基于自注意力机制预测ICU脓毒症患者的死亡率

doi:10.6040/j.issn.1671-7554.0.2025.0117

摘要/Abstract

摘要： 目的基于自注意力机制模型预测重症加强护理病房(Intensive Care Unit, ICU)脓毒症患者死亡率。方法在MIMIC-IV数据库中选取符合Sepsis-3标准的脓毒症患者,使用多重logistic回归分析种族对脓毒症患者死亡率的影响,通过构建纳入或未纳入种族特征的预测模型,并且比较其性能差异,进一步评估是否将种族纳入预测模型。将数据集按1∶1比例分为训练集和验证集。使用二分类交叉熵损失函数和Adam优化器在训练集上进行1 000次迭代训练,并在验证集上评估模型性能。性能指标包括受试者工作特征(receiver operating characteristic, ROC)曲线下面积(area under the curve, AUC)和准确率。结果共纳入16 521例脓毒症患者。多重logistic回归分析结果显示,种族与ICU脓毒症患者死亡率无显著关联。训练集中构建的纳入或未纳入种族特征的模型在验证集中AUC均为0.82,准确率均为0.88,优于传统评分系统(如OASIS AUC: 0.70;LODS AUC: 0.74;SAPSII AUC: 0.75)。结论种族特征对于脓毒症患者死亡率的预测无明显影响,基于自注意力机制构建的预测模型显著提高了对ICU脓毒症患者死亡率的预测性能,表现优于传统评分系统。

关键词: 重症加强护理病房, 脓毒症, 预测, 死亡率, 自注意力机制

Abstract: Objective To predict sepsis mortality in Intensive Care Unit(ICU)using a self-attention mechanism model. Methods Sepsis patients who meet the Sepsis-3 criteria were selected from the MIMIC-IV database. Multiple logistic regression analysis was performed to assess the impact of ethnicity on sepsis mortality. To further validate this relationship, parallel predictive models(with and without ethnicity)were constructed and their performance metrics were compared. The dataset was split into training and validation sets at a 1∶1 ratio. A binary cross-entropy loss function and the Adam optimizer were used to train the model for 1,000 iterations on the training set, and the models performance was evaluated on the validation set. Performance metrics included the area under the curve(AUC)of receiver operating characteristic(ROC)and accuracy. Results A total of 16,521 sepsis patients were included. The multiple logistic regression analysis showed no significant relationship between ethnicity and mortality, so ethnicity was not included in the subsequent model. Both models(with/without ethnicity)achieved identical AUC(0.82)and accuracy(0.88)on the validation set, outperforming traditional scoring systems(e.g., OASIS AUC: 0.70; LODS AUC: 0.74; SAPSII AUC: 0.75). Conclusion Ethnicity has no significant effect on the prediction of mortality in patients with sepsis. The prediction model built on the self-attention mechanism significantly improves the prediction performance of ICU sepsis mortality, outperforming traditional scoring systems.

Key words: Intensive Care Unit, Sepsis, Prediction, Mortality, Self-attention mechanism

中图分类号:

R631+.3

李晓琪,刘佩丽,成红,赵艳艳. 基于自注意力机制预测ICU脓毒症患者的死亡率[J]. 山东大学学报 (医学版), 2025, 63(8): 86-93.

LI Xiaoqi, LIU Peili, CHENG Hong, ZHAO Yanyan. Predicting ICU sepsis mortality using self-attention mechanism[J]. Journal of Shandong University (Health Sciences), 2025, 63(8): 86-93.

参考文献

[1] Singer M, Deutschman CS, Seymour CW, et al. The third international consensus definitions for sepsis and septic shock(sepsis-3)[J]. JAMA, 2016, 315(8): 801-810.
[2] Rudd KE, Johnson SC, Agesa KM, et al. Global, regional, and national sepsis incidence and mortality, 1990-2017: analysis for the Global Burden of Disease Study[J]. Lancet, 2020, 395(10219): 200-211.
[3] Kong G, Lin K, Hu Y. Using machine learning methods to predict in-hospital mortality of sepsis patients in the ICU[J]. BMC Med Inform Decis Mak, 2020, 20: 251. doi:10.1186/s12911-020-01271-2
[4] Gao J, Lu Y, Ashrafi N, et al. Prediction of sepsis mortality in ICU patients using machine learning methods[J]. BMC Med Inform Decis Mak, 2024, 24: 228. doi:10.1186/s12911-024-02630-z
[5] Khojandi A, Tansakul V, Li X, et al. Prediction of sepsis and in-hospital mortality using electronic health records[J]. Methods Inf Med, 2018, 57(4): 185-193.
[6] Bao C, Deng F, Zhao S. Machine-learning models for prediction of sepsis patients mortality[J]. Med Intensiva(Engl Ed), 2023, 47(6): 315-325.
[7] Zhang Y, Xu W, Yang P, et al. Machine learning for the prediction of sepsis-related death: a systematic review and Meta-analysis[J]. BMC Med Inform Decis Mak, 2023, 23(1): 283. doi:10.1186/s12911-023-02383-1
[8] 詹贤春, 程恒亮, 李维华. 基于注意力的融合模型预测脓毒症患者死亡率[J]. 云南大学学报(自然科学版), 2024, 46(5): 829-837. ZHAN Xianchun, CHENG Hengliang, LI Weihua. Attention-based fusion model to predict mortality of sepsis patients[J]. Journal of Yunnan University: Natural Sciences Edition, 2024, 46(5): 829-837.
[9] 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): 462. doi:10.1186/s12967-020-02620-5
[10] Nikravangolsefid N, Reddy S, Truong HH, et al. Machine learning for predicting mortality in adult critically ill patients with sepsis: a systematic review[J]. J Crit Care, 2024, 84: 154889. doi:10.1016/j.jcrc.2024.154889
[11] Sarraf E, Sadr AV, Abedi V, et al. Enhancing sepsis prognosis: integrating social determinants and demographic variables into a comprehensive model for critically ill patients[J]. J Crit Care, 2024, 83: 154857. doi:10.1016/j.jcrc.2024.154857
[12] Black LP, Hopson C, Puskarich MA, et al. Racial disparities in septic shock mortality: a retrospective cohort study[J]. Lancet Reg Health Am, 2023, 29. doi:10.1016/j.lana.2023.100646
[13] Galiatsatos P, Sun J, Welsh J, et al. Health disparities and sepsis: a systematic review and Meta-analysis on the influence of race on sepsis-related mortality[J]. J Racial Ethn Health Disparities, 2019, 6(5): 900-908.
[14] Erickson SE, Vasilevskis EE, Kuzniewicz MW, et al. The effect of race and ethnicity on outcomes among patients in the intensive care unit: a comprehensive study involving socioeconomic status and resuscitation preferences[J]. Crit Care Med, 2011, 39(3): 429-435.
[15] Jones JM, Fingar KR, Miller MA, et al. Racial disparities in sepsis-related in-hospital mortality: using a broad case capture method and multivariate controls for clinical and hospital variables, 2004-2013[J]. Crit Care Med, 2017, 45(12): e1209-e1217.
[16] Chaudhary NS, Donnelly JP, Wang HE. Racial differences in sepsis mortality at U.S. academic medical center-affiliated hospitals[J]. Crit Care Med, 2018, 46(6): 878-883.
[17] Sandoval E, Chang DW. Association between race and case fatality rate in hospitalizations for sepsis[J]. J Racial Ethn Health Disparities, 2016, 3(4): 625-634.
[18] Zhang Y, Liu C, Liu M, et al. Attention is all you need: utilizing attention in AI-enabled drug discovery[J]. Brief Bioinform, 2024, 25(1): bbad467. doi:10.1093/bib/bbad467
[19] Johnson A, Bulgarelli L, Pollard T, et al. MIMIC-IV(version 2.2)[EB/OL].(2023-01-06)[2025-01-26]. https://doi.org/10.13026/6mm1-ek67
[20] Kalimouttou A, Lerner I, Cheurfa C, et al. Machine-learning-derived sepsis bundle of care[J]. Intens Care Med, 2023, 49(1): 26-36.
[21] Sarkar R, Martin C, Mattie H, et al. Performance of intensive care unit severity scoring systems across diffe-rent ethnicities in the USA: a retrospective observational study[J]. Lancet Digit Health, 2021, 3(4): 241-249.
[22] Angus DC, Van Der Poll T. Severe sepsis and septic shock[J]. N Engl J Med, 2013, 369(9): 840-851.
[23] Vincent JL, Sakr Y, Sprung CL, et al. Sepsis in European intensive care units: results of the SOAP study[J]. Crit Care Med, 2006, 34(2): 344-353.
[24] Azkárate I, Choperena G, Salas E, et al. Epidemiology and prognostic factors in severe sepsis/septic shock. Evolution over six years[J]. Med Intensiva, 2016, 40(1): 18-25.
[25] van Vught LA, Klouwenberg PMCK, Spitoni C, et al. Incidence, risk factors, and attributable mortality of secondary infections in the Intensive Care Unit after admission for sepsis[J]. JAMA, 2016, 315(14): 1469-1479.
[26] Vincent JL, Moreno R, Takala J, et al. The SOFA(Sepsis-related Organ Failure Assessment)score to describe organ dysfunction/failure. On behalf of the Working Group on Sepsis-Related Problems of the European Society of Intensive Care Medicine[J]. Intensive Care Med, 1996, 22(7): 707-710.
[27] Le Gall JR, Lemeshow S, Saulnier F. A new Simplified Acute Physiology Score(SAPS II)based on a Euro-pean/North American multicenter study[J]. JAMA, 1993, 270(24): 2957-2963.
[28] Evans L, Rhodes A, Alhazzani W, et al. Surviving sepsis campaign: international guidelines for management of sepsis and septic shock 2021[J]. Crit Care Med, 2021, 49(11): 1063-1143.
[29] Taylor SP, Karvetski CH, Templin MA, et al. Hospital differences drive antibiotic delays for black patients compared with white patients with suspected septic shock[J]. Crit Care Med, 2018, 46(2): 126-131.
[30] Rosenblatt F. The perceptron: a probabilistic model for information storage and organization in the brain[J]. Psychol Rev, 1958, 65(6): 386-408.
[31] Mao A, Mohri M, Zhong Y. Cross-entropy loss functions: theoretical analysis and applications. International conference on Machine learning[EB/OL].(2023-06-20)[2025-01-26]. https://arxiv.org/abs/2304.07288
[32] Kingma DP, Ba J. Adam: a method for stochastic optimization[EB/OL].(2017-01-30)[2025-01-26]. https://doi.org/10.48550/arXiv.1412.6980
[33] Koköfer A, Mamandipoor B, Flamm M, et al. The impact of ethnic background on ICU care and outcome in sepsis and septic shock-a retrospective multicenter analysis on 17,949 patients[J]. BMC Infect Dis, 2023, 23(1): 194. doi:10.1186/s12879-023-08170-7
[34] Prest J, Sathananthan M, Jeganathan N. Current trends in sepsis-related mortality in the United States[J]. Crit Care Med, 2021, 49(8): 1276-1284.
[35] McGowan SK, Sarigiannis KA, Fox SC, et al. Racial disparities in ICU outcomes: a systematic review[J]. Crit Care Med, 2022, 50(1): 1-20.
[36] Limaye NP, Matias WR, Rozansky H, et al. Limited English proficiency and sepsis mortality by race and ethnicity[J]. JAMA Netw Open, 2024, 7(1): e2350373. doi:10.1001/jamanetworkopen.2023.50373
[37] Zimmerman JE, Kramer AA, McNair DS, et al. Acute Physiology and Chronic Health Evaluation(APACHE)IV: hospital mortality assessment for todays critically ill patients[J]. Crit Care Med, 2006, 34(5): 1297-1310.

多维度评价

Viewed

Full text

Abstract

Cited

Shared

Discussed