Analysis of influencing factors of diabetic foot ulcer recurrence and construction of the Bayesian network model
-
摘要:
目的 探讨DFU复发的影响因素,构建预测糖尿病足溃疡（DFU）复发的贝叶斯网络模型,分析各影响因素间的关系及其对DFU复发的影响程度。 方法 该研究为回顾性队列研究。2016年4月—2020年4月,天津医科大学朱宪彝纪念医院糖尿病足科收治1 656例符合入选标准的住院DFU患者,其中男1 127例、女529例,年龄18~80岁。根据出院后5年内DFU复发情况,将患者分为复发组（552例）和未复发组（1 104例）。比较2组患者的性别,糖尿病病程、视网膜病变情况、溃疡部位、足部畸形、足截肢史,入院当天身体质量指数（BMI）、收缩压及D-二聚体、纤维蛋白原、总胆固醇、低密度脂蛋白（LDL）水平,住院期间糖化血红蛋白（HbA1c）水平,降糖方案、使用阿司匹林和他汀类药物情况。通过多因素logistic回归模型筛选影响患者DFU复发的独立危险因素；采用限制性立方样条模型分析在调整协变量后住院期间HbA1c水平与DFU复发的剂量-反应关系；构建预测DFU复发的贝叶斯网络模型,分析各影响因素间的关系及其对DFU复发的影响程度。绘制受试者操作特征（ROC）曲线评估贝叶斯网络模型对DFU复发的预测效能。 结果 2组患者性别、视网膜病变情况、溃疡部位、足部畸形、足截肢史、降糖方案、使用阿司匹林和他汀类药物情况（χ2分别为5.87、6.14、24.97、159.63、4.03、39.07、26.60、4.69,P<0.05）,糖尿病病程,入院当天BMI、收缩压和D-二聚体、纤维蛋白原、总胆固醇、LDL水平及住院期间HbA1c水平（Ζ值分别为-7.20、-6.14、-5.57、-2.24、-9.09、-2.08、-7.74、-11.32,P<0.05）比较,差异均有统计学意义。多因素logistic回归分析显示,入院当天BMI、收缩压、纤维蛋白原水平、LDL水平,住院期间HbA1c水平,糖尿病病程、溃疡部位、足部畸形,注射胰岛素、口服降糖药物+胰岛素等降糖方案,以及使用阿司匹林是DFU患者复发的独立危险因素（OR分别为1.06、1.01、1.31、1.47、1.19、1.04、1.71、6.59、1.82、1.60、1.85,95%CI分别为1.03~1.10、1.01~1.02、1.21~1.43、1.29~1.68、1.13~1.26、1.03~1.06、1.26~2.31、4.67~9.38、1.19~2.80、1.18~2.16、1.45~2.38）。限制性立方样条模型显示,调整协变量后,DFU患者住院期间HbA1c水平≥9.0%与DFU复发风险存在非线性剂量-反应关系（χ2=15.01,P<0.05）。构建的贝叶斯网络模型由10个节点、14条有向边组成,溃疡部位、使用阿司匹林情况、足部畸形、入院当天LDL水平和BMI与DFU复发直接相关；足部畸形对DFU复发的影响最大,其次是溃疡部位和使用阿司匹林情况。贝叶斯网络模型的ROC曲线下面积为0.74（95%CI为0.70~0.78）。 结论 入院当天BMI、收缩压及纤维蛋白原、LDL水平,住院期间HbA1c水平,糖尿病病程、溃疡部位、足部畸形,以及降糖方案、使用阿司匹林情况是患者DFU复发的独立危险因素,溃疡部位、使用阿司匹林情况、足部畸形、入院当天LDL水平和BMI与DFU复发直接相关。 Abstract:Objective To explore the influencing factors of diabetic foot ulcer (DFU) recurrence, construct the Bayesian network model for predicting DFU recurrence, and investigate the interrelationships among these influencing factors and their effects on DFU recurrence. Methods This study was a retrospective cohort study. From April 2016 to April 2020, 1 656 hospitalized DFU patients who met the inclusion criteria were admitted to the Diabetic Podiatry Department of Tianjin Medical University Chu Hsien-I Memorial Hospital, including 1 127 males and 529 females, aged 18-80 years. According to DFU recurrence status within 5 years after discharge, the patients were divided into recurrence group (552 cases) and non-recurrence group (1 104 cases). The following indicators were compared between the two groups, including gender, duration of diabetes, retinopathy status, ulcer location, foot deformity, history of foot amputation, body mass index (BMI), systolic blood pressure, levels of D-dimer, fibrinogen, total cholesterol, and low-density lipoprotein (LDL) on the day of admission, level of glycated hemoglobin A1c (HbA1c) during hospitalization, hypoglycemic regimen, and use of aspirin and statins. A multivariate logistic regression model was used to identify independent risk factors for DFU recurrence. A restricted cubic spline model was applied to analyze the dose-response relationship between level of HbA1c during hospitalization and DFU recurrence after adjusting covariates. A Bayesian network model for predicting DFU recurrence was established to analyze the interrelationships among these influencing factors and their effects on DFU recurrence. The receiver operating characteristic (ROC) curve was plotted to evaluate the predictive performance of the Bayesian network model for DFU recurrence. Results There were statistically significant differences between the two groups of patients in gender, retinopathy status, ulcer location, foot deformity, history of foot amputation, hypoglycemic regimen, and use of aspirin and statins (with χ² values of 5.87, 6.14, 24.97, 159.63, 4.03, 39.07, 26.60, and 4.69, respectively, P<0.05), as well as duration of diabetes, BMI, systolic blood pressure, and levels of D-dimer, fibrinogen, total cholesterol, and LDL on the day of admission, and level of HbA1c during hospitalization (with Z values of -7.20, -6.14, -5.57, -2.24, -9.09, -2.08, -7.74, and -11.32, respectively, P<0.05). Multivariate logistic regression analysis showed that BMI, systolic blood pressure, levels of fibrinogen and LDL on the day of admission, level of HbA1c during hospitalization, duration of diabetes, ulcer location, foot deformity, hypoglycemic regimens of injecting insulin and oral hypoglycemic agents + injecting insulin, and use of aspirin were the independent risk factors for DFU recurrence (with OR values of 1.06, 1.01, 1.31, 1.47, 1.19, 1.04, 1.71, 6.59, 1.82, 1.60, and 1.85, respectively, 95%CI of 1.03-1.10, 1.01-1.02, 1.21-1.43, 1.29-1.68, 1.13-1.26, 1.03-1.06, 1.26-2.31, 4.67-9.38, 1.19-2.80, 1.18-2.16, and 1.45-2.38, respectively, P<0.05). Restricted cubic spline model showed a non-linear dose-response relationship between level of HbA1c ≥9.0% during hospitalization and DFU recurrence risk after adjusting covariates (χ²=15.01, P<0.05). The constructed Bayesian network model consisted of 10 nodes and 14 directed edges. The model demonstrated that ulcer location, use of aspirin, foot deformity, and levels of LDL and BMI on the day of admission were directly associated with DFU recurrence; foot deformity posed the greatest risk for DFU recurrence, followed by ulcer location and use of aspirin. The area under the ROC curve for the Bayesian network model was 0.74 (95%CI of 0.70-0.78). Conclusions The BMI, systolic blood pressure, levels of fibrinogen and LDL on the day of admission, level of HbA1c during hospitalization, duration of diabetes, ulcer location, foot deformity, hypoglycemic regimens, and use of aspirin were the independent risk factors for DFU recurrence. The ulcer location, use of aspirin, foot deformity, and levels of LDL and BMI on the day of admission were directly associated with DFU recurrence. -
Key words:
- Diabetic foot /
- Ulcer /
- Recurrence /
- Root cause analysis /
- Bayesian network /
- Restricted cubic spline
-
参考文献
(37) [1] ArmstrongDG, TanTW, BoultonAJM, et al. Diabetic foot ulcers: a review[J]. JAMA, 2023, 330(1):62-75. DOI: 10.1001/jama.2023.10578. [2] ArmstrongDG, BoultonAJM, BusSA. Diabetic foot ulcers and their recurrence[J]. N Engl J Med, 2017, 376(24):2367-2375. DOI: 10.1056/NEJMra1615439. [3] KhalifaWA.Risk factors for diabetic foot ulcer recurrence: a prospective 2-year follow-up study in Egypt[J].Foot (Edinb),2018,35:11-15.DOI: 10.1016/j.foot.2017.12.004. [4] HuangZH,LiSQ,KouY,et al.Risk factors for the recurrence of diabetic foot ulcers among diabetic patients: a meta-analysis[J].Int Wound J,2019,16(6):1373-1382.DOI: 10.1111/iwj.13200. [5] 胡慧萍,陈明卫.糖尿病足溃疡愈合及截肢的相关因素分析[J].安徽医科大学学报,2016(11):1634-1637.DOI: 10.19405/j.cnki.issn1000-1492.2016.11.018. [6] 王柯鑫,张国艳,丁利萍,等.基于CiteSpace慢性病经济毒性研究热点的可视化分析[J].医学与哲学,2023,44(22):76-80.DOI: 10.12014/j.issn.1002-0772.2023.22.17. [7] 中华医学会烧伤外科学分会,长三角一体化糖尿病足专病联盟,《中华烧伤与创面修复杂志》编辑委员会.中国糖尿病足防治实践指南(Ⅰ)[J].中华烧伤与创面修复杂志,2025,41(11):1029-1049.DOI: 10.3760/cma.j.cn501225-20250801-00345. [8] 中华医学会烧伤外科学分会,长三角一体化糖尿病足专病联盟,《中华烧伤与创面修复杂志》编辑委员会. 中国糖尿病足防治实践指南(Ⅱ)[J]. 中华烧伤与创面修复杂志,2025,41(12):1111-1131. DOI: 10.3760/cma.j.cn501225-20251029-00448. [9] Monteiro-SoaresM,BoykoEJ,RibeiroJ,et al.Risk stratification systems for diabetic foot ulcers: a systematic review[J].Diabetologia,2011,54(5):1190-1199.DOI: 10.1007/s00125-010-2030-3. [10] 康艳楠 糖尿病足复发的影响因素与风险预测模型研究 天津 天津中医药大学 2022 DOI: 10.27368/d.cnki.gtzyy.2022.000286 康艳楠.糖尿病足复发的影响因素与风险预测模型研究[D].天津:天津中医药大学,2022. DOI:10.27368/d.cnki.gtzyy.2022.000286.
[11] PetersenBJ, Linde-ZwirbleWT, TanTW, et al. Higher rates of all-cause mortality and resource utilization during episodes-of-care for diabetic foot ulceration[J]. Diabetes Res Clin Pract, 2022,184:109182. DOI: 10.1016/j.diabres.2021.109182. [12] ChenD,WangM,ShangX,et al.Development and validation of an incidence risk prediction model for early foot ulcer in diabetes based on a high evidence systematic review and meta-analysis[J].Diabetes Res Clin Pract,2021,180:109040.DOI: 10.1016/j.diabres.2021.109040. [13] WangM,ChenD,FuH,et al.Development and validation of a risk prediction model for the recurrence of foot ulcer in type 2 diabetes in China: a longitudinal cohort study based on a systematic review and meta-analysis[J].Diabetes Metab Res Rev,2023,39(4):e3616.DOI: 10.1002/dmrr.3616. [14] 王美君,许洪梅,葛甜甜,等.糖尿病足患者溃疡愈合后5年复发和死亡的随访结果及其危险因素分析[J].中华糖尿病杂志,2021,13(3):227-232.DOI: 10.3760/cma.j.cn115791-20200707-00422. [15] ChenD,WangM,ShangX,et al.Development and validation of an incidence risk prediction model for early foot ulcer in diabetes based on a high evidence systematic review and meta-analysis[J].Diabetes Res Clin Pract,2021,180:109040.DOI: 10.1016/j.diabres.2021.109040. [16] 莫泽纬,陈道雄,高勇义,等.初发糖尿病足溃疡患者足溃疡复发影响因素分析[J].中国热带医学,2018,18(7):716-719.DOI: 10.13604/j.cnki.46-1064/r.2018.07.21. [17] 王银荣,邱晓堂,潘莉. 糖尿病足溃疡患者愈合后复发的影响因素分析及风险预测模型构建[J]. 军事护理,2023,40(5):62-65. DOI: 10.3969/j.issn.2097-1826.2023.05.015. [18] 刘星雨,饶庆雨,刘祖婷,等.基于贝叶斯网络的个体化风险评估模型及R语言实现[J].中国循证医学杂志,2024,24(11):1325-1331. DOI: 10.7507/1672-2531.202405072. [19] LauCL,SmithCS.Bayesian networks in infectious disease eco-epidemiology[J].Rev Environ Health,2016,31(1):173-177.DOI: 10.1515/reveh-2015-0052. [20] ManciniF,SousaFS,HummelAD,et al.Classification of postural profiles among mouth-breathing children by learning vector quantization[J].Methods Inf Med,2011,50(4):349-357.DOI: 10.3414/ME09-01-0039. [21] 魏珍,张雪雷,饶华祥,等. 禁忌搜索算法的贝叶斯网络模型在冠心病影响因素分析中的应用[J]. 中华流行病学杂志,2016,37(6):895-899. DOI: 10.3760/cma.j.issn.0254-6450.2016.06.031. [22] BusSA,LaveryLA,Monteiro-SoaresM,et al.Guidelines on the prevention of foot ulcers in persons with diabetes (IWGDF 2019 update)[J].Diabetes Metab Res Rev,2020,36Suppl 1:e3269.DOI: 10.1002/dmrr.3269. [23] van NettenJJ,BusSA,ApelqvistJ,et al.Definitions and criteria for diabetes-related foot disease (IWGDF 2023 update)[J].Diabetes Metab Res Rev,2024,40(3):e3654.DOI: 10.1002/dmrr.3654. [24] van NettenJJ, BusSA, ApelqvistJ, et al. Definitions and criteria for diabetic foot disease[J]. Diabetes Metab Res Rev, 2020, 36Suppl 1:e3268. DOI: 10.1002/dmrr.3268. [25] 罗剑锋,金欢,李宝月,等.限制性立方样条在非线性回归中的应用研究[J].中国卫生统计,2010,27(3):229-232.DOI: 10.3969/j.issn.1002-3674.2010.03.002. [26] 解其华,魏迎凤,徐园园,等.糖尿病足溃疡感染患者血清脂联素、D二聚体及NLR水平与病情严重程度及预后的相关性[J].中华内分泌外科杂志,2023,17(6):670-674.DOI: 10.3760/cma.j.cn.115807-20220722-00198. [27] 梅柯强,刘泽慧,祝蓉,等.2型糖尿病足溃疡加重的危险因素及细菌感染特征分析[J].重庆医科大学学报,2025,50(6):770-777.DOI: 10.13406/j.cnki.cyxb.003761. [28] 郭晓蕙,纪立农,陆菊明,等.2009年中国成人2型糖尿病患者口服降糖药联合胰岛素治疗后血糖达标状况调查[J].中华糖尿病杂志,2012,4(8):474-478. DOI: 10.3760/cma.j.issn.1674-5809.2012.08.007. [29] LiXH, GuanLY, LinHY. Fibrinogen: a marker in predicting diabetic foot ulcer severity[J]. J Diabetes Res, 2016:2358321. DOI: 10.1155/2016/2358321. [30] 黄泳清,沈洁,袁思捷,等.红细胞分布宽度、纤维蛋白原与D-二聚体预测糖尿病足发生发展风险的价值[J].中国组织工程研究,2019,23(23):3716-3721.DOI: 10.3969/j.issn.2095-4344.1319. [31] 王娟,赵子源,李凤翱.糖尿病足感染患者炎症因子与D-二聚体和CD64及纤维蛋白原对病情严重程度评估及预后价值[J].中华医院感染学杂志,2020,30(3):373-377.DOI: 10.11816/cn.ni.2020-190583. [32] PinzurM, FreelandR, JuknelisD. The association between body mass index and foot disorders in diabetic patients[J]. Foot Ankle Int, 2005, 26(5):375-377.DOI: 10.1177/107110070502600506. [33] AhadiM,MolooghiK,MasoudifarN,et al.A review of non-alcoholic fatty liver disease in non-obese and lean individuals[J].J Gastroenterol Hepatol,2021,36(6):1497-1507.DOI: 10.1111/jgh.15353. [34] KeukenkampR,Busch-WestbroekTE,BarnR,et al.Foot ulcer recurrence, plantar pressure and footwear adherence in people with diabetes and Charcot midfoot deformity: a cohort analysis[J].Diabet Med,2021,38(4):e14438.DOI: 10.1111/dme.14438. [35] 申金付,蒋瑞妹,王卓群,等. 2型糖尿病患者糖尿病足溃疡复发情况及影响因素[J]. 中华烧伤杂志,2020,36(10):947-952. DOI: 10.3760/cma.j.cn501120-20190726-00315. [36] SunY,ZhouY,DaiY,et al.Predictors of post-healing recurrence in patients with diabetic foot ulcers: a systematic review and meta-analysis[J].J Tissue Viability,2024,33(4):542-549.DOI: 10.1016/j.jtv.2024.07.002. [37] ZhangJ, ShangS, LiuW, et al. Effect of aspirin on platelet-rich plasma of diabetes mellitus with lower extremity atherosclerosis[J]. Future Sci OA, 2024,10(1):2413827. -
图 1 1 656例患者住院期间HbA1c水平与糖尿病足溃疡复发的关系限制性立方样条曲线
注：红色虚线表示参考线,红色实线表示OR,红色区域表示95%CI,黑色虚线对应OR为1.0；调整的协变量包括入院当天身体质量指数、收缩压、纤维蛋白原水平、低密度脂蛋白水平及糖尿病病程、溃疡部位、足部畸形、降糖方案、使用阿司匹林情况；HbA1c为糖化血红蛋白
图 2 预测1 656例患者糖尿病足溃疡复发的贝叶斯网络模型
注：复发指出院后5年内糖尿病足溃疡复发；每个节点对应的百分比为患者占比；线条越粗代表对DFU复发的影响越大；颜色越红代表该因素对DFU复发的敏感性越高；①、②分别表示入院当天纤维蛋白原水平2.0~4.0 g/L、<2.0或>4.0 g/L,③、④分别表示入院当天收缩压<120 mmHg（1 mmHg=0.133 kPa）、≥120 mmHg,⑤、⑥分别表示住院期间糖化血红蛋白（HbA1c）水平<9.0%、≥9.0%,⑦、⑧、⑨、⑩分别表示糖尿病病程<10年、10~19年、>19~29年、>29年,⑪、⑫、⑬分别表示降糖方案中口服降糖药、胰岛素、口服降糖药+胰岛素,⑭、⑮分别表示入院当天低密度脂蛋白（LDL）水平<3.4 mmol/L、≥3.4 mmol/L,⑯、⑰、⑱、⑲分别表示入院当天身体质量指数（BMI）<18.5 kg/m2、18.5~24.9 kg/m2、>24.9~29.9 kg/m2、>29.9 kg/m2
图 3 预测1 656例患者糖尿病足溃疡复发的贝叶斯网络模型推理
注：①、②分别表示入院当天纤维蛋白原2.0~4.0 g/L、<2.0或>4.0 g/L,③、④分别表示入院当天收缩压<120 mmHg（1 mmHg=0.133 kPa）、≥120 mmHg,⑤、⑥分别表示住院期间糖化血红蛋白（HbA1c）水平<9.0%、≥9.0 %,⑦、⑧、⑨、⑩分别表示糖尿病病程<10年、10~19年、>19~29年、>29年,⑪、⑫、⑬分别表示降糖方案中口服降糖药、胰岛素、口服降糖药+胰岛素,⑭、⑮分别表示入院当天低密度脂蛋白（LDL）水平<3.4 mmol/L、≥3.4 mmol/L,⑯、⑰、⑱、⑲分别表示入院当天身体质量指数（BMI）<18.5 kg/m2、18.5~24.9 kg/m2、>24.9~29.9 kg/m2、>29.9 kg/m2
Table 1. 2组糖尿病足溃疡患者一般资料比较
组别 例数 年龄[岁,M（Q1,Q3）] 性别（例） BMI[kg/m2,M（Q1,Q3）] 收缩压[mmHg,M（Q1,Q3）] 舒张压[mmHg,M（Q1,Q3）] 男 女 复发组 552 64.00（56.25,69.00） 354 198 26.2（23.6,28.9） 140（130,155） 79（70,81） 非复发组 1 104 64.00（55.00,71.00） 773 331 24.8（22.3,27.4） 135（125,149） 80（70,83） 统计量值 Ζ=-0.57 χ2=5.87 Ζ=-6.14 Ζ=-5.57 Ζ=-0.74 P值 0.570 0.009 0.001 0.001 0.458 注：复发指出院后5年内糖尿病足溃疡复发；除糖化血红蛋白（HbA1c）为住院期间的水平,其他检测指标均为入院当天的水平；BMI为身体质量指数,LDL为低密度脂蛋白,HDL为高密度脂蛋白；1 mmHg=0.133 kPa 
下载: 导出CSV
Table 2. 1 656例患者DFU复发的多因素logistic回归分析结果
因素 OR 95%CI P值 性别 1.00 0.77~1.31 0.984 入院当天BMI（kg/m2） 1.06 1.03~1.10 <0.001 入院当天收缩压（mmHg） 1.01 1.01~1.02 0.002 糖尿病病程（年） 1.04 1.03~1.06 <0.001 视网膜病变（以无视网膜病变为参照） 1.09 0.85~1.41 0.494 溃疡部位（以非足底为参照） 1.71 1.26~2.31 <0.001 足部畸形（以无足部畸形为参照） 6.59 4.67~9.38 <0.001 足截肢史（以无截肢史为参照） 0.90 0.69~1.16 0.413 住院期间HbA1c水平（%） 1.19 1.13~1.26 <0.001 入院当天纤维蛋白原水平（g/L） 1.31 1.21~1.43 <0.001 入院当天LDL水平（mmol/L） 1.47 1.29~1.68 <0.001 降糖方案（以口服降糖药物为参照） 注射胰岛素 1.82 1.19~2.80 <0.001 口服降糖药物+胰岛素 1.60 1.18~2.16 <0.001 使用阿司匹林（以未使用阿司匹林为参照） 1.85 1.45~2.38 <0.001 注：BMI为身体质量指数,HbA1c为糖化血红蛋白,LDL为低密度脂蛋白；1 mmHg=0.133 kPa
下载: 导出CSV
-
下载:
图(5) / 表(2)
计量
- 文章访问数: 1
- HTML全文浏览量: 0
- PDF下载量: 0
- 被引次数: 0
