川南地区老年髋部骨折术后常见并发症发生率及死亡率预测模型的初步建立与价值分析

发布时间：2018-04-30 22:13

本文选题：老年髋部骨折 + 肺部感染　；参考：《西南医科大学》2017年硕士论文

【摘要】：目的:建立川南地区老年髋部骨折术后常见并发症发生率和死亡率的预测模型,并检验其预测价值。方法:1.建立4份资料收集表,包括老年髋部骨折术后肺部感染、术后认知功能障碍(POCD)、术后下肢深静脉血栓形成(术后LEDVT)、术后死亡;2.收集2012年1月到2016年10月于西南医科大学附属医院住院手术治疗的此类患者临床数据,填入相对应的资料收集表;3.然后利用Epidata3.1软件建立相应4个数据库并将相对应的资料收集表中的临床数据录入数据库;将数据导入spss19.0软件进行统计分析:计量资料采用t检验、计数资料采用χ2检验进行变量的单因素分析,获得有统计学意义的变量(以?=0.05为检验水准,P值(27)0.05变量有统计学意义);4.在生理学和手术严重度评分系统(physical and operation severity score for the enumeration of mortality and morbidity,POSSUM)的基础上,将这些变量分为生理学指标和手术严重性指标两类,建立起老年髋部骨折术后肺部感染、POCD、术后LEDVT、术后死亡的评分系统。通过Logistic回归分析得出此类患者术后肺部感染、POCD、术后LEDVT、术后死亡发生率预测模型;5.最后用实际值与预测值的比值、ROC曲线、Hosmer-Lemeshow检验来评估其预测价值。结果:1.术后肺部感染组:1)生理学指标中的年龄、白细胞、ASA分级、COPD、心功能分级、合并症数量,以及手术严重性指标中的术前准备时间、手术时间、术中失血量、麻醉方式是术后肺部感染的危险因素。2)术后肺部感染风险评分系统预测模型:Ln[R/(1-R)]=-7.187+0.226×PS+0.161×OS。3)该模型术后肺部感染率的预测值平均8.93%,实际值9.89%,实际值/预测值1.11,两者之间差异无统计学意义(χ2=0.279,P=0.6730.05)。ROC曲线结果显示灵敏度(Se)=82.7%,特异度(Sp)=72.4%,误诊率(?)=27.6%,漏诊率(β)=17.3%,ROC曲线下面积为0.814。对该预测模型进行Hosmer-Lemeshow检验,结果显示,此评分系统预测术后肺部并发症发生率(H2=7.707,df=8,P=0.4630.05)效果良好,数据中的信息被充分提取。2.术后LEDVT组:1)生理学指标中的年龄、FIB、血清甘油三酯、BMI、静脉曲张、高血压、冠心病、糖尿病、脑卒中、感染以及手术严重性指标中的麻醉、术前准备时间、出血量、手术时间是术后LEDVT的危险因素。2)术后LEDVT风险评分系统预测模型:Ln[R/(1-R)]=-11.493+0.347×PS+0.327×OS。3)该模型术后LEDVT发生率预测值为平均12.57%,实际值为13.38%,实际值/预测值为1.06,两者之间差异无统计学意义(χ2=0.144,P=0.7760.05)。ROC曲线结果显示Se=74.20%,Sp=86.20%,?=13.80%,β=25.80%,ROC曲线下面积为0.87。对该预测模型进行Hosmer-Lemeshow检验,结果显示,该评分系统预测术后LEDVT发生率(H2=3.309,df=8,P=0.9140.05)效果良好,数据中的信息被充分提取。3.POCD组:1)生理学指标中的年龄、血压(收缩压)、白蛋白、氧分压、合并症数量,COPD、脑卒中以及手术严重性指标重的手术时间、失血量、麻醉方式是POCD的危险因素。2)POCD风险评分系统预测模型:Ln[R/(1-R)]=-6.88+0.191×PS+0.302×OS。3)该模型POCD发生率的预测值为平均12.38%,实际为14.28%,实际值/预测值为1.15,两者之间差异无统计学意义(χ2=0.330,P=0.6670.05)。ROC曲线结果显示Se=53.3%,Sp=90%,?=10%,β=46.7%,ROC曲线下面积为0.759。对该预测模型进行Hosmer-Lemeshow检验,结果显示,此评分系统预测术后POCD发生率(H2=7.707,df=8,P=0.4630.05)效果良好,数据中的信息被充分提取。4.术后死亡组:1)生理学指标中的年龄、白细胞、白蛋白、血压(收缩压)、肌酐、ASA分级、心功能分级、合并症数量、COPD、脑卒中、糖尿病以及手术严重性指标中的手术方式、术前准备时间、手术时间、术中失血量是术后死亡的危险因素。2)术后死亡风险评分系统预测模型:Ln[R/(1-R)]=-11.565+0.265×PS+0.121×OS。3)该模型术后死亡率的预测值为平均3.99%,实际为5.18%,实际值/预测值为1.3,两者之间差异无统计学意义(χ2=0.820,P=0.4510.05)。ROC曲线结果显示Se=96.2%,Sp=88.8%,?=11.2%,β=3.8%,ROC曲线下面积为0.967。对该预测模型进行Hosmer-Lemeshow检验,结果显示,该评分系统预测术后死亡发生率(H2=10.869,df=8,P=0.2090.05)效果良好,数据中的信息被充分提取。结论:本课题初步建立起川南地区老年髋部骨折术后常见并发症发生率和死亡率的评分系统及其预测模型:1)术后肺部感染风险评分系统预测模型:Ln[R/(1-R)]=-7.187+0.226×PS+0.161×OS。2)术后LEDVT风险评分系统预测模型:Ln[R/(1-R)]=-11.493+0.347×PS+0.327×OS。3)POCD风险评分系统预测模型:Ln[R/(1-R)]=-6.88+0.191×PS+0.302×OS。4)术后死亡风险评分系统预测模型:Ln[R/(1-R)]=-11.565+0.265×PS+0.121×OS。并用实际值与预测值的比值、ROC曲线、Hosmer—Lemeshow检验来评估其预测价值,结果显示四个预测模型均具有良好的准确度。本课题针对术后并发症的范围广泛的问题将其进行细化,对常见的不同并发症,具体问题具体分析,得出各自的手术风险因素,排除一些与本并发症无关的风险因素,针对性更强,效率更高。但样本样不足,未进行前瞻性研究,没有将本次研究结果与PPOSSUM以及骨科POSSUM的统计结果相比较,故仍需要进一步的深入研究。
[Abstract]:Objective: to establish a predictive model for the incidence and mortality of common complications after hip fracture in the south of Sichuan, and to test its predictive value. Methods: 1., 4 data collection tables were established, including postoperative pulmonary infection, postoperative cognitive dysfunction (POCD), postoperative deep venous thrombosis (postoperative LEDVT), postoperative death, and postoperative death; 2. The clinical data of these patients who were hospitalized in the Affiliated Hospital of Southwest Medical University from January 2012 to October 2016 were collected and the corresponding data collection table was filled. 3. then the corresponding 4 databases were established by using Epidata3.1 software and the clinical data in the corresponding data collection table were recorded into the database; the data were introduced into the spss19.0 software. Statistical analysis: the measurement data were tested by T, and the count data were analyzed by the x 2 test for single factor analysis. The statistical variables were statistically significant (with =0.05 as the test level and the P value (27) 0.05 variables were statistically significant); 4. was in the physiological and surgical severity score system (physical and operation severity score for the enumeration. On the basis of of mortality and morbidity, POSSUM), these variables are divided into two categories of physiological index and surgical severity index. The scoring system for postoperative lung infection, POCD, postoperative LEDVT, and postoperative death of the aged hip fractures is established. The postoperative pulmonary infection, POCD, postoperative LEDVT, and postoperative death are obtained by Logistic regression analysis. Incidence prediction model; 5. finally, using the ratio of actual value to predicted value, ROC curve, and Hosmer-Lemeshow test to evaluate its predictive value. Results: 1. postoperative pulmonary infection group: 1) age, leukocyte, ASA classification, COPD, cardiac function classification, number of complications, and preoperative preparation time in surgical severity index, operation time, operation time, operation time, operation time, operation time, operation time, operation time, operation time, operation time, operation time, operation time, operation time The amount of blood loss during the operation, the way of anesthesia was the risk factor of pulmonary infection after operation.2) the prediction model of the lung infection risk score system after operation: Ln[R/ (1-R)]=-7.187+0.226 x PS+0.161 x OS.3) the predicted value of the pulmonary infection rate of the model was 8.93%, the actual value was 9.89%, the actual value / prediction value was 1.11, there was no statistical difference between the two (x 2=0.27). 9, P=0.6730.05).ROC curve showed sensitivity (Se) =82.7%, specificity (Sp) =72.4%, misdiagnosis rate (?) =27.6%, missed diagnosis rate (beta) =17.3%, and 0.814. under ROC curve for Hosmer-Lemeshow test of the prediction model. The results showed that the rate of lung complications after the pre test of this scoring system was good, data were good, data were good, data The information in the LEDVT group was fully extracted after.2.: 1) age, FIB, serum triglycerides, serum triglycerides, BMI, varicose veins, hypertension, coronary heart disease, diabetes, stroke, infection, and surgical severity indicators, preoperative preparation time, bleeding volume, and operation time, the risk factor of LEDVT after operation,.2), LEDVT risk score after operation. System prediction model: Ln[R/ (1-R)]=-11.493+0.347 x PS+0.327 x OS.3) the prediction value of LEDVT incidence of the model after operation is 12.57%, the actual value is 13.38%, the actual value / prediction value is 1.06, there is no statistical difference between the two (P=0.7760.05).ROC curve fruit display Se=74.20%, Sp=86.20%, beta, beta area under the curve area Hosmer-Lemeshow test was performed on the 0.87. model for the prediction model. The results showed that the scoring system predicted the LEDVT incidence (H2=3.309, df=8, P=0.9140.05) after operation. The information in the data was fully extracted from the.3.POCD group: 1) age, blood pressure (systolic pressure), albumin, oxygen pressure, number of complications, COPD, stroke and hands in the physiological indexes. The operation time, the amount of blood loss and the way of anesthesia were the risk factor of POCD.2) POCD risk scoring system prediction model: Ln[R/ (1-R)]=-6.88+0.191 x PS+0.302 x OS.3) the predicted value of the POCD incidence of the model was 12.38%, the actual value was 14.28%, the actual value / prediction value was 1.15, there was no statistical difference between the two (x 2=0.330,) P=0.6670.05) the results of the.ROC curve show Se=53.3%, Sp=90%, =10%, beta =46.7%, and 0.759. under ROC curve for the Hosmer-Lemeshow test of this prediction model. The results show that this scoring system predicts the incidence of POCD after operation (H2=7.707, df=8,) is good, and the information in the data is fully extracted after the operation: 1) physiological index The age, white blood cell, albumin, blood pressure (systolic pressure), creatinine, ASA classification, cardiac function classification, complication number, COPD, stroke, diabetes and surgical severity index, preoperative preparation time, operation time, intraoperative blood loss are the risk factors for postoperative death.2) the prediction model of postoperative death risk score system: Ln[ R/ (1-R)]=-11.565+0.265 x PS+0.121 x OS.3) the predictive value of postoperative mortality of the model was 3.99%, the actual value was 5.18%, the actual value / prediction value was 1.3. There was no statistical difference between the two models (P=0.4510.05).ROC curve results showed Se=96.2%, Sp=88.8%, =11.2%, beta =3.8%. Smer-Lemeshow test, the results showed that the scoring system predicted the incidence of postoperative mortality (H2=10.869, df=8, P=0.2090.05), and the information in the data was fully extracted. Conclusion: this project initially established a scoring system for the incidence and mortality of common complications after hip fracture in the south of Sichuan Province and its prediction model: 1) lung after operation. Ln[R/ (1-R)]=-7.187+0.226 x PS+0.161 x OS.2) prediction model of LEDVT risk scoring system after operation: Ln[R/ (1-R)]=-11.493+0.347 x PS+0.327 x OS.3) POCD risk scoring system prediction model: the prediction model of postoperative mortality risk score system 565+0.265 x PS+0.121 x OS. and the ratio of the actual value to the predicted value, the ROC curve and the Hosmer Lemeshow test to evaluate their predictive value. The results show that the four prediction models have good accuracy. Analyze the risk factors of the operation and eliminate some risk factors that are not related to the complications, which are more pertinent and more efficient. However, there is no prospective study on the sample sample and no comparison between the results of this study and the statistical results of the PPOSSUM and Department of orthopedics in the Department of orthopedics, so further in-depth study is needed.

【学位授予单位】：西南医科大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：R687.3

【参考文献】