A decision tree model for predicting restenosis after PCI based on clinical characteristics in patients with coronary heart disease
Du Jianqing
Zhang Kangjian
Guo Erbing
Gong Linghui
Abstract：Objective To establish a decision tree model for predicting restenosis after percutaneous coronary intervention (PCI based on clinical characteristics in patients with coronary heart disease (CHD).Methods CHD patients undergone PCI (n=120) were chosen from Department of Cardiology in People's Hospital of Pudong New District of Shanghai City from July 2023 to Mar.2024.The patients were divided,according to restenosis occurrence status after PCI,into restenosis group (n=27) and non-restenosis group (n=93).The basic clinical materials were compared between 2 groups.The independent risk factors for restenosis occurrence after PCI were analyzed by using binary Logistic regression model.A decision tree model was established and its predictive value to restenosis was analyzed by using ROC curve.Results There were significant differences in smoking,diabetes,stent number,stent diameter,uric acid (UA),high-sensitivity C-reactive protein (hs-CRP),lipoprotein a[Lp(a)],total cholesterol (TC) and low-density lipoprotein (LDL) between 2 groups (P＜0.05).The results of binary Logistic regression analysis showed that complicated diabetes,higher UA level,hs-CRP＞10 mg/L and Lp(a)＞300 mg/L were independent risk factors for restenosis occurrence after PCI in CHD patients (P＜0.05),which could increase the risk by 7.09 times,5.93 times,9.22 times and 5.17 times,respectively.The results of decision tree model showed that the percentage of restenosis occurrence after PCI was higher in patients with hs-CRP＞10 mg/L than that in those with hs-CRP≤10 mg/L.At the node of hs-CRP＞10 mg/L,the percentage of restenosis occurrence after PCI was higher in patients with Lp(a)＞300 mg/L than that in those with Lp(a)≤300 mg/L,at the nodes of Lp(a)≤300 mg/L and hs-CRP≤10 mg/L,the percentage of restenosis occurrence after PCI was higher in patients with higher UA level than that in those with normal UA level,and at the nodes of hs-CRP≤10 mg/L and higher UA level,the percentage of restenosis occurrence after PCI was higher in patients with complicated diabetes than that in those without complicated diabetes.The AUC of Logistic regression model was 0.884,and AUC of decision tree model was 0.862.After taken cut-off value,sensitivity of Logistic regression model was 0.926 and specificity was 0.710,and sensitivity of decision tree model was 0.852 and specificity was 0.817.The results of Delong test showed that there was no significant difference in AUC between 2 models (Z=0.793,P=0.428).Conclusion The independent risk factors for restenosis occurrence after PCI include complicated diabetes,higher UA level,hs-CRP＞10 mg/L and Lp(a)＞300 mg/L,and the decision tree model based on above factors can accurately predict restenosis occurrence after PCI in CHD patients.
Keywords：Coronary heart diseasePercutaneous coronary interventionRestenosisDecision tree model
Publication Date：2024-12-28
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：5( 1538-1542 )