摘要
目的 建立老年髋部骨折患者术后3年全因死亡风险预测模型并验证其效能.方法 回顾性分析2020年1月至2021年12月中国人民解放军总医院第四医学中心收治的325例老年髋部骨折患者的资料,男90例,女235例;年龄(80.9±8.9)岁;股骨转子间骨折191例,股骨颈骨折134例.将2020年1月至2021年6月收治的髋部骨折患者分配至建模组(244例),将2021年7-12月收治的髋部骨折患者分配至验证组(81例).记录两组患者年龄、性别、骨折类型、伤前疾病、血液学指标、受伤至手术时间、既往抗凝药物使用情况、骨密度、美国麻醉医师协会(ASA)分级、体重指数、输血量.利用建模组数据,采用logistic回归分析及受试者工作特征曲线(ROC)预测、筛选出老年髋部骨折术后3年死亡的独立危险因素,并构建预测模型;利用验证组数据,使用ROC曲线分析其预测效能,应用Hosmer-Lemeshow拟合优度检验评价其拟合程度.结果 建模组与验证组术后3年死亡患者数分别有75、29例.两组患者各基线资料比较差异均无统计学意义(P＞0.05),具有可比性.logistic单因素分析显示:性别、年龄、伤前冠状动脉粥样性心脏病、伤前脑卒中、伤前肺部感染、血红蛋白、白蛋白、高密度脂蛋白、体重指数、受伤至手术时间和ASA分级与患者术后3年死亡有关(P＜0.05).ROC曲线显示有预测能力的变量包括:年龄、血红蛋白、白蛋白、体重指数、受伤至手术时间.logistic多因素分析显示:男性、年龄≥83岁、伤前脑卒中、受伤至手术时间≥4.5d、ASA Ⅲ级、ASA Ⅳ级与老年髋部骨折患者术后3年死亡显著相关(P＜0.05).基于以上指标构建预测模型,ROC曲线下面积为0.792(95％CI:0.730-0.855,P＜0.001),截断值为0.459,灵敏度和特异度分别为56％和90.5％,且该模型的拟合程度良好(x2=5.818,P=0.668).结论 男性、年龄≥ 83岁、伤前脑卒中、受伤至手术时间≥4.5d、ASA Ⅲ级、Ⅳ级是老年髋部骨折患者术后3年死亡的危险因素.基于以上指标构建的老年髋部骨折术后死亡风险预测模型具有良好的预测效能.
Abstract
Objective To develop and validate a risk model of postoperative 3-year all-cause mortal-ity for the elderly patients with hip fracture.Methods A retrospective study was conducted to analyze the clinical data of 325 elderly patients with hip fracture who had been treated at The Fourth Medical Center,The General Hospital of PLA from January 2020 to December 2021.There were 90 men and 235 women with an age of(80.9±8.9)years,and 191 intertrochanteric fractures and 134 femoral neck fractures.The patients with hip fractures admitted from January 2020 to June 2021 were assigned to a modeling group(244 cases),and those admitted from July to December 2021 to a validation group(81 cases).Recorded were age,gender,fracture type,pre-injury disease,hematological indicators,time from injury to surgery,previous use of antico-agulant drugs,bone density,American Society of Anesthesiologists(ASA)grading,body mass index,and blood transfusion volume in the 2 groups of patients.The data in the modeling group were used to screen the independent risk factors for the postoperative 3-year all-cause mortality in the elderly patients with hip fracture through logistic regression analysis and receiver operating characteristic(ROC)curve analysis and consequently to construct a prediction model.The data in the validation group were used to analyze the predictive perfor-mance of the model through the ROC curve,and the fitting degree of the model was evaluated using the Hosmer-Lemeshow goodness-of-fit test.Results Respectively,75 cases in the modeling group and 29 cases in the validation group died 3 years after surgery.There was no statistically significant difference in the baseline data between the 2 groups,indicating comparability(P＞0.05).Logistic univariate analysis showed that gender,age,coronary atherosclerotic heart disease before injury,stroke before injury,pulmonary infection before in-jury,hemoglobin,albumin,high-density lipoprotein,body mass index,time from injury to surgery and ASA grade were related to the mortality of patients 3 years after surgery(P＜0.05).The ROC curve analysis showed that the variables with predictive ability included age,hemoglobin,albumin,body mass index,and time from injury to surgery.Logistic multivariate analysis showed male,age≥83 years,pre-injury stroke,time from injury to surgery ≥ 4.5 days,ASA grade Ⅲ,and ASA grade Ⅳ were significantly related to the postoperative 3-year all-cause mortality in the elderly patients with hip fracture(P＜0.05).In the prediction model based on the above indicators,the area under the ROC curve was 0.792(95％CI:0.730 to 0.855,P＜0.001),the cutoff value 0.459,sensitivity 56％,and specificity 90.5％,showing a good fit of the model(x2=5.818,P=0.668).Conclusions Male,age≥83 years,pre-injury stroke,time from injury to surgery ≥ 4.5 days,and ASA grades Ⅲ and Ⅳ are risk factors for postoperative 3-year all-cause mortality in the elderly patients with hip fracture.The risk prediction model based on these indicators demonstrates good predictive efficacy.
NETL
NSTL
万方数据