肘关节骨折术后关节功能恢复不良Nomogram预测模型的建立与验证

Establishment and verification of a Nomogram prediction model for poor postoperative joint functional recovery after elbow fractures

目的探究肘关节骨折术后关节功能恢复不良Nomogram预测模型的建立与验证。方法回顾性分析2019年6月—2022年9月唐山市第二医院骨病科收治的210例肘关节骨折患者,收集患者的一般资料,包括:性别、年龄、损伤侧别、骨折类型、骨折AO分型、合并血管损伤、合并神经损伤、合并糖尿病、骨折至手术时间、手术入路方式、术后并发症、术后是否进行早期康复锻炼。依据肘关节恢复情况分为恢复良好组和恢复不良组,通过Logistic回归分析影响术后肘关节功能恢复不良的独立影响因素,据此独立影响因素构建Nomogram预测模型,采用R软件中C指数、受试者工作特征(ROC)曲线及校准曲线验证术后关节功能恢复不良风险的Nomogram模型效能。结果210例患者中,关节功能恢复不良48例,占比22.86%,余162例均恢复良好,纳入恢复良好组。恢复不良组年龄(≥60岁)、开放性骨折、C型骨折、合并血管损伤、合并神经损伤、合并糖尿病、术后出现并发症人数占比均多于恢复良好组,进行早期康复锻炼人数占比少于恢复良好组(P<0.05)。经Logistic回归分析显示,年龄≥60岁、开放性骨折、C型骨折、合并血管损伤、合并神经损伤、合并糖尿病、术后出现并发症、未进行早期康复训练是影响术后肘关节功能恢复的独立危险因素(P<0.05);绘制ROC曲线结果显示术后肘关节功能恢复不良的AUC值均>0.700,OR>1,说明上述指标对于术后肘关节功能恢复不良具有较好的预测价值;基于以上8个独立危险因素建立Lasso-Nomogram预测模型,校准曲线C-index值为0.820,ROC曲线训练组和测试组的AUC值为0.822和0.701,说明该Nomogram模型具有良好的区分度及预测能效。结论基于术后肘关节功能恢复的独立影响因素构建的Nomogram预测模型,能直观地预测术后肘关节功能恢复不良发生的概率。

Objective To explore the influencing factors of poor joint functional recovery of elbow joint fractures after surgery and to establish and verify the Nomogram prediction model.From Jun.2019 to Sep.2022,210 patients with elbow fractures in our hospital were retrospectively analyzed,in terms of gender,age,injury side,fracture type,AO classification,combination of vascular or nerve injuries,combination of diabetes,fracture to operation time,operation approaches,postoperative complications,and conduction of rehabilitation exercises early after surgery.According to the functional recovery of the elbow joint,patients were divided into good recovery group and poor recovery group.Logistic regression analysis was used to analyze independent influencing factors for poor postoperative elbow joint functional recovery,and based on the results a Nomogram prediction model was constructed,whose efficiency was further verified by the C index,receiver operating characteristic(ROC)curve and calibration curve in R software.Results Among the 210 patients,48 have poor joint functional recovery(poor recovery group),accounting for 22.86%.The other 162 cases were included in the good recovery group.Compared with the good recovery group,the poor recovery group showed much higher proportions of patients with an age≥60 years,open fractures,AO type C fractures,combined vascular injury,combined nerve injury,diabetes history,and postoperative complications(all P<0.05),and less proportion of patients taking early rehabilitation exercises(P<0.05),which were all confirmed as independent risk factors for poor functional recovery of elbow fractures after surgery by Logistic regression analysis(all P<0.05).The ROC curve showed that for each risk factor,the area under the curve was>0.700 and the odds ratio was>1,indicating a good value of all the factors in predicting poor postoperative elbow joint functional recovery.Based on the above 8 independent risk factors,the Lasso-Nomogram prediction model was established.The C-index value of the calibration curve was 0.820,and the AUC of the ROC curve training group and test group reached 0.822 and 0.701 respectively,which indicated that the Nomogram model had good discrimination and prediction efficiency.Conclusion A Nomogram prediction model based on 8 independent influencing factors of elbow joint functional recovery after surgery is effective.