老年髋部骨折手术延迟的影响因素分析及风险预测模型构建

Influencing factors and a risk forecasting model of delayed operation for hip fractures in the aged

目的:探讨老年髋部骨折手术延迟的影响因素,构建老年髋部骨折手术延迟风险预测模型。方法:选取2019年11月至2022年11月采用手术治疗的老年髋部骨折患者的病例资料进行研究,将纳入研究的患者按照2∶1的比例随机分为训练集(用于模型构建)和验证集(用于模型验证)。从病历系统中提取纳入患者的信息,包括年龄、性别、体质量指数、骨折类型、美国麻醉医师协会(American Society of Anesthesiologists, ASA)分级、伤前日常活动能力(activities of daily living, ADL)、是否服用影响凝血功能的药物、入院至手术时间、手术方式,是否合并精神障碍、高血压、糖尿病、呼吸系统疾病、心功能不全、肝功能不全、肾功能不全、电解质紊乱、尿酮体异常、下肢静脉血栓、凝血功能异常,以及入院后血清肿瘤坏死因子-α、C反应蛋白水平等。将训练集中的患者根据入院至手术时间分为早期手术组(入院至手术时间<48 h)和延迟手术组(入院至手术时间≥48 h)。先对2组患者的相关信息进行单因素对比分析,再对单因素分析中组间差异有统计学意义的因素进行多因素Logistic回归分析及多重共线性诊断;采用R软件基于贝叶斯网络模型构建老年髋部骨折手术延迟风险预测模型,并采用Netica软件进行贝叶斯网络模型推理。采用受试者操作特征(receiver operating characteristic, ROC)曲线评价老年髋部骨折手术延迟风险预测模型的区分度,采用校准曲线评价老年髋部骨折手术延迟风险预测模型的校准度。结果:(1)分组结果。共纳入老年髋部骨折患者318例,训练集212例、验证集106例。根据入院至手术时间,训练集中早期手术组78例、延迟手术组134例。(2)老年髋部骨折手术延迟影响因素的单因素分析结果。2组患者ASA分级、是否服用影响凝血功能的药物及是否合并精神障碍、高血压、糖尿病、呼吸系统疾病、心功能不全、电解质紊乱、凝血功能异常的比较,组间差异均有统计学意义(χ~2=3.862,P=0.049;χ~2=26.806,P=0.000;χ~2=29.852,P=0.000;χ~2=21.743,P=0.000;χ~2=25.226,P=0.000;χ~2=5.415,P=0.020;χ~2=11.683,P=0.001;χ~2=14.686,P=0.000;χ~2=6.057,P=0.014)。(3)老年髋部骨折手术延迟影响因素的多因素分析及多重共线性诊断结果。多因素Logistic回归分析结果显示,服用影响凝血功能的药物及合并精神障碍、高血压、糖尿病、呼吸系统疾病、心功能不全、电解质紊乱、凝血功能异常均是老年髋部骨折手术延迟的影响因素[β=0.328,P=0.000,OR=5.112,95%CI(2.686,9.728);β=0.322,P=0.000,OR=5.425,95%CI(2.884,10.203);β=0.302,P=0.000,OR=3.956,95%CI(2.189,7.148);β=0.312,P=0.000,OR=4.560,95%CI(2.476,8.398);β=0.291,P=0.021,OR=1.962,95%CI(1.108,3.474);β=0.296,P=0.001,OR=2.713,95%CI(1.520,4.844);β=0.303,P=0.000,OR=3.133,95%CI(1.729,5.679);β=0.296,P=0.015,OR=2.061,95%CI(1.154,3.680)];多重共线性诊断结果显示,上述影响因素均不存在共线性(VIF=1.134,VIF=1.266,VIF=1.465,VIF=1.389,VIF=1.342,VIF=1.183,VIF=1.346,VIF=1.259)。(4)基于贝叶斯网络模型的老年髋部骨折手术延迟风险预测模型的构建与推理结果。基于贝叶斯网络模型构建的老年髋部骨折手术延迟风险预测模型包括8个节点、8条有向边。模型显示,服用影响凝血功能的药物及合并精神障碍、呼吸系统疾病、电解质紊乱、凝血功能异常直接影响手术延迟的发生,合并心功能不全、高血压、糖尿病间接影响手术延迟的发生;推理结果显示,患者合并心功能不全、凝血功能异常及精神障碍时,手术延迟发生率为64.1%。(5)老年髋部骨折手术延迟风险预测模型的评价结果。采用训练集数据进行老年髋部骨折手术延迟风险预测模型评价,ROC曲线下面积为0.861[P=0.000,95%CI(0.810,0.912)],灵敏度为91.29%,特异度为93.35%;校准曲线显示其一致性指数为0.866[P=0.000,95%CI(0.702,0.943)];采用验证集数据进行老年髋部骨折手术延迟风险预测模型评价,ROC曲线下面积为0.848[P=0.000,95%CI(0.795,0.901)],灵敏度为91.62%,特异度为92.46%;校准曲线显示其一致性指数为0.879[P=0.000,95%CI(0.723,0.981)]。结论:服用影响凝血功能的药物以及合并精神障碍、高血压、糖尿病、呼吸系统疾病、心功能不全、电解质紊乱、凝血功能异常均为老年髋部骨折手术延迟的影响因素,基于上述因素构建的老年髋部骨折手术延迟风险预测模型具有较高的应用价值。

Objective:To explore the factors influencing delay of operation for hip fractures in the aged,and to build a risk forecasting model.Methods:The medical records of aged patients who underwent operation for hip fractures from November 2019 to November 2022 were selected.The information of the included patients,including age,gender,body mass index,fracture type,American Society of Anesthesiologists(ASA)classification,activities of daily living(ADL)before injury,whether taking medications affecting blood coagulation function,time from hospital admission to operation,operation method,whether combined with mental disorder,hypertension,diabetes mellitus,respiratory system diseases,cardiac insufficiency,hepatic insufficiency,renal insufficiency,electrolyte disorders,urinary ketone body abnormalities,lower-extremity venous thrombosis,and coagulation abnormalities,as well as the serum levels of tumor necrosis factor-α(TNF-α)and C-reaction protein(CRP)after hospital admission,was extracted from the electronic medical record system(EMRS).The included patients were randomized into a training set(for model building)and a validation set(for model validation)in a ratio of 2∶1.According to the time from hospital admission to operation,the patients in the training set were further assigned into early operation group(time from hospital admission to operation of<48 hours)and delayed operation group(time from hospital admission to operation of≥48 hours).The single-factor comparative analysis was performed on the relevant information of patients in the 2 groups,followed by multi-factor logistic regression analysis and multicollinearity diagnosis on the factors with statistically significant differences between the 2 groups.A Bayesian network-based model for predicting the risk of delayed operation in aged patients with hip fractures was constructed by using the R software,and the model was inferred by using the Netica software;meanwhile,the discrimination and calibration performance of the risk forecasting model were analyzed and evaluated by using receiver operating characteristic(ROC)curve and calibration curve,respectively.Results:①Three hundred and eighteen aged patients with hip fractures were enrolled in the study,212 ones in training set,and 106 ones in validation set;furthrtmore,according to the time from hospital admission to operation,78 ones of training set were assigned into early operation group,and 134 ones of training set into delayed operation group.②The difference was statistically significant in ASA classification,whether taking medications affecting blood coagulation function,whether combined with mental disorder,hypertension,diabetes mellitus,respiratory system diseases,cardiac insufficiency,electrolyte disorders,and coagulation abnormalities between the 2 groups(χ2=3.862,P=0.049;χ2=26.806,P=0.000;χ2=29.852,P=0.000;χ2=21.743,P=0.000;χ2=25.226,P=0.000;χ2=5.415,P=0.020;χ2=11.683,P=0.001;χ2=14.686,P=0.000;χ2=6.057,P=0.014).③The results of multi-factor logistic regression analysis revealed that taking medications affecting blood coagulation function,combined with mental disorder,hypertension,diabetes mellitus,respiratory system diseases,cardiac insufficiency,electrolyte disorders,and coagulation abnormalities were the factors influencing delay of operation for hip fractures in the aged(β=0.328,P=0.000,OR=5.112,95%CI(2.686,9.728);β=0.322,P=0.000,OR=5.425,95%CI(2.884,10.203);β=0.302,P=0.000,OR=3.956,95%CI(2.189,7.148);β=0.312,P=0.000,OR=4.560,95%CI(2.476,8.398);β=0.291,P=0.021,OR=1.962,95%CI(1.108,3.474);β=0.296,P=0.001,OR=2.713,95%CI(1.520,4.844);β=0.303,P=0.000,OR=3.133,95%CI(1.729,5.679);β=0.296,P=0.015,OR=2.061,95%CI(1.154,3.680)).The results of multicollinearity diagnosis showed that there was no collinearity between the above-mentioned influencing factors(VIF=1.134,VIF=1.266,VIF=1.465,VIF=1.389,VIF=1.342,VIF=1.183,VIF=1.346,VIF=1.259).④The Bayesian network-based risk forecasting model consisted of 8 nodes and 8 directed edges.The model showcased that taking medications affecting blood coagulation function,combined with mental disorder,respiratory system diseases,electrolyte disorders,and coagulation abnormalities directly affected the occurrence of operation delay;while,combined with cardiac insufficiency,hypertension,and diabetes mellitus indirectly affected the occurrence of operation delay.The results of inference showed that the incidence rate of delayed operation was 64.1%in aged patients with cardiac insufficiency,coagulation abnormalities,and mental disorders.⑤The evaluation on risk forecasting model with the training set data showed that the area under the ROC curve,sensitivity,specificity,and consistency index was 0.861(P=0.000,95%CI(0.810,0.912)),91.29%,93.35%,and 0.866(P=0.000,95%CI(0.702,0.943)),respectively.The evaluation on risk forecasting model with the validation set data showed that the area under the ROC curve,sensitivity,specificity,and consistency index was 0.848(P=0.000,95%CI(0.795,0.901)),91.62%,92.46%,and 0.879(P=0.000,95%CI(0.723,0.981)),respectively.Conclusion:Taking medications affecting blood coagulation function,combined with mental disorder,hypertension,diabetes mellitus,respiratory system diseases,cardiac insufficiency,electrolyte disorders,and coagulation abnormalities are the factors influencing the delay of operation for hip fractures in the aged.The risk forecasting model built based on the above factors demonstrates a high application value.