放射性颌骨坏死术后感染预警模型的建立与验证

目的: 构建预测放射性颌骨坏死（osteoradionecrosis of the jaw,ORNJ）术后出现手术部位感染（surgical site infection,SSI）风险的列线图模型,并对该模型进行外部验证,评估其预测能力。方法: 纳入1993年3月—2014年10月于中山大学附属口腔医院手术治疗的507例ORNJ患者作为建模组,另纳入2000年1月—2021年11月于南方医科大学口腔医院手术治疗的112例ORNJ病例作为验证组,收集临床资料。采用SPSS 22.0软件包的单因素及多因素Logistic回归模型,分析ORNJ患者并发SSI的影响因素;应用R软件建立预测ORNJ患者并发SSI风险的列线图模型,并进行内部和外部验证。结果: Logistic回归分析显示,性别、放疗间隔时间≤2年、开口度≤1 cm、术前感染程度、手术方式是ORNJ患者并发SSI的独立影响因素。对列线图模型进行验证,其在建模组与验证组的Harrell一致性指数（C-index）分别为0.745和0.751;校准曲线显示该模型具有良好的区分度与精准度;ROC曲线显示该模型在建模组与验证组预测ORNJ患者并发SSI风险的曲线下面积分别为0.7446（95%CI：0.6934~0.7959）和0.7514（95%CI：0.6556~0.8473）。结论: 结合性别、放疗间隔时间、开口度、术前感染程度、手术方式等因素所构建的列线图可以较准确地预测ORNJ患者术后围术期发生SSI的风险,具有较高的临床应用价值。

Construction and validation of early warning model for postoperative infection of radionecrosis of the jaw

PURPOSE: The purpose of this study was to construct a rosette model to predict the risk of surgical site infection (SSI) after surgical treatment of osteoradio-necrosis of the jaw (ORNJ), and to conduct external validation of the model to evaluate its predictive power. METHODS: Five hundred and seven ORNJ surgical cases from March 1993 to October 2014 in Hospital of Stomatology, Sun Yat-sen University were included as model group. In addition, 112 ORNJ surgical cases from January 2000 to November 2021 in Hospital of Stomatology, Southern Medical University were included as validation group. The patients' clinical data were collected. Univariate and multivariate Logistic regression model in SPSS 22.0 software package was used to analyze the contributing factors of patients of ORNJ complicated with SSI. Then a nomogram was established using R programming language to predict the risk of ORNJ being complicated with SSI, internal and external validations were finally carried out. RESULTS: Logistic regression analysis showed that gender, radiation interval (≤2 years), mouth opening(≤1 cm), preoperative infection and procedures were the independent contributing factors of ORNJ complicated with SSI. Harrell concordance index was 0.745 and 0.751 in model group and in validation group, respectively. Calibration curve showed that this model had good discrimination and accuracy. The area under the curve (AUC) of ROC was 0.7446(95%CI: 0.6934-0.7959) and 0.7514(95%CI: 0.6556-0.8473) in model group and in validation group, respectively for predicting the risk of ORNJ complicated with SSI. CONCLUSIONS: Nomogram combining factors like gender, radiation interval, mouth opening, preoperative infection and operation methods can provide us with a comparatively precise way to predict the risk of surgical site infection in patients with ORNJ during perioperative period, which has distinct clinical application value.