通讯作者: 吴夏阳, 56425477@qq.com

DOI：10.12201/bmr.202503.00066

Mortality risk assessment and interpretability analysis of preterm infants in the NICU using machine learning models

• 摘要：目的 旨在利用机器学习算法预测早产儿ICU死亡风险,为临床医生提供早期诊断和风险评估的辅助决策工具。方法 回顾性地收集PIC数据库中早产儿病例的临床数据。按照ICU预后情况分为死亡组和生存组。基于LASSO回归分析和多因素Logistic回归分析的结果,筛选出可能影响早产儿预后的关键临床特征。研究通过SMOTE算法平衡数据,结合7种机器学习模型(如LightGBM、随机森林等),构建预测模型并评估其性能。使用 Shapley Additive Explanations (SHAP)算法进行模型解释。 结果 最终纳入患儿923人。生存组886人,死亡组37人,共收集38个临床特征。LASSO筛选出8个与早产儿ICU死亡密切相关的变量包括乳酸、氯离子浓度、中性粒细胞、红细胞分布宽度等。多因素Logistic回归分析显示：乳酸、呼吸频率是早产儿ICU预后的独立影响因素。LightGBM模型的AUC达到0.972,在准确性、精确性等指标上均优于其他模型同时通过SHAP分析提高了模型的解释性。研究结果显示,呼吸频率和乳酸对早产儿死亡风险的预测贡献最大。 结论 本研究为早产儿预后的早期识别和干预提供了可靠工具,强调了关键生理指标的重要性。未来需要多中心数据验证以增强模型的普适性,并进一步优化算法性能。

  关键词： 早产儿、ICU死亡风险、机器学习、LightGBM模型、风险预测

   

  Abstract: Objective: Aimed at using machine learning algorithms to predict the risk of neonatal ICU mortality, providing clinicians with an early diagnosis and risk assessment tool to assist in decision-making. Methods: Retrospectively collecting clinical data of preterm infants from the PIC database. Cases were divided into mortality and survival groups based on ICU outcomes. Key clinical characteristics potentially affecting preterm infant outcomes were screened using LASSO regression analysis and multivariate logistic regression analysis. The study balanced the data using the SMOTE algorithm and constructed predictive models using seven machine learning models (e.g., LightGBM, random forest), evaluating their performance. Model interpretation was performed using the Shapley Additive Explanations (SHAP) algorithm. Results: A total of 923 infants were included in the final analysis. The survival group comprised 886 infants, and the death group comprised 37 infants. A total of 38 clinical characteristics were collected. LASSO screening identified 8 variables significantly associated with neonatal ICU mortality, including lactate, chloride concentration, neutrophils, and red blood cell distribution width. Multivariate logistic regression analysis revealed that lactate and respiratory rate were independent predictors of neonatal ICU outcomes. The LightGBM model achieved an AUC of 0.972 and outperformed other models in terms of accuracy and precision. Furthermore, SHAP analysis enhanced model interpretability. The results indicated that respiratory rate and lactate contributed most significantly to the prediction of infant mortality risk. Conclusion: This study provides reliable tools for early identification and intervention of preterm infant outcomes, emphasizing the importance of key physiological indicators. Future multi-center data validation is needed to enhance the models generalizability and further optimize algorithm performance.

  Key words: Premature infants, ICU mortality risk, machine learning, LightGBM model, risk prediction

  提交时间：2025-03-21

  版权声明：作者本人独立拥有该论文的版权，预印本系统仅拥有论文的永久保存权利。任何人未经允许不得重复使用。
• html

• 图表

• 何永虹, 张伟, 覃大卫, 田文军, 陈玲, 严蜜, 谷秀, 符和建, 田昌军. 早产儿羊水胎粪污染的危险因素及临床结局. 2025. doi: 10.12201/bmr.202502.00007

  张丽亚, 陈黎丽. ω－3PUFA对早产儿脑损伤免疫功能调节机制研究. 2024. doi: 10.12201/bmr.202408.00011

  胡欣. 剖宫产术中寒战风险预测模型及列线图的构建. 2025. doi: 10.12201/bmr.202501.00053

  王一凡, 石超君, 马安宁. Ⅱ型糖尿病并发动脉粥样硬化风险预测模型比较. 2024. doi: 10.12201/bmr.202404.00007

  何源, 杨树杰, 曾英, 易海英, 黄杰, 甘斌, 魏贤, 唐艳林. 微创给药在极早产儿呼吸窘迫综合征的疗效观察. 2024. doi: 10.12201/bmr.202410.00018

  周妮, 谷强. 儿童大叶性肺炎病程的危险因素及风险预测. 2025. doi: 10.12201/bmr.202504.00059

  陈冰融, 舒文秀, 罗柳飞, 童嘉琦, 乐静. 骨髓瘤患者发生3-4级骨髓瘤骨病的风险预测模型构建. 2024. doi: 10.12201/bmr.202410.00057

  韦云师, 莫伟, 向娅, 廖秋姣, 何柳, 凌朝灵, 陆启祥, 刘芳印. 老年髋部骨折患者术后谵妄风险预测模型的研究进展. 2024. doi: 10.12201/bmr.202409.00029

  兰雨姗, 郑思, 李姣. 机器学习方法在因果推断中混杂因素控制的应用. 2022. doi: 10.12201/bmr.202203.00015

  冯利, 岳小飞. 9种结合式机器学习算法在基于高维数据的肿瘤早期诊断准确性的比较研究. 2021. doi: 10.12201/bmr.202108.00016

• 序号	提交日期	编号	操作
  1	2025-02-19	bmr.202503.00066V1	下载

• 公开评论  匿名评论  仅发给作者 提交评论