摘要
目的探讨退变性腰椎滑脱患者术后腰背功能康复效果的影响因素,基于多维度指标构建预警模型,并评价其预测效能。方法前瞻性选取2022年1月至2023年12月河南省洛阳正骨医院(河南省骨科医院)收治的253例退变性腰椎滑脱患者作为研究对象,所有患者均采取显微内镜辅助下微创经椎间孔腰椎椎体间融合术治疗。术后随访12个月(10例失访),根据日本骨科协会(JOA)评分评价腰背功能康复效果,将患者分为康复良好组(n=202)和不良组(n=41)。比较两组患者的人口学特征、围术期指标和炎症因子等,并利用逻辑回归(LR)与极端梯度提升(XGBoost)算法筛选退变性腰椎滑脱患者术后腰背功能康复效果的影响因素,基于此应用R软件分别构建LR模型和XGBoost模型,绘制受试者工作(ROC)曲线,并计算曲线下面积(AUC)、敏感度、特异度、准确性以评估两个预测模型的综合预测效能。另选取2024年1月至2024年3月60例退变性腰椎滑脱患者作为验证集进行外部验证。结果不良组患者的体质量指数(BMI)、糖尿病占比、恐动症占比、多裂肌脂肪浸润Goutallier分级4级占比分别为(22.88±1.13)kg/m^(2)、53.66%、63.41%、51.22%,明显高于良好组的(21.87±1.46)kg/m^(2)、31.19%、42.08%、15.35%,而术后1周内康复锻炼占比为58.54%,明显低于良好组的75.74%,差异均有统计学意义(P<0.05);LR回归方程分析结果显示,BMI、恐动症、多裂肌脂肪浸润Goutallier分级、术后1周内康复锻炼是术后腰背功能康复不良的影响因素(P<0.05);XGBoost算法筛选结果显示,术后腰背功能康复不良的前4个重要变量依次为恐动症、术后1周内康复锻炼、多裂肌脂肪浸润Goutallier分级、糖尿病;验证结果显示,XGBoost算法模型在内部、外部验证中的曲线下面积(AUC)分别为0.925(0.884~0.955)、0.942(0.850~0.986),明显高于LR模型的0.849(0.798~0.892)、0.785(0.660~0.881)(P<0.05),且综合预测性能最佳。结论XGBoost算法模型在退变性腰椎滑脱患者术后腰背功能康复效果预测中的预测性能最佳,其影响因素为恐动症、术后1周内康复锻炼、多裂肌脂肪浸润Goutallier分级、糖尿病,据此能及早识别该类患者术后腰背功能康复不良的潜在风险人群,可为临床预防措施的制定提供参考依据。
Objective To explore the influencing factors of postoperative low back function rehabilitation in patients with degenerative lumbar spondylolisthesis(DLS),construct an early warning model based on multidimensional indicators,and evaluate its predictive efficacy.Methods A prospective study was conducted on 253 DLS patients admitted to Luoyang Orthopedic-Traumatological Hospital of Henan Province from January 2022 to December 2023.All patients underwent microendoscopy-assisted minimally invasive transforaminal lumbar interbody fusion(MI-TLIF).Patients were followed up for 12 months postoperatively(10 cases lost to follow-up),and their rehabilitation outcome was assessed using the Japanese Orthopaedic Association(JOA)score.Patients were divided into a good recovery group(n=202)and a poor recovery group(n=41).Demographic characteristics,perioperative indicators,and inflammatory factors were compared between the two groups.Logistic Regression(LR)and eXtreme Gradient Boosting(XGBoost)algorithms were used to identify influencing factors for poor rehabilitation outcome.Based on these factors,LR and XGBoost models were constructed using R software.Receiver operating characteristic(ROC)curves were plotted,and the area under the curve(AUC),sensitivity,specificity,and accuracy were calculated to evaluate the comprehensive predictive dation set for external validation.Results The poor recovery group had significantly higher BMI,proportions of diabetes,kinesiophobia,and multifidus fat infiltration Goutallier Grade 4,and a significantly lower proportion of patients initiating rehabilitation exercise within 1 week postoperatively compared to the good recovery group(all P<0.05):(22.88±1.13)kg/m^(2) vs(21.87±1.46)kg/m^(2),53.66%vs 31.19%,63.41%vs 42.08%,51.22%vs 15.35%,and 58.54%vs 75.74%,respectively.LR analysis showed that BMI,kinesiophobia,multifidus fat infiltration Goutallier grade,and rehabilitation exercise within 1 week postoperatively were influencing factors for poor rehabilitation outcome(P<0.05).The XGBoost algorithm identified the top four important variables as follows:kinesiophobia,rehabilitation exercise within 1 week postoperatively,multifidus fat infiltration Goutallier grade,and diabetes.Validation results showed that the XGBoost model had AUCs of 0.925(0.884-0.955)and 0.942(0.850-0.986)in internal and external validation,respectively,significantly higher than the LR model's AUCs of 0.849(0.798-0.892)and 0.785(0.660-0.881),P<0.05,and demonstrated the best overall predictive performance.Conclusion The XGBoost model demonstrates superior predictive performance for forecasting postoperative low back function rehabilitation outcomes in DLS patients.The key influencing factors are kinesiophobia,rehabilitation exercise initiation within 1 week postoperatively,multifidus fat infiltration Goutallier grade,and diabetes.This model can facilitate early identification of patients at potential risk for poor rehabilitation outcomes,providing a reference for developing clinical preventive measures.
作者
李磊
张志乾
陈钱
LI Lei;ZHANG Zhi-qian;CHEN Qian(Department of Spinal SurgeryⅡ,Zhengzhou Branch,Luoyang Orthopedic-Traumatological Hospital of Henan Province(Henan Provincial Orthopedic Hospital),Zhengzhou 450000,Henan,CHINA)
出处
《海南医学》
2025年第24期3546-3552,共7页
Hainan Medical Journal
基金
河南省中医药拔尖人才培养项目(编号:豫卫中医函2021-15号)。
关键词
多维度指标
退变性腰椎滑脱
腰背功能
康复效果
风险预警模型
Multidimensional indicators
Degenerative lumbar spondylolisthesis
Low back function
Rehabilitation effects
Risk warning model
