| 摘要: |
| 目的:探讨早产儿脑损伤的危险因素,并构建列线图预测模型。 方法:选取 2020 年 1 月至 2024 年 8 月哈励逊国际和平医院 新生儿科重症监护室收治的早产儿 360 例,按 7 ∶ 3 比例将早产儿随机分为训练集(n= 252)和验证集(n= 108),根据诊断标准将训 练集中早产儿分为脑损伤组(n= 63)和非脑损伤组(n= 189)。 同期纳入河北省盐山县人民医院的 116 例早产儿进行外部验证。 结 果:训练集和验证集一般资料比较差异均无统计学意义(P>0. 05)。 脑损伤组患儿的母亲妊娠期糖尿病、胎龄<28 周、有创通气、出 生后感染、新生儿呼吸窘迫综合征(NRDS)比例高于非脑损伤组,日均体质量增长、淋巴细胞水平低于非脑损伤组(P<0. 05)。 母 亲妊娠期糖尿病、胎龄<28 周、有创通气、日均体质量增长、NRDS、淋巴细胞是早产儿脑损伤发生的影响因素(P<0. 05)。 内部验 证:Hosmer-Lemeshow 检验和校准曲线结果显示,实际预测曲线和修正后预测曲线拟合较好;受试者特征(ROC)曲线的曲线下面积 (AUC)分别为 0. 939、0. 920;决策曲线分析显示,列线图预测模型的阈值概率范围为 0. 06~0. 82、0. 03~0. 98,该范围内对早产儿进 行临床干预的净获益最高。 外部验证:Hosmer-Lemeshow 检验和校准曲线结果显示,实际预测曲线和修正后预测曲线拟合较好; ROC 曲线的 AUC 为 0. 902;决策曲线分析结果显示,列线图预测模型的阈值概率范围为 0. 02~0. 98,该范围内对早产儿进行临床干 预的净获益最高。 结论:基于多因素 logistic 回归分析结果构建并验证了早产儿脑损伤发生风险的列线图预测模型,该模型预测效 能较高。 |
| 关键词: 早产儿 脑损伤 危险因素 列线图 预测模型 |
| DOI：10.13407/j.cnki.jpp.1672-108X.2026.03.012 |
|
| 基金项目: |
|
| Construction of Column Chart Prediction Model and Early Drug Prevention and Treatment Strategies for Brain Injury in Premature Infants |
| Liu Dan1 , Meng Li1 , Liu Shuyan1 , Xu Kun2 , Dong Shasha3 |
| (1. Harrison International Peace Hospital, Hebei Hengshui 053000, China; 2. Hebei Yanshan County People’ s Hospital, Hebei Cangzhou 061300, China; 3. Baoding Fifth Hospital, Hebei Baoding 071052, China) |
| Abstract: |
| Objective: To explore the risk factors of brain injury in premature infants and construct the column chart prediction model.Methods: From Jan. 2020 to Aug. 2024, totally 360 premature infants in the Neonatal Intensive Care Unit of Harrison International Peace Hospital were extracted to be stochastically assigned into the training set (n = 252) and validation set (n = 108) in a 7 ∶ 3 ratio. The premature infants in the training set were assigned into the brain injury group ( n = 63) and non-brain injury group ( n = 189) according to the diagnostic criteria. And 116 premature infants admitted into Hebei Yanshan County People’s Hospital during the same period were extracted for external validation. Results: There was no statistically significant difference in the general data between the training set and validation set (P>0. 05). The proportions of gestational diabetes mellitus of the mothers, gestational age <28 weeks, invasive ventilation, postnatal infection, neonatal respiratory distress syndrome (NRDS) in the brain injury group were higher than those in the non-brain injury group, and the daily body mass growth and lymphocytes were lower than those in the non-brain injury group (P< 0. 05). Gestational diabetes mellitus of the mothers, gestational age <28 weeks, invasive ventilation, daily body mass growth, NRDS, lymphocytes were the influencing factors of brain injury in premature infants (P<0. 05). Internal verification: the Hosmer-Lemeshow test and calibration curve results revealed that the actual prediction curve and the corrected prediction curve were in good agreement. The area under the curve (AUC) of receiver operating characteristic (ROC) curve was 0. 939 and 0. 920, respectively. Decision curve analysis showed that the threshold probability range of the column chart prediction model was from 0. 06 to 0. 82 and from 0. 03 to 0. 98, within which the net benefit of clinical intervention for premature infants was the highest. External verification: the Hosmer-Lemeshow test and calibration curve results revealed that the actual prediction curve and the corrected prediction curve were in good agreement, and the AUC of the ROC curve was 0. 902. Decision curve analysis showed that the threshold probability range of the column chart prediction model was from 0. 02 to 0. 98, within which the net benefit of clinical intervention for premature infants was the highest. Conclusion: Based on the results of multivariate logistic regression analysis, the column chart prediction model for the risk of brain injury in preterm infants is constructed and verified, and the model has high prediction efficiency. |
| Key words: premature infants brain injury risk factors column chart prediction model |
