Machine learning algorithms using the inflammatory prognostic index for contrast-induced nephropathy in NSTEMI patients

<b>Aim:</b> Inflammatory prognostic index (IPI), has been shown to be related with poor outcomes in cancer patients. We aimed to investigate the predictive role of IPI for contrast-induced nephropathy (CIN) development in non-ST segment elevation myocardial infarction patients using a nomogram and performing machine learning (ML) algorithms. <b>Materials &amp; methods:</b> A total of 178 patients with CIN (+) and 1511 with CIN (-) were included. <b>Results:</b> CIN (+) patients had higher IPI levels, and IPI was independently associated with CIN. A risk prediction nomogram including IPI had a higher predictive ability and good calibration. Naive Bayes and k-nearest neighbors were the best ML algorithms for the prediction of CIN patients. <b>Conclusion:</b> IPI might be used as an easily obtainable marker for CIN prediction using ML algorithms. Contrast-induced nephropathy (CIN) is a significant complication that can arise from the use of contrast media during diagnostic or invasive procedures. Inflammatory markers such as the neutrophil-to-lymphocyte ratio and the C-reactive protein-to-albumin ratio were previously identified as good predictors of CIN, leading to the idea that the combination of these markers might be more useful for predicting CIN development than each parameter individually. A novel inflammatory marker (inflammatory prognostic index), which is the product of neutrophil/lymphocyte ratio (NLR) and C-reactive protein/albumin ratio (CAR) was an independent predictor of CIN in non-ST segment elevation myocardial infarction patients (NSTEMI). A nomogram including inflammatory-prognostic index (IPI) had high predictive and calibration abilities for estimating the risk of CIN development. IPI had higher discriminative ability than both NLR and CAR for discrimination of CIN (+) patients from CIN (-) ones. Naive-Bayes and k-nearest neighbors were the best algorithms for prediction of CIN development in NSTEMI. As a prediction tool machine learning (ML) algorithms including IPI, especially k-nearest neighbors might have net clinical benefit for detecting CIN development risk. In recent years, ML algorithms have been shown to better predict worse cardiovascular outcomes by analyzing the available healthcare data. The present study revealed that the ML algorithms might accurately predict the CIN developments when combined with well-known inflammatory marker in NSTEMI patients. A nomogram including IPI might be useful as a prediction tool for CIN development after coronary interventions in NSTEMI patients.

目的：炎症预后指数（Inflammatory prognostic index，IPI）已被证实与癌症患者的不良预后相关。本研究旨在通过列线图（nomogram）及机器学习（machine learning，ML）算法，探究IPI对非ST段抬高型心肌梗死（non-ST segment elevation myocardial infarction）患者对比剂诱导肾病（Contrast-induced nephropathy，CIN）发生的预测价值。

材料与方法：共纳入178例CIN阳性患者及1511例CIN阴性患者。

结果：CIN阳性患者的IPI水平更高，且IPI与CIN独立相关。包含IPI的风险预测列线图具有较高的预测效能及良好的校准度。朴素贝叶斯（Naive Bayes）和k近邻算法（k-nearest neighbors）是预测CIN患者的最佳ML算法。

结论：IPI可作为一种易于获取的标志物，结合ML算法用于CIN预测。对比剂诱导肾病（CIN）是诊断或有创操作中使用对比剂可能引发的严重并发症。中性粒细胞与淋巴细胞比值（neutrophil-to-lymphocyte ratio，NLR）及C反应蛋白与白蛋白比值（C-reactive protein-to-albumin ratio，CAR）等炎症标志物此前已被证实是CIN的良好预测因子，由此推测这些标志物的联合应用可能比单一参数更有助于预测CIN的发生。炎症预后指数（IPI）作为一种新型炎症标志物，由NLR与CAR的乘积计算得出，是非ST段抬高型心肌梗死（NSTEMI）患者发生CIN的独立预测因子。包含IPI的列线图在评估CIN发生风险方面具有较高的预测效能及校准度。在区分CIN阳性与阴性患者时，IPI的鉴别效能优于NLR及CAR。朴素贝叶斯和k近邻算法是预测NSTEMI患者发生CIN的最佳算法。作为预测工具，包含IPI的ML算法（尤其是k近邻算法）在检测CIN发生风险方面可能具有净临床获益。近年来，研究表明ML算法可通过分析现有医疗数据更好地预测不良心血管预后。本研究显示，ML算法结合已知炎症标志物可准确预测NSTEMI患者的CIN发生情况。包含IPI的列线图或可作为NSTEMI患者冠状动脉介入治疗后CIN发生风险的有效预测工具。