prediction and risk factor analysis of in-stent restenosis and revascularization after coronary stenting based on machine learning.

Abstract

Background Effective prediction of in-stent restenosis and revascularization after coronary stent implantation and interventions targeting risk factors that may lead to these events in patients are crucial for their prevention and management. Methods Based on a C5.0 decision tree approach, data from 2,326 patients from two centers were included. We comprehensively analyzed 34 risk factors that may affect in-stent restenosis and revascularization after stent implantation and conducted predictions and risk factor analyses for in-stent restenosis and revascularization following coronary stent implantation. Results The accuracy of predicting in-stent restenosis following coronary stent implantation with a median follow-up period of 30 months was as follows: area under the curve (AUC) in the training set; 0.996, AUC in the internal validation set; 0.988, and AUC in the external validation set; 0.889, with an f1 value of 0.95, a sensitivity of 99.16%, and a specificity of 91.72%. Additionally, the accuracy of revascularization prediction was as follows: AUC in the training set; 0.984, AUC in the internal validation set; 0.956, and AUC in the external validation set; 0.876, with an f1 value of 0.84, a sensitivity of 96.43%, and a specificity of 25%. We also conducted a risk factor analysis. Conclusion We successfully constructed a predictive and risk factor analysis model for in-stent restenosis and revascularization following coronary stent implantation. This model may be helpful for clinical decision-making.