脊柱结核病灶清除植骨融合内固定术后住院时间延长的危险因素及预测模型

Risk factors and predictive model for prolonged length of stay after spinal tuberculosis lesion removal and bone graft fusion with internal fixation

中文关键词:  脊柱结核  脊柱融合术  住院时间  危险因素  列线图

英文关键词:Spinal tuberculosis  Spinal fusion  Length of stay  Risk factors  Nomogram

中文摘要:

  【摘要】 目的：分析脊柱结核患者行病灶清除植骨融合内固定术后住院时间（length of stay，LOS）延长的危险因素，建立预测模型并进行验证。方法：回顾性分析2016年2月~2020年12月在西安交通大学附属红会医院行病灶清除植骨融合内固定术的152例脊柱结核患者的临床资料，根据患者术后LOS是否超过整体研究队列第75%分位的术后LOS分为LOS延长组（PLOS组）和LOS正常组（NLOS组）。对两组患者的性别、年龄、高血压、糖尿病、截瘫、抗凝史、结核耐药、术前抗结核时间、输血、手术部位、手术入路、融合椎体数目、手术时间、术中出血量（intraoperative blood loss，IBL）、术后并发症、输血费用、住院费用、C反应蛋白（C-reactive protein，CRP）、血沉（erythrocyte sedimentation rate，ESR）、白蛋白（albumin，ALB）、血常规、凝血功能等进行单因素分析。根据套索（Lasso）回归，选择与脊柱结核术后LOS延长显著相关的危险因素；随后将筛选出来的危险因素纳入多因素Logistic回归分析，最终依据多因素Logistic回归分析结果建立预测模型，并通过绘制列线图对模型进行可视化，以此来预测脊柱结核术后LOS延长的风险概率。使用自举法（Bootstrap）进行模型内部验证，绘制受试者工作特征（receiver operating characteristic，ROC）曲线、校准曲线和决策曲线分析（decision curve analysis，DCA）验证该模型的区分度、准确度以及临床适用性。结果：纳入研究的152例患者中位LOS为10d，75%LOS为14d，PLOS组96例，NLOS组56例。单因素分析显示，两组患者的年龄、高血压、糖尿病、抗凝史、结核耐药、术前抗结核时间、手术部位、手术入路、手术时间、IBL、术后并发症、CRP、ESR、术前ALB、血常规、凝血功能等均无统计学差异（P>0.05），两组患者的性别、截瘫、输血、融合椎体数目、输血费用、住院费用差异有统计学差异（P<0.05）。将患者手术时间、IBL、术前Hb、术前ALB，按ROC的约登指数为分割点，手术时间临界值为198（min）、IBL临界值为1000（mL）、术前Hb临界值为118（g/L）、术前ALB 临界值为38.8（g/L）。筛选出与脊柱结核术后LOS延长密切相关的危险因素为女性、输血、融合椎体数目≥3、手术时间≥198min和IBL≥1000mL、术前Hb＜118g/L和术前ALB＜38.8g/L。多因素Logistic回归分析显示，女性、融合椎体数目≥3、手术时间≥198min和IBL≥1000ml是脊柱结核患者术后LOS延长的危险因素（P<0.05）。构建Logistic回归的可视化列线图模型，列线图中的预测因子包括女性、融合椎体数目、手术时间和IBL。进行1000次Bootstrap自助抽样以完成模型内部验证，C指数值为0.882，ROC曲线下面积（area under curve，AUC）为0.884（95%CI：0.782～0.985）。校准曲线显示模型的表观曲线与偏差校正后的曲线拟合良好。DCA曲线显示在0.2～0.9的阈值区间具有最大临床效益。结论：女性、融合椎体数目≥3、手术时间≥198min和IBL≥1000ml是脊柱结核患者行后路病灶清除植骨融合内固定术后LOS延长的主要危险因素，基于以上危险因素所绘制的连线图可以帮助医生做出临床决策并优化围术期管理。

英文摘要:

  【Abstract】 Objectives: To analyze the risk factors for prolonged length of stay(LOS) after lesion removal and bone graft fusion internal fixation in patients with spinal tuberculosis, and to develop and validate a predictive model. Methods: The clinical data of 152 patients with spinal tuberculosis who underwent lesion removal and bone grafting and fusion internal fixation at Honghui Hospital affiliated to Xi′an Jiaotong University from February 2016 to December 2020 were retrospectively analyzed. The patients were divided into the prolonged LOS(PLOS) group and normal LOS(NLOS) group according to whether their postoperative LOS exceeded the postoperative LOS of the 75th percentile of the overall study cohort. Univariate analyses were performed for gender, age, hypertension, diabetes, paraplegia, anticoagulation history, tuberculosis resistance, preoperative anti-tuberculosis time, blood transfusion, surgical site, surgical approach, number of fused vertebrae, operative time, intraoperative blood loss(IBL), postoperative complications, blood transfusion cost, hospital cost, C-reactive protein(CRP), erythrocyte sedimentation rate(ESR), albumin(ALB), blood routine, and coagulation function in both groups. Based on Lasso regression, the risk factors significantly associated with postoperative LOS prolongation in spinal tuberculosis were selected and incorporated into a multivariate logistic regression analysis, and thereby a prediction model was established based on the results of multivariate logistic regression analysis. The model was visualized by plotting a nomogram as a means of predicting the probability of risk for prolongation of LOS after spinal tuberculosis surgery. Internal validation of the model was performed using extended Bootstrap, where receiver operating characteristic(ROC) curves, calibration curves and decision curve analysis(DCA) were plotted to verify the discrimination, accuracy and clinical applicability. Results: The 152 patients enrolled in the study had a median LOS of 10d, and the 75% LOS was 14d. There were 96 patients in the PLOS group and 56 in the NLOS group. Univariate analysis showed that the differences in age, hypertension, diabetes, anticoagulation history, tuberculosis resistance, preoperative anti-tuberculosis time, surgical site, surgical approach, surgical time, IBL, postoperative complications, CRP, ESR, preoperative ALB, blood routine, and coagulation function were not statistically significant between the two groups(P>0.05), while the differences in gender, paraplegia, blood transfusion, number of fused vertebrae, blood transfusion cost, and hospitalization cost were statistically significant(P<0.05). The operative time, IBL, preoperative Hb, and preoperative ALB were divided according to the Yoden index of ROC as the dividing point, and the cut-off value of operative time was 198(min), the cut-off value of IBL was 1000(mL), the cut-off value of preoperative Hb was 118(g/L), and the cut-off value of preoperative ALB was 38.8(g/L). Through Lasso regression model, the risk factors closely related to the prolongation of LOS after spinal tuberculosis were female, blood transfusion, the number of fused vertebrae≥3, operative time≥198min and IBL≥1000mL, preoperative Hb <118g/L and preoperative ALB <38.8g/L. Multivariate logistic regression analysis showed that female, number of fused vertebrae≥3, operative time≥198min and IBL≥1000mL were the risk factors for postoperative LOS prolongation in patients with spinal tuberculosis(P<0.05). A visual nomogram model for logistic regression was constructed, and the predictors included female, number of fused vertebrae, operative time, and IBL. A Bootstrap self-sampling of 1,000 times was performed to complete the internal validation of the model, with a C-index value of 0.882 and an area under the curve(AUC) of ROC of 0.884(95% CI: 0.782 to 0.985). The calibration curve showed that the apparent curve of the model fitted well with the curve after deviation correction. The DCA curve showed that the threshold range of 0.2 to 0.9 had the greatest clinical benefit. Conclusions: Female, number of fused vertebrae≥3, operative time≥198min and IBL≥1000mL are the main risk factors for prolonged LOS after lesion removal and bone graft fusion with internal fixation in patients with spinal tuberculosis, and the predictive model based on the above risk factors can help physicians to make clinical decisions and optimize the perioperative management.