切换至 "中华医学电子期刊资源库"

中华神经创伤外科电子杂志 ›› 2023, Vol. 09 ›› Issue (01) : 19 -25. doi: 10.3877/cma.j.issn.2095-9141.2023.01.004

临床研究

基于决策树法构建创伤性颅脑损伤术后硬脑膜下积液的风险预测模型
张华1, 刘广明1, 刘国成1, 张姝红1, 陈大莉1, 蒲小龙1, 王志友1, 李倩1,()   
  1. 1. 617300 成都市郫都区人民医院重症医学科
  • 收稿日期:2022-09-16 出版日期:2023-02-15
  • 通信作者: 李倩

Construction of a risk prediction model for post-traumatic subdural hygroma based on decision tree method

Hua Zhang1, Guangming Liu1, Guocheng Liu1, Shuhong Zhang1, Dali Chen1, Xiaolong Pu1, Zhiyou Wang1, Qian Li1,()   

  1. 1. Department of Critical Medicine, Chengdu Pidu District People's Hospital, Chengdu 617300, China
  • Received:2022-09-16 Published:2023-02-15
  • Corresponding author: Qian Li
引用本文:

张华, 刘广明, 刘国成, 张姝红, 陈大莉, 蒲小龙, 王志友, 李倩. 基于决策树法构建创伤性颅脑损伤术后硬脑膜下积液的风险预测模型[J]. 中华神经创伤外科电子杂志, 2023, 09(01): 19-25.

Hua Zhang, Guangming Liu, Guocheng Liu, Shuhong Zhang, Dali Chen, Xiaolong Pu, Zhiyou Wang, Qian Li. Construction of a risk prediction model for post-traumatic subdural hygroma based on decision tree method[J]. Chinese Journal of Neurotraumatic Surgery(Electronic Edition), 2023, 09(01): 19-25.

目的

探讨创伤性颅脑损伤(TBI)术后硬脑膜下积液(PTSH)的影响因素,并构建TBI术后PTSH的决策树风险预测模型。

方法

选取成都市郫都区人民医院重症医学科自2019年6月至2022年6月收治的344例TBI患者作为研究对象,根据是否发生PTSH将其分为观察组和对照组,分析2组患者的临床资料,采用单因素和Logistic回归分析筛选TBI患者发生PTSH的影响因素,采用SPSS Modeler软件构建TBI术后PTSH的决策树预测模型,并分析决策树预测模型的诊断效能。

结果

344例TBI患者行DC后发生PTSH者68例,发生率为19.77%;单因素和多因素Logistic回归分析结果显示,蛛网膜撕裂、入院GCS评分>8分、中线移位、大骨瓣、骨窗面积>150 cm2、骨瓣边缘距中线距离>2 cm是TBI患者行DC后发生PTSH的影响因素(P<0.05);决策树模型选出了蛛网膜撕裂、入院GCS评分>8分、中线移位、大骨瓣、骨窗面积>150 cm2、骨瓣边缘距中线距离>2 cm等6个解释变量作为模型的节点,其中蛛网膜撕裂是最重要的预测因子。ROC曲线分析显示,决策树模型的曲线下面积(AUC)值为0.895,Logistic回归模型的AUC值为0.881,决策树模型的预测效能优于Logistic回归模型(Z=2.423,P=0.013)。

结论

蛛网膜撕裂、入院GCS评分>8分、中线移位、大骨瓣、骨窗面积>150 cm2、骨瓣边缘距中线距离>2 cm是TBI患者行DC后发生PTSH的影响因素,构建的决策树模型有助于筛查DC后发生PTSH高风险人群和指导临床制定科学的防治策略,具有较高的临床价值。

Objective

To investigate the influencing factors of post-traumatic subdural hygroma (PTSH) and to construct a decision tree risk prediction model for PTSH after traumatic brain injury (TBI) operation.

Methods

From June 2019 to June 2022, 344 patients with TBI who were admitted to Intensive Care Medicine Department of Pidu District People's Hospital of Chengdu were selected as the study objects. According to whether PTSH occurred, they were divided into observation group and control group. The clinical data of the two groups were analyzed. The influencing factors of PTSH development in TBI patients were screened by univariate and Logistic regression analysis. The decision tree prediction model of PTSH after TBI was constructed by SPSS Modeler software. The diagnostic efficacy of the decision tree prediction model was analyzed.

Results

Sixty-eight of 344 TBI patients developed PTSH after DC, with an incidence of 19.77%. The results of univariate and multivariate Logistic regression analysis showed that arachnoid tear, admission GCS score >8 points, midline shift, large bone flap, bone window area >150 cm2, distance from the edge of bone flap to the midline >2 cm were the influencing factors of PTSH in TBI patients after DC (P<0.05). Six explanatory variables were selected as nodes of the decision tree model, including arachnoid tear, admission GCS score >8 points, midline shift, large bone flap, bone window area >150 cm2, and distance from the edge of bone flap to the midline >2 cm, among which arachnoid tear is the most important predictor of the model. The ROC curve analysis showed that the area under curve (AUC) value of the decision tree model was 0.895, and the AUC value of the Logistic regression model was 0.881. The prediction efficiency of the decision tree model was better than that of the Logistic regression model (Z=2.423, P=0.013).

Conclusion

Arachnoid tear, admission GCS score >8 points, midline shift, large bone flap, bone window area >150 cm2, and the distance from the edge of bone flap to the midline >2 cm are the influential factors of PTSH in patients with TBI after DC. The decision tree model can help screen people at high risk of PTSH after DC, and guide clinical formulation of scientific prevention and treatment strategies. It has high clinical value.

表1 2组患者临床资料比较[例(%)]
Tab.1 Comparison of clinical data of two groups [n(%)]
项目 观察组(n=68) 对照组(n=276) χ2 P
性别     0.041 0.839
38(55.88) 158(57.25)    
30(41.12) 118(42.75)    
年龄(岁)     0.568 0.451
>60 35(51.47) 128(46.38)    
≤60 33(48.53) 148(53.62)    
BMI(kg/m2     0.213 0.644
>28 26(38.24) 114(41.30)    
≤28 42(61.76) 162(58.70)    
蛛网膜撕裂     12.031 0.001
44(64.71) 114(41.30)    
24(35.29) 162(58.70)    
蛛网膜下腔出血     3.383 0.066
40(58.82) 128(46.38)    
28(41.18) 148(53.62)    
脑室内出血     3.806 0.051
37(54.41) 114(41.30)    
31(45.59) 162(58.70)    
脑内血肿     3.116 0.078
49(72.06) 167(60.51)    
19(27.94) 109(39.49)    
硬膜下血肿     2.041 0.153
43(63.24) 148(53.62)    
25(36.76) 128(46.38)    
硬膜外血肿     0.282 0.596
39(57.35) 168(60.87)    
29(42.65) 108(39.13)    
入院GCS评分(分)     11.938 0.001
>8 28(41.18) 177(64.13)    
≤8 40(58.82) 99(35.87)    
皮层切开     3.610 0.057
32(47.06) 163(59.06)    
36(52.94) 113(40.94)    
中线移位     11.146 0.001
40(58.82) 101(36.59)    
28(41.18) 175(63.41)    
大骨瓣     10.342 0.002
37(54.41) 92(33.33)    
31(45.59) 184(66.67)    
去骨瓣侧边     0.282 0.595
双侧 37(54.41) 160(57.97)    
单侧 31(45.59) 116(42.03)    
骨窗面积(cm2     10.597 0.001
>150 38(55.88) 95(34.42)    
≤150 30(41.12) 181(65.58)    
骨瓣边缘距中线距离(cm)   11.833 0.001
>2 27(39.71) 173(62.68)    
≤2 41(60.29) 103(37.32)    
颅内感染     2.192 0.139
31(45.59) 99(35.87)    
37(54.41) 177(64.13)    
脑梗死     3.623 0.057
30(41.12) 88(31.88)    
38(55.88) 188(68.12)    
脑疝     2.240 0.134
25(36.76) 76(27.54)    
43(63.24) 200(72.46)    
脱水药使用的时间(d)     0.493 0.483
>7 32(47.06) 143(51.81)    
≤7 41(52.94) 133(48.19)    
表2 多因素Logistic回归分析的变量赋值表
Tab.2 Variable assignment table of multi-factor Logistic regression analysis
表3 TBI患者行DC后发生PTSH的多因素Logistic回归分析
Tab.3 Multivariate logistic regression analysis of PTSH in TBI patients after DC
图1 TBI患者行DC后发生PTSH决策树模型
Fig.1 Decision tree model of PTSH after DC in TBI patients
图2 决策树模型与Logistic回归模型的ROC曲线
Fig.2 ROC curve of decision tree model and Logistic regression model
表4 决策树模型与Logistic回归模型的分类效果比较
Tab.4 Comparison of classification effects between decision tree model and Logistic regression model
[1]
冯华,李文龚,谭亮,等.颅脑创伤病情评估多模态监测系统[J].中国现代神经疾病杂志, 2020, 20(7): 568-576. DOI: 10.3969/j.issn.1672-6731.2020.07.002.
[2]
Brown AC, Lavik E, Stabenfeldt SE. Biomimetic strategies to treat traumatic brain injury by leveraging fibrinogen[J]. Bioconjug Chem, 2019, 30(7): 1951-1956. DOI: 10.1021/acs.bioconjchem.9b00360.
[3]
Wei S. Clinical value of bilateral balanced frontotemporoparietal decompressive craniectomy in severe diffuse traumatic brain injury[J]. J Craniofac Surg, 2022, 33(1): 279-283. DOI: 10.1097/SCS.0000000000008121.
[4]
田润发,董金千,刘伟明,等.颅脑外伤后无瞳孔散大患者行去骨瓣减压术后的早期结果及出院死亡率的预后相关因素分析[J].中华神经创伤外科电子杂志, 2021, 7(2): 75-81. DOI: 10.3877/cma.j.issn.2095-9141.2021.02.003.
[5]
徐亚,陈华群,朱卫国,等.颅脑创伤后硬脑膜下积液的危险因素分析[J].临床神经外科杂志, 2018, 15(1): 65-67. DOI: 10.3969/j.issn.1672-7770.2018.01.016.
[6]
陈文鑫,晋聪,王婷,等.基于决策树模型与logistic回归模型的HBV宫内传播影响因素分析[J].中华流行病学杂志, 2022, 43(1): 85-91. DOI: 10.3760/cma.j.cn112338-20210630-00511.
[7]
樊雯婧,楼冬洁,卢新,等.基于决策树法和Logistic回归预测神经外科老年住院患者医院感染风险的研究[J].中华医院感染学杂志, 2020, 30(6): 878-883. DOI: 10.11816/cn.ni.2020-193101.
[8]
Gill MR, Reiley DG, Green SM. Interrater reliability of Glasgow coma scale scores in the emergency department[J]. Ann Emerg Med, 2004, 43(2): 215-223. DOI: 10.1016/s0196-0644(03)00814-x.
[9]
王雷平,吴崇光,姚军.颅脑损伤去骨瓣减压术后并发创伤后脑积水的危险因素[J].中华创伤杂志, 2014, 30(4): 307-310. DOI: 10.3760/cma.j.issn.1001-8050.2014.04.005.
[10]
易盼,王存祖,陈品,等.创伤性硬膜下积液转化为慢性硬膜下血肿危险因素分析[J].中国现代神经疾病杂志, 2020, 20(7): 602-607. DOI: 10.3969/j.issn.1672-6731.2020.07.008.
[11]
侯小山,金鹏,魏文峰,等.颅脑外伤去骨瓣减压术后继发硬膜下积液的危险因素分析[J].国际神经病学神经外科学杂志, 2017, 44(5): 468-471. DOI: 10.16636/j.cnki.jinn.2017.05.004.
[12]
范英俊,王琼芬,杨中鑫,等.不同方案治疗颅脑损伤去骨瓣减压术后并发进展型硬膜下积液的临床疗效观察[J].第三军医大学学报, 2020, 42(8): 835-840. DOI: 10.16016/j.1000-5404.201912170.
[13]
Lu VM, Carlstrom LP, Perry A, et al. Prognostic significance of subdural hygroma for post-traumatic hydrocephalus after decompressive craniectomy in the traumatic brain injury setting: a systematic review and meta-analysis[J]. Neurosurg Rev, 2021, 44(1): 129-138. DOI: 10.1007/s10143-019-01223-z.
[14]
褚正民,王耿焕,沈建国,等.早期颅骨修补术治疗去骨瓣减压术后难治性硬膜下积液[J].中华创伤杂志, 2019, 35(11): 1024-1026. DOI: 10.3760/cma.j.issn.1001-8050.2019.11.011.
[15]
Yuan Q, Wu X, Yu J, et al. Subdural hygroma following decompressive craniectomy or non-decompressive craniectomy in patients with traumatic brain injury: clinical features and risk factors[J]. Brain Inj, 2015, 29(7-8): 971-80. DOI: 10.3109/02699052.2015.1004760.
[16]
申向竹,张山,韩亚非,等.联合手术对脑外伤患者的临床效果及术后硬膜下积液的危险因素分析[J].实用临床医药杂志, 2021, 25(11): 22-25. DOI: 10.7619/jcmp.20210484.
[17]
赵思任,张建斌,张吉论,等.创伤性颅脑损伤患者去骨瓣术后脑积水风险列线图模型构建及其价值评估[J].创伤外科杂志, 2022, 24(4): 259-264. DOI: 10.3969/j.issn.1009-4237.2022.04.005.
[18]
杜勇,王革生,王雷,等.去骨瓣减压术治疗创伤性颅脑损伤术后并发症发生情况及危险因素分析[J].解放军医药杂志, 2017, 29(7): 41-43, 53. DOI: 10.3969/j.issn.2095-140X.2017.07.011.
[19]
吴小明,罗雨,王存祖,等.颅脑外伤去骨瓣减压术后并发硬脑膜下积液危险因素分析[J].江苏医药, 2018, 44(10): 1179-1181. DOI: 10.19460/j.cnki.0253-3685.2018.10.025.
[20]
Kang HJ, Kang H, Kim B, et al. Evaluation of the diagnostic performance of a decision tree model in suspected acute appendicitis with equivocal preoperative computed tomography findings compared with Alvarado, Eskelinen, and adult appendicitis scores: a STARD compliant article[J]. Medicine (Baltimore), 2019, 98(40): e17368. DOI: 10.1097/MD.0000000000017368.
[21]
闫瑞平,王习亮,姚粉霞,等.决策树模型与Logistic回归分析模型识别高血压危险因素的效果比较[J].中华疾病控制杂志, 2022, 26(2): 218-222. DOI: 10.16462/j.cnki.zhjbkz.2022.02.016.
[1] 高建松, 陈晓晓, 冯婷, 包剑锋, 魏淑芳, 潘林. 基于超声瞬时弹性成像的多参数决策树模型评估慢性乙型肝炎患者肝纤维化等级[J]. 中华医学超声杂志(电子版), 2023, 20(09): 923-929.
[2] 刘欢颜, 华扬, 贾凌云, 赵新宇, 刘蓓蓓. 颈内动脉闭塞病变管腔结构和血流动力学特征分析[J]. 中华医学超声杂志(电子版), 2023, 20(08): 809-815.
[3] 马艳波, 华扬, 刘桂梅, 孟秀峰, 崔立平. 中青年人颈动脉粥样硬化病变的相关危险因素分析[J]. 中华医学超声杂志(电子版), 2023, 20(08): 822-826.
[4] 黄应雄, 叶子, 蒋鹏, 詹红, 姚陈, 崔冀. 急性肠系膜静脉血栓形成致透壁性肠坏死的临床危险因素分析[J]. 中华普通外科学文献(电子版), 2023, 17(06): 413-421.
[5] 张俊, 罗再, 段茗玉, 裘正军, 黄陈. 胃癌预后预测模型的研究进展[J]. 中华普通外科学文献(电子版), 2023, 17(06): 456-461.
[6] 张再博, 王冰雨, 焦志凯, 檀碧波. 胃癌术后下肢深静脉血栓危险因素的Meta分析[J]. 中华普通外科学文献(电子版), 2023, 17(06): 475-480.
[7] 唐旭, 韩冰, 刘威, 陈茹星. 结直肠癌根治术后隐匿性肝转移危险因素分析及预测模型构建[J]. 中华普外科手术学杂志(电子版), 2024, 18(01): 16-20.
[8] 吴方园, 孙霞, 林昌锋, 张震生. HBV相关肝硬化合并急性上消化道出血的危险因素分析[J]. 中华普外科手术学杂志(电子版), 2024, 18(01): 45-47.
[9] 陈旭渊, 罗仕云, 李文忠, 李毅. 腺源性肛瘘经手术治疗后创面愈合困难的危险因素分析[J]. 中华普外科手术学杂志(电子版), 2024, 18(01): 82-85.
[10] 晏晴艳, 雍晓梅, 罗洪, 杜敏. 成都地区老年转移性乳腺癌的预后及生存因素研究[J]. 中华普外科手术学杂志(电子版), 2023, 17(06): 636-638.
[11] 莫闲, 杨闯. 肝硬化患者并发门静脉血栓危险因素的Meta分析[J]. 中华普外科手术学杂志(电子版), 2023, 17(06): 678-683.
[12] 毛永欢, 仝瀚文, 缪骥, 王行舟, 沈晓菲, 喻春钊. 造口旁疝危险因素预测模型构建[J]. 中华疝和腹壁外科杂志(电子版), 2023, 17(06): 682-687.
[13] 陆猛桂, 黄斌, 李秋林, 何媛梅. 蜂蛰伤患者发生多器官功能障碍综合征的危险因素分析[J]. 中华临床医师杂志(电子版), 2023, 17(9): 1010-1015.
[14] 李达, 张大涯, 陈润祥, 张晓冬, 黄士美, 陈晨, 曾凡, 陈世锔, 白飞虎. 海南省东方市幽门螺杆菌感染现状的调查与相关危险因素分析[J]. 中华临床医师杂志(电子版), 2023, 17(08): 858-864.
[15] 李琪, 黄钟莹, 袁平, 关振鹏. 基于某三级医院的ICU多重耐药菌医院感染影响因素的分析[J]. 中华临床医师杂志(电子版), 2023, 17(07): 777-782.
阅读次数
全文


摘要