Research on risk prediction model of spontaneous premature delivery in pregnant women with diabetes mellitus based on machine learning
-
摘要:
目的 基于4种机器学习算法构建妊娠期糖尿病(GDM)孕妇自发性早产(SPB)风险预测模型，筛选最优模型，为识别GDM孕妇发生SPB的高危群体提供参考。 方法 收集2023年1月—2025年1月在温州市人民医院分娩的221例GDM孕妇的病历资料，根据是否早产分为早产组(68例)和正常组(153例)，采用多因素logistic分析研究GDM孕妇发生SPB的危险因素；将221例患者以7∶3比例随机拆分为训练集和验证集，分别采用决策树(DT)、K近邻(KNN)、随机森林(RF)、支持向量机(SVM)4种机器学习算法构建GDM孕妇SPB风险预测模型，并评估、筛选最优模型。 结果 多因素logistic分析显示，年龄≥35岁、BMI≥24、妊娠期高血压、糖尿病家族史、妊娠期阴道感染、孕晚期白细胞计数升高是GDM孕妇SPB的独立危险因素，血糖控制良好则是GDM孕妇SPB的保护因素。RF模型的AUC为0.925、准确率为0.863、精确率为0.887、灵敏度为0.729、特异度为0.975、F1分数为0.826，均优于其他3种模型。 结论 基于DT、KNN、RF、SVM 4种机器学习算法构建GDM孕妇SPB风险预测模型，RF预测效果最佳，其可有效筛查GDM孕妇发生SPB的高危人群，为临床干预方案制定提供支持。 Abstract:Objective To construct the risk prediction model of spontaneous preterm birth (SPB) in pregnant women with diabetes mellitus (GDM) based on the four machine learning algorithms, and to screen the optimal model to provide a reference for identifying the SPB population of GDM pregnant women. Methods The case data of 221 pregnant women with GDM who gave birth in the Obstetrics and Gynecology Department of Wenzhou People' s Hospital from January 2023 to January 2025 were collected. They were divided into a premature birth group (68 cases) and a normal group (153 cases) based on whether they were premature. Multivariate logistic analysis was used to identify the risk factors for SPB in GDM pregnant women. The patients were randomly divided into a 70% training set (155 cases) and a 30% validation set (66 cases). Four machine learning algorithms, namely decision tree (DT), K-nearest neighbor (KNN), random forest (RF), and support vector machine (SVM), were used to construct a GDM pregnant woman SPB risk prediction model, and the optimal model was evaluated and screened. Results Multivariate logistic analysis showed that age≤35 years old, BMI≥24, Hypertension during pregnancy, family history of diabetes, guided infection during pregnancy, and elevated WBC count in the third trimester were independent risk factors for SPB in GDM pregnant women, while good blood glucose control was a protective factor for SPB in GDM pregnant women. The AUC value of the RF model was 0.925, the accuracy was 0.863, the precision was 0.887, the sensitivity was 0.729, the specificity was 0.975, and the F1 score was 0.826, all of which are superior to the other three models. Conclusion A GDM pregnant women' s SPB risk prediction model is constructed based on four machine learning algorithms: decision tree (DT), K-nearest neighbor (KNN), random forest (RF), and support vector machine (SVM). RF prediction performs the best and can effectively screen high-risk populations for SPB in GDM pregnant women, providing support for clinical intervention plan formulation. -
Key words:
- Gestational diabetes /
- Machine learning /
- Spontaneous premature birth /
- Risk profile
-
表 1 2组GDM孕妇一般资料比较
Table 1. Comparison of general data of GDM pregnant women in the two groups
项目 早产组(68例) 正常组(153例) 统计量 P值 年龄[例(%)] 8.765a 0.003 ＜35岁 29(42.65) 98(64.05) ≥35岁 39(57.35) 55(35.95) BMI[例(%)] 4.123a 0.042 ＜24 37(54.41) 105(68.63) ≥24 31(45.59) 48(31.37) 产妇类型[例(%)] 0.243a 0.622 初产妇 30(44.12) 73(47.71) 经产妇 38(55.88) 80(52.29) 受教育程度[例(%)] 0.173a 0.678 高中及以下 19(27.94) 47(30.72) 大学及以上 49(72.06) 106(69.28) 妊娠期高血压[例(%)] 8.595a 0.003 是 26(38.24) 30(19.61) 否 42(61.76) 123(80.39) 糖尿病家族史[例(%)] 4.522a 0.033 是 27(39.71) 39(25.49) 否 41(60.29) 114(74.51) 妊娠期阴道感染[例(%)] 6.525a 0.011 是 36(52.94) 53(34.64) 否 32(47.06) 100(65.36) 合并子痫前期[例(%)] 0.360a 0.548 是 9(13.24) 16(10.46) 否 59(86.76) 137(89.54) 孕次[例(%)] 0.166a 0.683 ＜2次 30(44.12) 63(41.18) ≥2次 38(55.88) 90(58.82) 自然流产史[例(%)] 8.665a 0.003 是 33(48.53) 43(28.10) 否 35(51.47) 110(71.90) 血糖控制[例(%)] 32.924a ＜0.001 良好 23(33.82) 114(74.51) 较差 45(66.18) 39(25.49) HbA1c(x±s，%) 6.85±1.04 5.43±1.21 8.067b ＜0.001 FBG(x±s，mmol/L) 6.60±1.25 5.71±0.98 5.708b ＜0.001 孕晚期白细胞计数(x±s，109/L) 10.18±1.33 8.95±1.16 6.949b ＜0.001 注：a为χ2值，b为t值。 
下载: 导出CSV
表 2 变量赋值情况
Table 2. Variable assignment
变量 赋值方法 年龄 ＜35岁=0，≥35岁=1 BMI ＜24=0，≥24=1 妊娠期高血压 否=0，是=1 糖尿病家族史 否=0，是=1 妊娠期阴道感染 否=0，是=1 孕晚期白细胞计数 以实际值赋值 血糖控制 良好=0，较差=1
下载: 导出CSV
表 3 GDM孕妇SPB发生的多因素logistic分析
Table 3. Multivariate logistic analysis of SPB occurrence in GDM pregnant women
变量 B SE Waldχ2 P值 OR值 95% CI 年龄≥35岁 0.917 0.476 8.261 0.021 2.502 1.754~3.306 BMI≥24 0.743 0.285 5.237 0.032 2.102 1.262~2.793 妊娠期高血压 1.108 0.693 8.732 0.010 3.028 1.64~4.450 糖尿病家族史 0.847 0.405 5.903 0.013 2.333 1.353~3.407 妊娠期阴道感染 0.925 0.342 10.254 ＜0.001 2.522 1.680~3.411 孕晚期白细胞计数升高 1.521 0.562 4.885 0.016 4.577 2.391~6.594 血糖控制良好 -0.427 0.926 9.007 0.003 0.652 0.442~0.858 注：本表仅列出差异有统计学意义的结果。
下载: 导出CSV
表 4 4种机器学习模型在训练集上的预测性能
Table 4. Prediction performance of four machine learning models on the training set
预测模型 AUC 准确率 精确率 灵敏度 特异度 F1分数 DT 0.782 0.813 0.695 0.558 0.837 0.516 KNN 0.836 0.784 0.643 0.672 0.910 0.745 RF 0.925 0.863 0.887 0.729 0.975 0.826 SVM 0.914 0.835 0.728 0.492 0.896 0.677
下载: 导出CSV
表 5 4种机器学习模型在验证集上的预测性能
Table 5. Prediction performance of four machine learning models on the validation set
预测模型 AUC 准确率 精确率 灵敏度 特异度 F1分数 DT 0.775 0.802 0.711 0.563 0.828 0.522 KNN 0.840 0.749 0.648 0.682 0.893 0.760 RF 0.931 0.880 0.893 0.751 0.946 0.835 SVM 0.897 0.826 0.735 0.524 0.874 0.692
下载: 导出CSV
-
[1] 王闪, 王永强, 张永梅. FIB、AT-Ⅲ和HbA1c水平与妊娠期糖尿病孕妇产后出血的相关性研究[J]. 热带医学杂志, 2025, 25(5): 679-681, 700.WANG S, WANG Y Q, ZHANG Y M. The correlation between FIB, AT-Ⅲ and HbA1c levels and postpartum hemorrhage in pregnant women with gestational diabetes mellitus[J]. Journal of Tropical Medicine, 2025, 25(5): 679-681, 700. [2] 裴燕, 尹燕娜, 门鸿芹. 糖化血清蛋白和糖化血红蛋白及其交互作用与妊娠期糖尿病孕妇胰岛素抵抗的关系[J]. 中国临床研究, 2025, 38(9): 1425-1429.PEI Y, YIN Y N, MEN H Q. Glycosylated serum protein, glycosylated hemoglobin and their interaction on insulin resistance in pregnant women with gestational diabetes mellitus[J]. Chinese Journal of Clinical Research, 2025, 38(9): 1425-1429. [3] 邓聪. 基于群组化保健模式在孕中期高龄妊娠期糖尿病孕妇中的应用及对妊娠结局的影响[J]. 实用妇科内分泌电子杂志, 2024, 11(21): 117-120.DENG C. Application of the group-based health care model in mid-pregnancy high-risk gestational diabetes mellitus pregnant women and its impact on pregnancy outcomes[J]. Electronic Journal of Practical Gynecological End-ocrinology, 2024, 11(21): 117-120. [4] 邹世姣, 曾爱萍. 妊娠期糖尿病孕妇不良妊娠结局的影响因素观察[J]. 大医生, 2025, 10(1): 25-27.ZOU S J, ZENG A P. 妊娠期糖尿病孕妇不良妊娠结局的影响因素观察[J]. Doctor, 2025, 10(1): 25-27. [5] 王丽云, 高天勤, 刘雨佳, 等. 基于机器学习产后压力性尿失禁风险预测模型的构建及验证[J]. 山东大学学报(医学版), 2025, 63(6): 55-66.WANG L Y, GAO T Q, LIU Y J, et al. Development and validation of a postpartum stress urinary incontinence risk prediction model based on machine learning[J]. Journal of Shandong University (Health Science), 2025, 63(6): 55-66. [6] 任夏, 刘罗杰, 查俊杰, 等. 基于机器学习算法构建重症监护病房急性胰腺炎并发急性呼吸窘迫综合征的风险预测模型[J]. 中国临床研究, 2025, 38(8): 1173-1181.REN X, LIU L J, ZHA J J, et al. Development of a prediction model for acute respiratory distress syndrome in ICU patients with acute pancreatitis based on machine learning algorithms[J]. Chinese Journal of Clinical Research, 2025, 38(8): 1173-1181. [7] 余善招, 包益萍, 赵灵, 等. 基于机器学习算法的成人神经外科重症患者肠内营养喂养不耐受风险预测模型的构建[J]. 中华全科医学, 2025, 23(3): 414-416, 452. doi: 10.16766/j.cnki.issn.1674-4152.003916YU S Z, BAO Y P, ZHAO L, et al. Construction of risk prediction model of enteral feeding intolerance in severe neurosurgical patients based on machine learning algorithm[J]. Chinese Journal of General Practice, 2025, 23(3): 414-416, 452. doi: 10.16766/j.cnki.issn.1674-4152.003916 [8] 刘吉莉, 王凤美, 刘阳, 等. 基于机器学习方法构建幽门螺杆菌感染的风险预测模型[J]. 药学前沿, 2025, 29(2): 265-276.LIU J L, WANG F M, LIU Y, et al. Construction of the risk prediction model of Helicobacter pylori infection based on Machine learning method[J]. Frontiers in Pharmaceutical Sciences, 2025, 29(2): 265-276. [9] 楼佳烨, 王艳梅, 潘欣欣, 等. 基于机器学习的糖尿病足发病风险预测模型构建[J]. 护理学杂志, 2025, 40(9): 26-30.LOU J Y, WANG Y M, PAN X X, et al. Construction of machine learning-based prediction models for diabetic foot risk in diabetes patients[J]. Journal of Nursing Science, 2025, 40(9): 26-30. [10] 林苗, 周平. 妊娠期糖尿病孕妇产后高血糖的危险因素分析[J]. 中华全科医学, 2022, 20(8): 1350-1352, 1356. doi: 10.16766/j.cnki.issn.1674-4152.002594LIN M, ZHOU P. Risk factors of postpartum hyperglycaemia in pregnant women with gestational diabetes mellitus[J]. Chinese Journal of General Practice, 2022, 20(8): 1350-1352, 1356. doi: 10.16766/j.cnki.issn.1674-4152.002594 [11] 陈芳芳, 张宜生, 李伟, 等. 基于机器学习的妊娠期糖尿病孕妇不良妊娠结局风险预测研究[J]. 医院管理论坛, 2023, 40(12): 69-72, 90.CHEN F F, ZHANG Y S, LI W, et al. Risk Prediction of Adverse Pregnancy Outcomes in Pregnant Women with Gestational Diabetes Mellitus Based on Machine Learning[J]. Hospital Management Forum, 2023, 40(12): 69-72, 90. [12] 邱青梅, 梁莉, 陆洁清, 等. 预测模型分析妊娠期糖尿病产妇不良妊娠结局影响因素的应用价值[J]. 安徽医学, 2025, 46(5): 582-588.QIU Q M, LIANG L, LU J Q, et al. Application value of the prediction model in analyzing the influencing factors of adverse pregnancy outcomes in puerperae with gestational diabetes mellitus[J]. Anhui Medical Journal, 2025, 46(5): 582-588. [13] 曾文玉, 王泯蓉, 邓夏, 等. 妊娠糖尿病孕妇早产风险的预测模型构建与验证[J]. 中国优生与遗传杂志, 2024, 32(8): 1626-1630.ZENG W Y, WANG M R, DENG X, et al. Construction and validation of prediction model for premature delivery risk of pregnant women with gestational diabetes mellitus[J]. Chinese Journal of Birth Health & Heredity, 2024, 32(8): 1626-1630. [14] 杨元元, 杨盼, 万瑞. 孕前体重指数及孕期血糖水平对妊娠期糖尿病孕妇围产期结局的影响[J]. 感染、炎症、修复, 2024, 25(32): 218-221.YANG Y Y, YANG P, WAN R. Effects of pre-pregnancy body mass index and blood glucose level on perinatal outcomes in pregnant women with gestational diabetes mellitus[J]. Infection Inflammation Repair, 2024, 25(32): 218-221. [15] 王娇连, 方彩玲, 陈梦蝶, 等. 妊娠期糖尿病孕妇自发性早产的二元logistic回归方程构建及预防策略[J]. 实用预防医学, 2023, 30(11): 1377-1380.WANG J L, FANG C L, CHEN M D, et al. Construction of binary logistic regression equation of spontaneous preterm delivery of pregnant women with gestational diabetes and prevention strategies[J]. Practical Preventive Medicine, 2023, 30(11): 1377-1380. [16] 曾欢, 李红霞, 杨永芹, 等. 妊娠期糖尿病患者血清微小RNA-875-5p与血清硫氧还蛋白还原酶1对母婴结局的预测价值[J]. 临床内科杂志, 2024, 41(10): 693-697.ZENG H, LI H X, YANG Y Q, et al. Predictive value of serum microRNA-875-5 p and serum thioredoxin reductase 1 in gestational diabetes mellitus for maternal and infant outcomes[J]. Journal of Clinical Internal Medicine, 2024, 41(10): 693-697. [17] WANG Y, LI Y C. Analysis of clinical differences between preterm delivery and full-term delivery among pregnant women with gestational diabetes mellitus[J]. JCNR, 2022, 6(6): 136-141. doi: 10.26689/jcnr.v6i6.4554 [18] 赵雅斐, 付文君. 妊娠期糖尿病发生不良妊娠结局的危险因素分析[J]. 医药论坛杂志, 2023, 44(18): 15-18.ZHAO Y F, FU W J. Analysis of risk factors for adverse pregnancy outcomes associated with gestational diabetes mellitus[J]. Journal of Medical Forum, 2023, 44(18): 15-18. [19] 夏雪梅, 周梦林, 陈丹青. 妊娠期糖尿病孕妇自发性早产的高危因素分析[J]. 实用妇产科杂志, 2022, 38(8): 621-624.XIA X M, ZHOU M L, CHEN D Q. Analysis of High Risk Factors of Spontaneous Preterm Birth of Patients with Gestational Diabetes mellitus[J]. Journal of Practical Obstetrics and Gynecology, 2022, 38(8): 621-624. [20] 胡敏, 韦莉霞, 陈江鸿. 宫颈机能不全孕妇行经阴道宫颈环扎术后早产或流产的危险因素[J]. 广西医学, 2022, 44(12): 1341-1345.HU M, WEI L X, CHEN J H. Risk factors for premature birth or miscarriage in pregnant women with cervical incompetence undergoing transvaginal cervical cerclage surgery[J]. Guangxi Medical Journal, 2022, 44(12): 1341-1345. -
点击查看大图
表(5)
计量
- 文章访问数: 10
- HTML全文浏览量: 4
- PDF下载量: 0
- 被引次数: 0
