medigraphic.com

2026, Number 1

<< Back Next >>

Med Crit 2026; 40 (1)

Prediction of sepsis phenotypes obtained through unsupervised machine learning in patient records in internal medicine and intensive care

Cárdenas CHE, Ramírez HJP, Mandujano SIJ, Merlos RJI, Sosa MMÁ

Language: Spanish
References: 42
Page: 14-24
PDF size: 302.10 Kb.


ABSTRACT

Introduction: sepsis is a complex disorder, understood as a syndrome rather than a disease, which has presented a challenge from the very moment of its definition. Therefore, clinical trials addressing its diagnosis, prognosis, and especially its treatment have been extremely challenging and have often resulted in failures, despite having a robust theoretical foundation. Thus, it is evident that the definition of sepsis is very common and broad, applicable to a heterogeneous group of patients who do not necessarily have the same disorder. Objective: to describe the prognosis of different sepsis phenotypes using unsupervised machine learning and clinical and biochemical variables obtained during the first 12 hours of hospitalization from the medical records of patients admitted to the Internal Medicine and Intensive Care units of the «Dr. Manuel Gea González» Hospital, during the period from January 1, 2022, to December 31, 2023. Material and methods: this was an observational, analytical, retrospective, cross-sectional, and retrospective study of 387 patient records of individuals admitted to the hospital with sepsis, according to the SEPSIS-3 consensus. Clinical and biochemical variables at admission were analyzed using unsupervised machine learning to distinguish different sepsis phenotypes. Results: of the 389 records included, 188 were categorized as cluster 1 and 201 as cluster 2, with a robustness of 0.0696 according to the silhouette index and 0.1002459 according to Dunn’s index. Acute kidney injury was documented in 77% (n = 144) of patients in cluster 1 and in 55% (n = 111) in cluster 2 (χ2 = 19.649, p = 0.0004998). Renal replacement therapy was recorded in 20% (n = 37) of cases in cluster 1 and in 4% (n = 8) in cluster 2 (χ2 = 23.408, p = 0.0004998). Conclusion: in the medical records of patients diagnosed with sepsis, the analysis of clinical and biochemical variables available during the first 12 hours results in a new sepsis categorization with prognostic utility regarding kidney injury and the need for renal replacement therapy.


REFERENCES

  1. Cecconi M, Evans L, Levy M, Rhodes A. Sepsis and septicshock. Lancet Lond Engl. 2018;392(10141):75-87.

  2. Singer M, Deutschman CS, Seymour CW, Shankar-Hari M,Annane D, Bauer M, et al. The third international consensusdefinitions for sepsis and septic shock (Sepsis-3). JAMA.2016;315(8):801-810.

  3. Shankar-Hari M, Phillips GS, Levy ML, Seymour CW, LiuVX, Deutschman CS, et al. Developing a new definition andassessing new clinical criteria for septic shock: for the thirdinternational consensus definitions for sepsis and septic shock(Sepsis-3). JAMA. 2016;315(8):775-787.

  4. Marik PE. Early management of severe sepsis: concepts andcontroversies. Chest. 2014;145(6):1407-1418.

  5. Ulloa L, Brunner M, Ramos L, Deitch EA. Scientific and clinicalchallenges in sepsis. Curr Pharm Des. 2009;15(16):1918-1935.

  6. Martin GS, Mannino DM, Eaton S, Moss M. The epidemiologyof sepsis in the United States from 1979 through 2000. N Engl JMed. 2003;348(16):1546-1554.

  7. Vincent JL, Sakr Y, Sprung CL, Ranieri VM, Reinhart K, GerlachH, et al. Sepsis in European intensive care units: results of theSOAP study. Crit Care Med. 2006;34(2):344-353.

  8. Rudd KE, Johnson SC, Agesa KM, Shackelford KA, Tsoi D,Kievlan DR, et al. Global, regional, and national sepsis incidenceand mortality, 1990-2017: analysis for the Global Burden ofDisease Study. Lancet Lond Engl. 2020;395(10219):200-211.

  9. Carrillo-Esper R, Carrillo-Córdova JR, Carrillo-Córdova LD.Epidemiological study of sepsis in Mexican intensive care units.Cir Cir. 2009;77(4):301-308; 279-285.

  10. Gorordo-Delsol LA, Merinos-Sánchez G, Estrada-Escobar RA,Medveczky-Ordoñez NI, Amezcua-Gutiérrez MA, Morales-Segura MA, et al. Sepsis and septic shock in emergencydepartments of Mexico: a multicenter point prevalence study.Gac Med Mex. 2020;156(6):486-492.

  11. McCarthy J, Minsky ML, Rochester N, Shannon CE. A proposalfor the Dartmouth summer research project on artificialintelligence. August 31, 1955. AI Mag. 2006;27:12-14.

  12. Calvo J, Guzman M, Ramos D. Machine Learning: una piezaclave en la transformación de los negocios [Internet]. www.ibm.com. Managment Solutions; 2024. Disponible en: https://www.ibm.com/es-es/topics/unsupervised-learning

  13. McCorduck P. Machines who think: a personal inquiry into the historyand prospects of artificial intelligence. 25th anniversary update. BocaRaton: CRC Press, Taylor & Francis Group; 2019. p. 565.

  14. El Naqa I, Murphy MJ. What is machine learning? In: El Naqa I,Li R, Murphy MJ, editors. Machine learning in radiation oncology.Cham: Springer International Publishing; 2015. pp. 3-11.

  15. LeCun Y, Bengio Y, Hinton G. Deep learning. Nature.2015;521(7553):436-444.

  16. Joshi S. Types of clustering algorithms in machine learningwith examples. Blogs & Updates on Data Science, BusinessAnalytics, AI Machine Learning. 2022.

  17. Chen M. What is unsupervised learning? [Internet]. Oracle.com.Oracle; 2024. Available in: https://www.oracle.com/mx/artificialintelligence/machine-learning/unsupervised-learning/

  18. Komorowski M, Celi LA, Badawi O, Gordon AC, Faisal AA. TheArtificial Intelligence Clinician learns optimal treatment strategiesfor sepsis in intensive care. Nat Med. 2018;24(11):1716-1720.

  19. De Corte T, Van Hoecke S, De Waele J. Artificial intelligencein infection management in the ICU. Crit Care Lond Engl.

  20. 2022;26(1):79.20. Fleuren LM, Klausch TLT, Zwager CL, Schoonmade LJ, GuoT, Roggeveen LF, et al. Machine learning for the prediction ofsepsis: a systematic review and meta-analysis of diagnostic testaccuracy. Intensive Care Med. 2020;46(3):383-400.

  21. Islam MM, Nasrin T, Walther BA, Wu CC, Yang HC, Li YC.Prediction of sepsis patients using machine learning approach: ameta-analysis. Comput Methods Programs Biomed. 2019;170:1-9.

  22. BISEPRO: inteligencia artificial para agilizar el diagnósticode la sepsis [Internet]. iSanidad; 2018. Disponible en: https://isanidad.com/126725/bisepro-inteligencia-artificial-para-agilizarel-diagnostico-de-la-sepsis/

  23. Lamping F, Jack T, Rubsamen N, Sasse M, Beerbaum P,Mikolajczyk RT, et al. Development and validation of a diagnosticmodel for early differentiation of sepsis and non-infectious SIRSin critically ill children - a data-driven approach using machinelearningalgorithms. BMC Pediatr. 2018;18(1):112.

  24. Sharafoddini A, Dubin JA, Lee J. Identifying subpopulations ofseptic patients: A temporal data-driven approach. Comput BiolMed. 2021;130:104182.

  25. Scicluna BP, van Vught LA, Zwinderman AH, Wiewel MA,Davenport EE, Burnham KL, et al. Classification of patients withsepsis according to blood genomic endotype: a prospectivecohort study. Lancet Respir Med. 2017;5(10):816-826.

  26. Geri G, Vignon P, Aubry A, Fedou AL, Charron C, Silva S, et al.Cardiovascular clusters in septic shock combining clinical andechocardiographic parameters: a post hoc analysis. IntensiveCare Med. 2019;45(5):657-667.

  27. Kattan E, Bakker J, Estenssoro E, Ospina-Tascón GA,Cavalcanti AB, Backer DD, et al. Hemodynamic phenotypebased,capillary refill time-targeted resuscitation in early septicshock: The ANDROMEDA-SHOCK-2 Randomized Clinical Trialstudy protocol. Rev Bras Ter Intensiva. 2022;34(1):96-106.

  28. Gardlund B, Dmitrieva NO, Pieper CF, Finfer S, Marshall JC,Taylor Thompson B. Six subphenotypes in septic shock: latentclass analysis of the PROWESS shock study. J Crit Care.2018;47:70-79.

  29. Zhao H, Kennedy JN, Wang S, Brant EB, Bernard GR, DeMerleK, et al. Revising host phenotypes of sepsis using microbiology.Front Med. 2021;8:775511.

  30. Papin G, Bailly S, Dupuis C, Ruckly S, Gainnier M, Argaud L,et al. Clinical and biological clusters of sepsis patients usinghierarchical clustering. PloS One. 2021;16(8):e0252793.

  31. Hu C, Li Y, Wang F, Peng Z. Application of machine learningfor clinical subphenotype identification in sepsis. Infect Dis Ther.2022;11(5):1949-1964.

  32. Seymour CW, Kennedy JN, Wang S, Chang CCH, Elliott CF,Xu Z, et al. Derivation, validation, and potential treatmentimplications of novel clinical phenotypes for sepsis. JAMA.2019;321(20):2003-2017.

  33. Pierrakos C, Velissaris D, Bisdorff M, Marshall JC, VincentJL. Biomarkers of sepsis: time for a reappraisal. Crit Care.2020;24(1):287.

  34. Póvoa P, Coelho L, Dal-Pizzol F, Ferrer R, Huttner A, ConwayMorris A, et al. How to use biomarkers of infection or sepsis at thebedside: guide to clinicians. Intensive Care Med. 2023;49(2):142-153.

  35. Huang Z, Fu Z, Huang W, Huang K. Prognostic value ofneutrophil-to-lymphocyte ratio in sepsis: a meta-analysis. Am JEmerg Med. 2020;38(3):641-647.

  36. Wu H, Cao T, Ji T, Luo Y, Huang J, Ma K. Predictive value ofthe neutrophil-to-lymphocyte ratio in the prognosis and risk ofdeath for adult sepsis patients: a meta-analysis. Front Immunol.2024;15:1336456.

  37. Thomas-Ruddel DO, Frohlich H, Schwarzkopf D, Bloos F,Riessen R. Sepsis and underlying comorbidities in intensivecare unit patients : Analysis of the cause of death by differentclinicians-a pilot study. Med Klin Intensivmed Notfmed.2024;119(2):123-128.

  38. Sharma N, Schwendimann R, Endrich O, Ausserhofer D, SimonM. Comparing Charlson and Elixhauser comorbidity indices withdifferent weightings to predict in-hospital mortality: an analysis ofnational inpatient data. BMC Health Serv Res. 2021;21(1):13.

  39. Eguia E, Cobb AN, Baker MS, Joyce C, Gilbert E, Gonzalez R, etal. Risk factors for infection and evaluation of Sepsis-3 in patientswith trauma. Am J Surg. 2019;218(5):851-857.

  40. Knaus WA, Marks RD. New phenotypes for sepsis: thepromise and problem of applying machine learning and artificialintelligence in clinical research. JAMA. 2019;321(20):1981-1982.

  41. La Via L, Sangiorgio G, Stefani S, Marino A, Nunnari G,Cocuzza S, et al. The global burden of sepsis and septic shock.Epidemiologia (Basel). 2024;5(3):456-478.

  42. Davenport EE, Burnham KL, Radhakrishnan J, Humburg P,Hutton P, Mills TC, et al. Genomic landscape of the individualhost response and outcomes in sepsis: a prospective cohortstudy. Lancet Respir Med. 2016;4(4):259-271. doi: 10.1016/S2213-2600(16)00046-1.




2020     |     www.medigraphic.com