Published in:
01-11-2019 | Myocardial Infarction | IM - ORIGINAL
Identifying patients with refusal of percutaneous coronary intervention for acute myocardial infarction: a classification and regression tree analysis
Authors:
Manyan Wu, Long Li, Sufang Li, Yuxia Cui, Dan Hu, Junxian Song, Chongyou Lee, Hong Chen
Published in:
Internal and Emergency Medicine
|
Issue 8/2019
Login to get access
Abstract
The purpose of the present study is to develop and validate a prediction tool to identify patients who refuse to receive percutaneous coronary intervention (PCI) rapidly. We developed a risk stratification model using the derivation cohort of 288 patients with ST segment elevation myocardial infarction (STEMI) in our hospital and validated it in a prospective cohort of 115 patients. There were 52 (18.1%) patients and 18 (15.7%) patients who refused PCI among derivation and validation cohort, respectively. A classification and regression tree (CART) analysis and multivariate logistic regression were used for statistical analysis. The decision-making factors for refusal of PCI were also investigated. The CART analysis and logistic regression both showed that self-rated mild symptom was the most significant predictor of not choosing PCI. The model generated three risk groups. The high-risk group included: self-rated mild symptoms; self-rated severe symptom, glomerular filtration rate < 60 ml/min/1.73m2. The intermediate-risk group included: self-rated severe symptom, glomerular filtration rate ≥ 60 ml/min/1.73m2 and age ≥ 75 years. The low-risk group included: self-rated severe symptom, glomerular filtration rate ≥ 60 ml/min/1.73m2 and age < 75 years. The prevalence for refusal of PCI of the three groups were 45%–44%, 18% and 4%, respectively. The sensitivity was 88% and the negative predictive value was 96%. And similar results were obtained when this prediction tool was applied prospectively to the validation cohort. Patients at low and high risk can be easily identified for refusal of PCI by the prediction tool using common clinical data. This practical model might provide useful information for rapid recognition and early response for this kind of crowd.