基于可解释性机器学习构建脑卒中患者卒中相关性肺炎风险预测模型

李春标 ,  王婷 ,  袁娟 ,  袁琳丽 ,  刘艺 ,  李大进 ,  尤敏

中国卫生统计 ›› 2026, Vol. 43 ›› Issue (4) : 489 -496.

PDF (2835KB)
中国卫生统计 ›› 2026, Vol. 43 ›› Issue (4) : 489 -496. DOI: 10.11783/j.issn.1002-3674.2026.04.002
论著

基于可解释性机器学习构建脑卒中患者卒中相关性肺炎风险预测模型

作者信息 +

Construct a Risk Prediction Model for Stroke-associated Pneumonia in Stroke Patients based on Interpretable Machine Learning

Author information +
文章历史 +
PDF (2902K)

摘要

目的 基于可解释性机器学习方法构建卒中相关性肺炎(stroke-associated pneumonia,SAP)风险预测模型,为临床预防SAP提供参考。方法 回顾性收集2023年12月1日至2024年8月31日就诊于安徽中医药大学第二附属医院的763例脑卒中患者的资料,将其按照6∶4的比例随机分为模型训练集(n=457)和模型验证集(n=306)。首先对纳入的数据进行整理,随后使用LASSO回归筛选纳入的变量。使用决策树(decision tree,DT)、逻辑回归(logistic regression,LR)、极限梯度提升算法(extreme gradient boosting,XGBoost)、支持向量机(support vector machine,SVM)、轻量级梯度提升机算法(light gradient boosting machine,lightGBM)5种机器学习方法构建模型,通过网格搜索5折交叉验证进行超参数优化,优化模型的性能指标。模型的预测性能通过AUC、校准曲线和决策曲线分析(decision curve analysis,DCA)等指标进行评估,最后使用Shapley加性解释(Shapley additive explanation,SHAP)来解释最优模型的预测结果。结果 本研究中SAP的发生率为31.72%(242/763),通过LASSO回归筛选出6个变量,包括使用鼻胃管治疗、年龄、日常生活活动能力量表(activity of daily living scale,ADL)评分、白蛋白、超敏C反应蛋白和血红蛋白。在验证集中表现最佳的模型为XGBoost模型,AUC为0.926,精确度为0.914,F1分数为0.889,其校准曲线和DCA均展示出良好的性能,SHAP算法显示重要性排序排名第一的为ADL评分。结论 使用XGBoost构建的模型具有良好的预测性能和临床适用性,有望为临床决策提供支持,改善患者的预后。

Abstract

Objective The risk prediction model of stroke-associated pneumonia (SAP) was constructed based on interpretable machine learning method to provide reference for clinical prevention of SAP.Methods Data of 763 stroke patients admitted to the Second Affiliated Hospital of Anhui University of Chinese Medicine from December 1, 2023 to August 31, 2024 were retrospectively collected and randomly divided into model training set (n=457) and model validation set (n= 306) according to a ratio of 6∶4. Firstly, the included data were sorted out, and then the included variables were screened using LASSO regression. Five machine learning methods, including decision tree (DT), logistic regression (LR), extreme gradient boosting (XGBoost), support vector machine (SVM) and lightweight gradient boosting machine (lightGBM), were used to construct the model. Hyperparameter optimization was carried out through grid search and 5-fold cross-validation to optimize the performance index of the model. The predictive performance of the model was evaluated by AUC, calibration curve and decision curve analysis (DCA), and finally Shapley additive explanation (SHAP) was used to explain the prediction results of the optimal model. Results In this study, the incidence of SAP was 31. 72% (242/763). Six variables were screened by LASSO regression, including nasogastric tube therapy, age, activity of daily living scale (ADL) score, albumin, hypersensitive C-reactive protein, and haemoglobin. The model with the best performance in the verification set is XGBoost model, with AUC of 0. 926, accuracy of 0. 914 and F1 score of 0. 889. Its calibration curve and DCA both show good performance. The SHAP algorithm shows that the model with the highest ranking in importance ranking is the ADL score. Conclusion The model built using XGBoost has good predictive performance and clinical applicability, which is expected to support clinical decision making and improve patient outcomes.

关键词

脑卒中 / 卒中相关性肺炎 / 机器学习 / 预测模型 / Shapley加性解释

Key words

Stroke / Stroke-associated pneumonia / Machine learning / Prediction model / Shapley additive explanation

引用本文

引用格式 ▾
李春标,王婷,袁娟,袁琳丽,刘艺,李大进,尤敏. 基于可解释性机器学习构建脑卒中患者卒中相关性肺炎风险预测模型[J]. 中国卫生统计, 2026, 43(4): 489-496 DOI:10.11783/j.issn.1002-3674.2026.04.002

登录浏览全文

4963

注册一个新账户 忘记密码

参考文献

[1]

Feigin VL, Stark BA, Johnson CO, et al. Global, regional, and national burden of stroke and its risk factors, 1990—2019: a systematic analysis for the Global Burden of Disease Study 2019 [J].Lancet Neurol, 2021, 20(10): 795-820.

[2]

Smith CJ, Kishore AK, Vail A, et al. Diagnosis of Stroke—Associated Pneumonia: Recommendations from the Pneumonia in Stroke Consensus Group [J].Stroke, 2015, 46(8): 2335-2340.

[3]

Ji R, Wang D, Shen H, et al. Interrelationship among common medical complications after acute stroke: pneumonia plays an important role [J].Stroke, 2013, 44(12): 3436-3444.

[4]

Smith CJ, Bray BD, Hoffman A, et al. Can a Novel Clinical Risk Score Improve Pneumonia Prediction in Acute Stroke Care? A UK Multicenter Cohort Study [J].J Am Heart Assoc, 2015, 4(1): e001307.

[5]

Teh WH, Smith CJ, Barlas RS, et al. Impact of stroke—associated pneumonia on mortality, length of hospitalization, and functional outcome [J].Acta Neurol Scand, 2018, 138(4): 293-300.

[6]

Yu YJ, Weng WC, Su FC, et al. Association between pneumonia in acute stroke stage and 3—year mortality in patients with acute first—ever ischemic stroke [J].J Clin Neurosci, 2016, 33: 124-128.

[7]

Houck PM, Bratzler DW, Nsa W, et al. Timing of antibiotic administration and outcomes for Medicare patients hospitalized with community—acquired pneumonia [J].Arch Intern Med, 2004, 164(6): 637-644.

[8]

Vermeij JD, Westendorp WF, Dippel DW, et al. Antibiotic therapy for preventing infections in people with acute stroke [J].Cochrane Database Syst Rev, 2018, 1(1): CD008530.

[9]

Ranstam J, Cook JA, Collins GS. Clinical prediction models [J].Br J Surg, 2016, 103(13): 1886.

[10]

Kwon HM, Jeong SW, Lee SH, et al. The pneumonia score: A simple grading scale for prediction of pneumonia after acute stroke [J].Am J Infect Control, 2006, 34(2): 64-68.

[11]

Hoffmann S, Malzahn U, Harms H, et al. Development of a Clinical Score (A2DS2) to Predict Pneumonia in Acute Ischemic Stroke [J].Stroke, 2012, 43(10): 2617-2623.

[12]

Ji R, Shen H, Pan Y, et al. Novel Risk Score to Predict Pneumonia After Acute Ischemic Stroke [J].Stroke, 2013, 44(5): 1303-1309.

[13]

谷鸿秋, 周支瑞, 章仲恒, 等. 临床预测模型:基本概念、应用场景及研究思路[J].中国循证心血管医学杂志, 2018, 10(12): 1454-1456.

[14]

Collin C, Wade DT, Davies S, et al. The Barthel ADL Index: A reliability study [J].Int Disabil Stud, 1988, 10(2): 61-63.

[15]

Osawa A, Maeshima S, Tanahashi N. Water—Swallowing Test: Screening for Aspiration in Stroke Patients [J].Cerebrovasc Dis, 2013, 35(3): 276-281.

[16]

Teasdale G, Jennett B. Assessment of coma and impaired consciousness. A practical scale [J].Lancet Lond Engl, 1974, 2(7872): 81-84.

[17]

Horan TC, Andrus M, Dudeck MA. CDC/NHSN surveillance definition of health care—associated infection and criteria for specific types of infections in the acute care setting [J].Am J Infect Control, 2008, 36(5): 309-332.

[18]

Riley RD, Ensor J, Snell KIE, et al. Calculating the sample size required for developing a clinical prediction model [J].BMJ, 2020: m441.

[19]

曹燚, 曾茜, 赵信飞, 等. 卒中相关性肺炎风险预测模型的系统评价[J].中国循证医学杂志, 2023, 23(11): 1259-1268.

[20]

Maeshima S, Osawa A, Hayashi T, et al. Elderly Age, Bilateral Lesions, and Severe Neurological Deficit Are Correlated with Stroke—associated Pneumonia [J].J Stroke Cerebrovasc Dis, 2014, 23(3): 484-489.

[21]

Matsusaka K, Kawakami G, Kamekawa H, et al. Pneumonia risks in bedridden patients receiving oral care and their screening tool: Malnutrition and urinary tract infection—induced inflammation [J].Geriatr Gerontol Int, 2018, 18(5): 714-722.

[22]

Schwarz M, Coccetti A, Murdoch A, et al. The impact of aspiration pneumonia and nasogastric feeding on clinical outcomes in stroke patients: A retrospective cohort study [J].J Clin Nurs, 2018, 27(1—2): e235-e241.

[23]

Brogan E, Langdon C, Brookes K, et al. Respiratory infections in acute stroke: nasogastric tubes and immobility are stronger predictors than dysphagia [J].Dysphagia, 2014, 29(3): 340-345.

[24]

Wang Q, Liu Y, Han L, et al. Risk factors for acute stroke—associated pneumonia and prediction of neutrophil—to—lymphocyte ratios [J].Am J Emerg Med, 2021, 41: 55-59.

[25]

Kalra L, Hodsoll J, Irshad S, et al. Association between nasogastric tubes, pneumonia, and clinical outcomes in acute stroke patients [J].Neurology, 2016, 87(13): 1352-1359.

[26]

Patel UK, Kodumuri N, Dave M, et al. Stroke—Associated Pneumonia: A Retrospective Study of Risk Factors and Outcomes [J].The Neurologist, 2020, 25(3): 39-48.

[27]

Huang L, Zhang R, Ji J, et al. Hypersensitive C—reactive protein—albumin ratio is associated with stroke—associated pneumonia and early clinical outcomes in patients with acute ischemic stroke [J].Brain Behav, 2022, 12(7): e2675.

[28]

Banait T, Wanjari A, Danade V, et al. Role of High—Sensitivity C—reactive Protein (hs—CRP) in Non—communicable Diseases: A Review [J].Cureus, 2022, 14(10): e30225.

[29]

Wang L, Wu L, Lang Y, et al. Association between high—sensitivity C—reactive protein levels and clinical outcomes in acute ischemic stroke patients treated with endovascular therapy [J].Ann Transl Med, 2020, 8(21): 1379.

[30]

Wei CC, Zhang ST, Tan G, et al. Impact of anemia on in—hospital complications after ischemic stroke [J].Eur J Neurol, 2018, 25(5): 768-774.

[31]

Song X, He Y, Bai J, et al. A nomogram based on nutritional status and A2DS2 score for predicting stroke—associated pneumonia in acute ischemic stroke patients with type 2 diabetes mellitus: A retrospective study [J].Front Nutr, 2022, 9: 1009041.

[32]

Yang X, Wang L, Zheng L, et al. Serum Albumin as a Potential Predictor of Pneumonia after an Acute Ischemic Stroke [J].Curr Neurovasc Res, 2020, 17(4): 385-393.

[33]

Lin G, Hu M, Song J, et al. High Fibrinogen to Albumin Ratio: A Novel Marker for Risk of Stroke—Associated Pneumonia [J].Front Neurol, 2022, 12: 747118.

[34]

Chen L, Xu M, Huang Q, et al. Clinical significance of albumin to globulin ratio among patients with stroke—associated pneumonia [J].Front Nutr, 2022, 9: 970573.

[35]

卢娅欣, 黄月, 李康. XGBoost算法在二分类非平衡高维数据分析中的应用[J].中国卫生统计, 2021, 38(1): 21-24.

[36]

Zhang Y, Xiang T, Wang Y, et al. Explainable machine learning for predicting 30—day readmission in acute heart failure patients [J].iScience, 2024, 27(7): 110281.

[37]

Islam MM, Rahman MJ, Rabby MS, et al. Predicting the risk of diabetic retinopathy using explainable machine learning algorithms [J].Diabetes Metab Syndr, 2023, 17(12): 102919.

[38]

Hou N, Li M, He L, et al. Predicting 30—days mortality for MIMIC—III patients with sepsis—3: a machine learning approach using Xgboost [J].J Transl Med, 2020, 18(1): 1-14.

基金资助

安徽省高校科研项目(2023AH050784)

安徽省科研编制计划项目(2022AH050463)

安徽省卫生健康科研项目(AHWJ2024Aa30205)

AI Summary AI Mindmap
PDF (2835KB)

0

访问

0

被引

详细

导航
相关文章

AI思维导图

/

〈 〉