PriMera Scientific Medicine and Public Health (ISSN: 2833-5627)

Research Article

Volume 5 Issue 3

Data sources (LLM) for a Clinical Decision Support Model (SSDC) using a Healthcare Interoperability Resources (HL7-FHIR) platform for an ICU Ecosystem

Bernardo Chávez Plaza*, Luis Chicuy Godoy, Mario Cuellar Martínez and Rodrigo Covarrubias Ganderats

August 18, 2024

DOI : 10.56831/PSMPH-05-161

View PDF

Abstract

The increasing development of digital technologies in recent years has led to a concomitant increase in the availability of data in ICUs. For this vast amount of information to be useful, it must be processed and analyzed to extract meaningful information. However, its size and complexity often exceed the capabilities of traditional tools, motivating ongoing research efforts to develop new analytical techniques better able to address these challenges. This effort has resulted in the maturation of disciplines such as artificial intelligence, machine learning (ML), data mining (LLM), parallel computing, and many others. Despite these advances, countless challenges of modern computing remain unaddressed and strategies to extract knowledge from complex data will undoubtedly persist as an active area of research in the years to come.

Real-time risk estimation of isolated pathologies provides interpretable information to understand the different risks of patients with multiple pathologies using electronic health records (EHRs) in an ICU patient; However, in this case there are fundamental problems when formulating hypotheses such as sample selection bias, imprecise variable definitions, implementation limitations, frequency of variable measurement, subjective treatment assignment and model overfitting.

Decision-making and predictive models (CDSS), on the other hand, are not yet widely developed with the current known health systems. However, their potential based on massive data sources, allows with the structured data of data lakes, to perform artificial intelligence (AI) to improve training and control for algorithms according to the different requirements and security that we must carry out and that we will build for the different syndromes. Here we present an CDSS model that captures data from public ICUs and we show in our report the potential data mining, for later analysis with different predictive models.

We highlight that with the current results of a public ICU, through a Smart ICU there is 0.003% that corresponds to EHR data and only 3.97% is structured data in data lakes that are susceptible to useful algorithms -at the present time- for an CDSS system.

Keywords: Data sources; Intensive Care Units; CDSS; Artificial Intelligence; Interoperability; EHR

References

  1. McKenzie Mary S., et al. “An Observational Study of Decision Making by Medical Intensivists”. Critical Care Medicine 43.8 (2015): 1660-1668.
  2. Lundgrén­Laine H., et al. “Managing daily activities in intensive care: an observational study of ad hoc decision making by attending nurses and intensivists”. Critical Care 15.4 (2011): R188.
  3. Curtis JR and Vincent J-L. “Ethics and end-of-life care for adults in the intensive care unit”. The Lancet 376.9749 (2010): 1347-53.
  4. Meskó B and Topol EJ. “The imperative for regulatory oversight of large language models (or generative AI) in healthcare”. NPJ Digit. Med 6 (2023): 120.
  5. Piers RD., et al. “Perceptions of Appropriateness of Care Among European and Israeli Intensive Care Unit Nurses and Physicians”. JAMA 306.24 (2011): 2694-703.
  6. Lapsley I and Melia K. “Clinical actions and financial constraints: the limits to rationing intensive care”. Sociology of Health & Illness 23.5 (2001): 729-46.
  7. Mingze Yuan., et al. “Large language models illuminate a progressive pathway to artificial intelligent healthcare assistant”. Medicine Plus (2024).
  8. Trentini F., et al. “Pressure on healthcare system and intensive care utilization during the COVID-19 outbreak in the Lombardy region: a retrospective observational study on 43,538 hospitalized patients”. Am J Epidemiol 191.1 (2022): 137-146.
  9. Thoral PJ., et al. “Explainable Machine Learning on Amsterdam UMCdb for ICU Discharge Decision Support: Uniting Intensivists and Data Scientists”. Critical Care Explorations 3.9 (2021): e0529.
  10. Cosgriff CV, Celi LA and Sauer CM. “Boosting Clinical Decision-making: Machine Learning for Intensive Care Unit Discharge”. Annals ATS 15.7 (2018): 804-5.
  11. Peng C., et al. “A study of generative large language model for medical research and healthcare”. NPJ Digit Med 6.1 (2023): 210.
  12. Wornow M., et al. “The shaky foundations of large language models and foundation models for electronic health records”. NPJ Digit Med 6 (2023): 135.
  13. Sauer Christopher M., et al. “Leveraging electronic health records for data science: common pitfalls and how to avoid them”. The Lancet Digital Health 4.12 (2022): e893-e898.
  14. Chávez Plaza B., et al. “HL7 FHIR Platform, Scalable, Reliable and Comprehensive of Clinical Databases Analyzed with Machine Learning for ICU Public Healthcare Center”. In: Pino, E., Magjarević, R., de Carvalho, P. (eds) International Conference on Biomedical and Health Informatics 2022. IFMBE Proceedings, Springer vol 108.
  15. Chávez B., et al. “Data Sources for use in a Healthcare Interoperability Resources (HL7-FHIR) Platform. A Decision Guide for Clinicians and Data Scientists in Public ICUs”. Clareus Scientific Medical Sciences 1.1 (2024): 35-42.
  16. Tobias Gentner., et al. “Data Lakes in Healthcare: Applications and Benefits from the Perspective of Data Sources and Players”. Procedia Computer Science 225 (2023): 1302-1311.
  17. Sutton RT., et al. “An overview of clinical decision support systems: benefits, risks, and strategies for success”. npj Digit. Med 3 (2020): 17.
  18. Hripcsak G. “Arden Syntax for Medical Logic Modules”. MD Comput 8.2 (1991): 76-8.
  19. HL7 International. Arden Syntax v 2.10 (Health Level Seven Arden Syntax for Medical Logic Systems, Version2.10).
  20. Johnson PD., et al. “A virtual medical record for guideline-based decision support”. Proc AMIA Symp 2001: 294-8.
  21. Topol EJ. “High-performance medicine: the convergence of human and artificial intelligence”. Nat Med 25 (2019): 44-56.
  22. Bailly S, Meyfroidt G and Timsit JF. “What's new in ICU in 2050: big data and machine learning”. Cuidados Intensitivos Med 44 (2018): 1524-1527.
  23. Ghassemi M, Celi LA and Stone DJ. “Cutting-edge review: the data revolution in intensive care”. Critical Care 19 (2015): 118.
  24. Beam AL and Kohane IS. “Translating artificial intelligence into clinical care”. JAMA 316 (2016): 2368-2369.
  25. Ince C. “Intensive care medicine in 2050: the ICU in vivo”. Intensive Care Med 43 (2017): 1700-1702.
  26. Nemati S., et al. “An interpretable machine learning model for accurate prediction of sepsis in the ICU”. Crit Care Med 46 (2018): 547-553.
  27. Pirracchio R., et al. “Mortality predicting mortality in intensive care units with the ICU super learning algorithm (SICULA): a population-based study”. Lancet Respir Med 3 (2015): 42-52.
  28. Gårdlund B., et al. “Six subphenotypes in septic shock: latent class analysis of the PROWESS Shock study”. J Crit Care 47 (2018): 70-79.
  29. Q Mao., et al. “Multicentre validation of a sepsis prediction algorithm using only vital sign data in the emergency department, general ward and ICU”. BMJ Open 8.1 (2018): e017833.
  30. Antcliffe DB., et al. “Transcriptomic signatures in sepsis and a differential response to steroids: from the VANISH randomized trial”. Am J Respir Crit Care Med 199 (2018): 980-986.
  31. Seymour CW., et al. “Derivation, validation, and potential treatment implications of novel clinical phenotypes for sepsis”. JAMA 321.20 (2019): 2003-2017.
  32. Sutton RS and Barto AG. “Reinforcement Learning: An Introduction”. 2nd ed. MIT Press, Cambridge (2018).
  33. Guiza F, Van Eyck J and Meyfroidt G. “Predictive data mining on intensive care unit monitoring data”. J Clin Monitor Comput 27 (2013): 449-453.
  34. Gottesman O., et al. “Guidelines for reinforcement learning in health care”. Nat Med 25 (2019): 16-18.
  35. Komorowski M., et al. “The Artificial Intelligence Clinician learns optimal treatment strategies for sepsis in intensive care”. Nat Med 24 (2018): 1716-1720.
  36. HL7 International. DSTU updates. http://hl7.org/fhir/DSTU1/
  37. HL7 International. Welcome to FHIR. http://hl7.org/fhir/R4/
  38. 21st Century Cures Act: Interoperability, information blocking, and the ONC Health IT Certification Program. Fed Regist 85 (2020): 25642-961.
  39. Jung CY, Sward KA and Haug PJ. “Executing medical logic modules expressed in ArdenML using Drools”. J Am Med Inform Assoc 19.4 (2012): 533-6.
  40. Bernardo Chávez P., et al. “SUB-PROJECT: “Monitoring with dashboards for centralized management of data in critical patients, ensuring the security and quality of services, with a robust, interoperable and resilient Artificial Intelligence and Machine Learning platform, at El Salvador Hospital, Santiago, Chile”. International Conference on Biomedical and Health Informatics 2022. IFMBE Proceedings (2022).