Construction and Validation of A Prognosis Prediction Model for Sudden Deafness
翟晓敏 1许雯雯 2张红蕾 3袁军 3郭睿3
扫码查看
点击上方二维码区域，可以放大扫码查看
作者信息
1. 河北北方学院研究生院(张家口 075000);中国人民解放军空军特色医学中心耳鼻咽喉头颈外科
2. 宜昌市第二人民医院耳鼻咽喉头颈外科
3. 中国人民解放军空军特色医学中心耳鼻咽喉头颈外科
折叠
摘要
目的 研究突发性聋患者的列线图,确定其预后因素.方法 回顾性选取2018年1月-2020年3月中国人民解放军空军特色医学中心耳鼻咽喉头颈外科收治的129例突发性聋患者,按照7:3随机分为训练组93例,验证组 36例.通过套索算法(the least absolute shrinkage and selection operator,LASSO)筛选变量,二元 logistic 回归分析年龄、性别、患侧、发病时间、耳鸣、眩晕和听力图分型对预后的影响,结合临床经验和二元logistic回归分析中有统计学意义的变量用于构建突发性聋预后模型.计算一致性指数(concordance index,C-index),并绘制受试者工作特征(receiver operating characteristic,ROC)曲线下面积、校准度和决策度来评估和验证模型的准确性.结果 眩晕和听力图分型是突发性聋的独立预后因素.训练组C-index为0.863(95％CI:0.760～0.931),验证组C-index为0.786(95％CI:0.627-0.944).训练组ROC曲线下面积为0.829,验证组ROC曲线下面积为0.786.预测模型观察值和预测值具有较高的一致性,良好的临床实用性.结论 本研究构建的模型可预测突发性聋患者的预后,为临床工作人员决策提供参考.
Abstract
Objective To report a nomogram model of prognostic factors in sudden deafness.Methods Data from patients with sudden deafness treated at the Air Force Medical Center from January 2018 to March 2020(n=129),randomly divided into a training set(n=93)and a validation set(n=36),were retrospectively analyzed.The least abso-lute shrinkage and selection operator(LASSO)regression was used to screen variables,and binary logistic regression was used to analyze effects of age,gender,disease side,time of onset,tinnitus,vertigo and audiogram pattern on prog-nosis.Combined with clinical experiences,variables identified via the binary logistic regression analysis were used to construct a prognosis prediction model.The accuracy of the model was assessed and validated by calculating the consis-tency index(C-index)and area under the receiver operating characteristic curve(ROC),as well as calibration and deci-sion degrees.Results Vertigo and audiogram pattern were independent prognostic factors in sudden deafness.The con-cordance indices(C-indices)of nomogram was 0.863(95％CI:0.760-0.931)in the training cohort and 0.786(95％CI:0.627-0.944)in the validation cohort,respectively.The(ROC)was 0.829 for the training cohort and 0.786 for the vali-dation cohort,respectively.Calibration curves suggested high consistency between observations and predictions.Deci-sion curve analysis(DCA)indicated good clinical practicability.Conclusion The model constructed in this study can predict the prognosis in sudden deafness,and can facilitate decision-making for clinical staff.
维普
万方数据