Clinical Phenotype Identification and Validation of Patients with Sepsis in the Intensive Care Unit
-
摘要: 目的 探讨重症监护病房脓毒症患者临床表型识别与验证,促进精准医疗发展。方法 采用人工智能算法处理重症监护医学信息数据库Ⅳ( Medical Information Mart for Intensive Care IV,MIMIC-IV)中脓毒症患者的临床数据,包括人口学特征、ICU首日实验室指标及治疗措施等89个临床特征,首先应用无监督机器学习算法( K均值聚类和层次聚类)进行表型识别,再应用有监督机器学习算法(轻量级梯度增强机)进行表型预测,并结合机器学习可解释性算法SHAP识别重要特征;最后通过传统统计学方法,从各表型间临床特征差异与临床结局差异两个角度进行验证。结果 在22517例脓毒症患者中发现了3种临床特征及结局显著不同的表型。其中表型1患者的死亡风险最高( 28 d死亡率为46.4%),以肾功能异常和疾病严重性评分升高为主;表型3患者的死亡风险最低( 28 d死亡率为11.2%),神经功能评分最佳。通过可解释性机器学习,识别出阴离子间隙、血尿素氮、肌酐、格拉斯哥昏迷评分、凝血酶原时间和序贯性器官功能衰竭评分6个特征(首日最差值)在表型识别( AUC≥0.89)和表型预后预测( AUC≥0.74)方面表现良好。在出ICU后28 d、60 d、90 d及1年内,表型3患者的死亡风险均最低( HR<1)。结论 利用机器学习算法成功识别了3种具有不同临床特征及预后的脓毒症临床表型,并筛选出6个关键临床特征,预期将在脓毒症亚型分类及预后评估中发挥重要作用,有助于患者的个体化治疗。Abstract: Objective To identify sepsis phenotypes and facilitate precision medicine. Methods We applied unsupervised machine learning algorithms (K-means clustering and hierarchical clustering) to identify the phenotypes of sepsis patients in the Medical Information Mart for Intensive Care IV (MIMIC-IV) database, based on 89 clinical features including demographic characteristics, laboratory indicators and treatment measures on the first day in ICU. Then, supervised machine learning algorithms (lightweight gradient augmenter) were used for the prediction of the patient's phenotypes , and further combined with SHAP (Shapely Additive eXplanations) for the identification of important features ; Finally , traditional statistical methods were used to validate the differences in clinical characteristics and clinical outcomes between the phenotypes. Results We identified three phenotypes in 22,517 sepsis patients. The phenotype 1 patient population had the highest risk of death (28-day mortality of 46.4%), dominated by abnormal renal function and elevated disease severity scores, while the phenotype 3 patient population had the lowest risk of death (28-day mortality of 11.2%), and it had the best neurological function score. Using interpretable machine learning, we identified six features (all the worst value on the first day) including anion gap, blood urea nitrogen, creatinine, Glasgow Coma Scale score, prothrombin time, and Sequential Organ Failure Assessment score showed good performance in phenotypic identification (AUC≥0.89) and phenotypic prognostic prediction (AUC≥0.74). The mortality risk of phenotype 3 patients was the lowest at 28 days, 60 days, 90 days, and 1 year after ICU discharge (HR<1). Conclusion Using machine learning methods, we successfully identified 3 clinical phenotypes of sepsis patients with different clinical characteristics and prognosis and screened out six key clinical features, which is expected to play an important role in the phenotype classification and prognostic assessment of sepsis and is conducive to individualized treatment.
-
Keywords:
- sepsis /
- phenotypic recognition /
- machine learning /
- precision medicine
-
-
[1] SINGER M, DEUTSCHMAN C S, SEYMOUR C W, et al. The Third International Consensus Definitions for Sepsis and Septic Shock (Sepsis-3) [J]. Jama, 2016, 315(8): 801-10.
[2] RUDD K E, JOHNSON S C, AGESA K M, 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): 200-11.
[3] YUK S A, SANCHEZ-RODRIGUEZ D A , TSIFANSKY M D, et al. Recent advances in nanomedicine for sepsis treatment [J]. Ther Deliv, 2018, 9(6): 435-50.
[4] LIANG C, PAN S, WU W, et al. Glucocorticoid therapy for sepsis in the AI era: a survey on current and future approaches [J]. Comput Struct Biotechnol J,2024, 24: 292-305.
[5] FLEISCHMANN C, SCHERAG A, ADHIKARI N K J, et al. Assessment of Global Incidence and Mortality of Hospital-treated Sepsis. Current Estimates and Limitations [J]. Am J Respir Crit Care Med, 2016, 193(3): 259-72.
[6] EVANS L, RHODES A, ALHAZZANI W, et al. Surviving sepsis campaign: international guidelines for management of sepsis and septic shock 2021[J]. Intensive care medicine, 2021, 47(11): 1181-247.
[7] PRESCOTT H C, ANGUS D C. Enhancing Recovery From Sepsis: A Review [J]. Jama, 2018, 319(1): 62-75.
[8] BRACHT H, HAFNER S, WEIß M. [Sepsis Update: Definition and Epidemiology] [J]. Anasthesiologie, Intensivmedizin, Notfallmedizin, Schmerztherapie : AINS, 2019, 54(1): 10-20.
[9] LELIGDOWICZ A, MATTHAY M A. Heterogeneity in sepsis: new biological evidence with clinical applications [J]. Crit Care, 2019, 23(1): 80.
[10] DELLINGER R P, BAGSHAW S M, ANTONELLI M, et al. Effect of Targeted Polymyxin B Hemoperfusion on 28-Day Mortality in Patients With Septic Shock and Elevated Endotoxin Level: The EUPHRATES Randomized Clinical Trial [J]. Jama, 2018, 320(14): 1455-63.
[11] SUPADY A, BRODIE D, WENGENMAYER T. Extracorporeal haemoadsorption: does the evidence support its routine use in critical care? [J]. The Lancet Respiratory medicine, 2022, 10(3): 307-12.
[12] KOMOROWSKI M, GREEN A, TATHAM K C, et al. Sepsis biomarkers and diagnostic tools with a focus on machine learning [J]. EBioMedicine, 2022, 86: 104394.
[13] GORDON A C, ALIPANAH-LECHNER N, BOS L D, et al. From ICU Syndromes to ICU Subphenotypes: Consensus Report and Recommendations For Developing Precision Medicine in ICU [J]. Am J Respir Crit Care Med, 2024.
[14] JOHNSON A E W, BULGARELLI L, SHEN L, et al. MIMIC-IV, a freely accessible electronic health record dataset [J]. Sci Data, 2023, 10(1): 1.
[15] Lê S, JOSSE J, HUSSON F. FactoMineR: An R Package for Multivariate Analysis [J]. Journal of Statistical Software, 2008, 25(1): 1- 18.
[16] MARTINS C, NEVES B, TEIXEIRA A S, et al. Identifying subgroups in heart failure patients with multimorbidity by clustering and network analysis [J]. BMC medical informatics and decision making, 2024, 24(1): 95.
[17] GAO C X, DWYER D, ZHU Y, et al. An overview of clustering methods with guidelines for application in mental health research [J]. Psychiatry Res, 2023, 327: 115265.
[18] BHATRAJU P K, PRINCE D K, MANSOUR S, et al. Integrated Analysis of Blood and Urine Biomarkers to Identify Acute Kidney Injury Subphenotypes and Associations With Long-term Outcomes [J]. American journal of kidney diseases : the official journal of the National Kidney Foundation, 2023, 82(3): 311-21.e1.
[19] ROUSSEEUW P J. Silhouettes: A graphical aid to the interpretation and validation of cluster analysis [J]. Journal of Computational and Applied Mathematics, 1987, 20: 53-65.
[20] CALIŃSKI T, HARABASZ J J C I S-T, METHODS. A dendrite method for cluster analysis [J]. 1974, 3: 1-27.
[21] DAVIES D L, BOULDIN D W. A Cluster Separation Measure [J]. IEEE transactions on pattern analysis and machine intelligence, 1979, PAMI-1(2): 224-7.
[22] DING L, MANE R, WU Z, et al. Data-driven clustering approach to identify novel phenotypes using multiple biomarkers in acute ischaemic stroke: A retrospective, multicentre cohort study [J]. EClinicalMedicine, 2022, 53: 101639.
[23] SEYMOUR C W, KENNEDY J N, WANG S, et al. Derivation, Validation, and Potential Treatment Implications of Novel Clinical Phenotypes for Sepsis [J]. Jama, 2019, 321(20): 2003-17.
[24] GOFTON T E, YOUNG G B. Sepsis-associated encephalopathy [J]. Nature reviews Neurology, 2012, 8(10): 557-66.
[25] WIDMANN C N, HENEKA M T. Long-term cerebral consequences of sepsis [J]. The Lancet Neurology, 2014, 13(6): 630-6.
[26] YANG K, CHEN J, WANG T, et al. Pathogenesis of sepsis-associated encephalopathy: more than blood-brain barrier dysfunction [J]. Molecular biology reports, 2022, 49(10): 10091-9.
[27] SCHULER A, WULF D A, LU Y, et al. The Impact of Acute Organ Dysfunction on Long-Term Survival in Sepsis [J]. Critical care medicine, 2018, 46(6): 843-9.
[28] LU X, QIN M, WALLINE J H, et al. CLINICAL PHENOTYPES OF SEPSIS-ASSOCIATED ENCEPHALOPATHY: A RETROSPECTIVE COHORT STUDY [J]. Shock (Augusta, Ga), 2023, 59(4): 583-90.
[29] CHEN J , SHI X , DIAO M , et al. A retrospective study of sepsis-associated encephalopathy: epidemiology, clinical features and adverse outcomes [J]. BMC emergency medicine, 2020, 20(1): 77.
[30] IWASHYNA T J, ELY E W, SMITH D M, et al. Long-term cognitive impairment and functional disability among survivors of severe sepsis [J]. Jama, 2010, 304(16): 1787-94.
[31] ANDONEGUI G, ZELINSKI E L, SCHUBERT C L, et al. Targeting inflammatory monocytes in sepsis-associated encephalopathy and long-term cognitive impairment [J]. JCI insight, 2018, 3(9).
[32] WöSTEN-VAN ASPEREN R M, LA ROI-TEEUW H M, VAN AMSTEL R B, et al. Distinct clinical phenotypes in paediatric cancer patients with sepsis are associated with different outcomes-an international multicentre retrospective study [J]. EClinicalMedicine, 2023, 65: 102252.
[33] SCHLAPBACH L J, WATSON R S, SORCE L R, et al. International Consensus Criteria for Pediatric Sepsis and Septic Shock [J]. Jama, 2024, 331(8): 665-74.
[34] RHODES A, EVANS L E, ALHAZZANI W, et al. Surviving Sepsis Campaign: International Guidelines for Management of Sepsis and Septic Shock: 2016[J]. Intensive care medicine, 2017, 43(3): 304-77.
[35] PULENDRAN B, DAVIS M M. The science and medicine of human immunology [J]. Science (New York, NY), 2020, 369(6511).
[36] BHAVANI S V, SEMLER M, QIAN E T, et al. Development and validation of novel sepsis subphenotypes using trajectories of vital signs [J]. Intensive care medicine, 2022, 48(11): 1582-92.
[37] HAO C, HAO R, ZHAO H, et al. Identification and validation of sepsis subphenotypes using time-series data [J]. Heliyon, 2024, 10(7): e28520.
[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]. Journal of translational medicine, 2020, 18(1): 462.
[39] FENG M, MCSPARRON J I, KIEN D T, et al. Transthoracic echocardiography and mortality in sepsis: analysis of the MIMIC-III database [J]. Intensive care medicine, 2018, 44(6): 884-92.
[40] ZHENG R, QIAN S, SHI Y, et al. Association between triglyceride-glucose index and in-hospital mortality in critically ill patients with sepsis: analysis of the MIMIC-IV database [J]. Cardiovasc Diabetol, 2023, 22(1): 307.
计量
- 文章访问数: 113
- HTML全文浏览量: 2
- PDF下载量: 19