Predictive tools in intensive care unit management: a protocol of a scoping review
Study Protocol

Predictive tools in intensive care unit management: a protocol of a scoping review

Roberta Eduarda Grolli1 ORCID logo, Greici Capellari Fabrizzio1 ORCID logo, Eduardo Cunha Cabral2 ORCID logo, Jayme Mattos de Souza3 ORCID logo, Lucas de Souza Vieira2 ORCID logo, Anderson Sales Zambeli2, Gabriela Machado Silva3 ORCID logo, Matheus Machado4 ORCID logo, Luis Gustavo Bornia4 ORCID logo, Simone Silmara Werner5 ORCID logo, Jônata Tyska Carvalho5 ORCID logo, Renato Fileto5 ORCID logo, André Wüst Zibetti5 ORCID logo

1Postgraduate Program in Nursing, Federal University of Santa Catarina, Florianópolis, SC, Brazil; 2Degree Computer Science, Department of Informatics and Statistics (INE), Federal University of Santa Catarina, Florianópolis, SC, Brazil; 3Postgraduate Program in Health Informatics, Federal University of Santa Catarina, Florianópolis, SC, Brazil; 4Postgraduate Program in Computer Science, Federal University of Santa Catarina, Florianópolis, SC, Brazil; 5Department of Informatics and Statistics (INE), Federal University of Santa Catarina, Florianópolis, SC, Brazil

Contributions: (I) Conception and design: All authors; (II) Administrative support: All authors; (III) Provision of study materials or patients: None; (IV) Collection and assembly of data: All authors; (V) Data analysis and interpretation: None; (VI) Manuscript writing: All authors; (VII) Final approval of manuscript: All authors.

Correspondence to: Roberta Eduarda Grolli. Postgraduate Program in Nursing, Federal University of Santa Catarina, Rua Agrônomo Andrei Cristian Ferreira, s/n Trindade, Florianópolis, SC 88034-001, Brazil. Email: robertaeduarda06@gmail.com.

Background: Effective management of intensive care unit (ICU) beds is a crucial factor for the efficiency of healthcare systems. This study proposes a scoping review aimed at identifying predictive models used in ICU bed management.

Methods: The review aims to answer the following research question “What are the predictive models used in ICU bed management?”, will follow the methodology of the Joanna Briggs Institute and the Preferred Reporting Items for Systematic Reviews and Meta-Analyses for Scoping Reviews guidelines. Will be considering publications since 2013, the year in which the National Patient Safety Program was implemented in Brazil. Predictive models, based on artificial intelligence techniques and the analysis of large datasets, have shown great potential for optimizing resource allocation, predicting bed occupancy, and improving patient care.

Discussion: The literature has already documented the application of these tools to predict patient transfers, support diagnosis, and manage hospital services. The findings of this review will provide a theoretical foundation for the development of a predictive model to be implemented in the state of Santa Catarina, Brazil, aimed to reduce waiting times for ICU beds and enhance the management of these units.

Trial Registration: Open Science Framework (https://doi.org/10.17605/OSF.IO/5XNPZ).

Keywords: Artificial intelligence (AI); intensive care units (ICUs); medical informatics; hospital administration

Received: 23 October 2024; Accepted: 17 February 2025; Published online: 11 April 2025.

doi: 10.21037/jmai-24-390

Introduction

The proposed scoping review constitutes the initial phase of a multi-stage research project aimed at identifying the predictive models used in intensive care unit (ICU) management.

An ICU is a specialized environment, essential for the efficiency of the healthcare system, particularly during periods of high demand. The availability of ICU beds is a critical indicator that directly impacts mortality rates and clinical outcomes. The World Health Organization (WHO) recommends 1 to 3 ICU beds for every 10,000 inhabitants. Although Brazil is aligned with this average, significant regional disparities exist, especially within the public healthcare system [Sistema Único de Saúde (SUS)]. The southern region, for example, has an overall figure of 2.2 beds per 10,000 inhabitants, but when looking only at the availability of public beds, this figure drops to 1.8 per 10,000 inhabitants. Moreover, the occupancy rate frequently exceeds 80%, placing the healthcare system in a constant state of critical alert (1).

Effective ICU bed management is a central challenge for many healthcare systems, as the allocation of resources directly influences patient care. In this context, various technologies are being explored to support this critical resource management. Studies have demonstrated the value of predictive models in decision-making, especially in scenarios with limited time and resources (2).

Despite growing interest and advancements in predictive models for ICU management, significant gaps remain regarding their application, efficacy, and adaptability across diverse healthcare settings. Existing studies often focus on specific aspects, such as patient-level predictions or operational improvements, without comprehensively addressing their development, validation, and implementation. The variability in methodologies, datasets, and performance metrics further complicates drawing generalized conclusions (2-4). A scoping review can bridge these gaps by mapping existing evidence, identifying trends and limitations, and providing a clearer understanding of the state of the art, ultimately guiding future research and informing decision-making in ICU resource allocation and patient care optimization (5,6).

Methods

This review will follow the methodology outlined by the Joanna Briggs Institute (JBI) for scoping reviews. The development of the review protocol is aligned with the Preferred Reporting Items for Systematic Reviews and Meta-Analyses for Scoping Reviews (PRISMA-ScR) guidelines and has been registered with the Open Science Framework (https://doi.org/10.17605/OSF.IO/5XNPZ).

Review question

The research question: “What are the predictive models used in ICU bed management?” was guided by the population, concept, and context (PCC) mnemonic (7): population: predictive models; concept: bed management; context: ICUs.

Participants

This scoping review will include studies from theses, dissertations, conference presentations, and scientific articles that describe the development and application of predictive models for ICU bed management.

Inclusion criteria

The study will include works published since 2013, marking the start of the National Patient Safety Program (PNSP), which established clear protocols and guidelines to enhance patient safety in hospital settings, particularly in supporting management and care. Only studies in Portuguese, English, and Spanish, with complete and free access, will be considered. Regarding exclusion criteria, articles that do not address the research question will not be included.

Population

This study will consider works that use predictive models as tools for ICU bed management.

Concept

Bed management, also known as hospital access management, involves the control of bed occupancy and turnover, as well as the management of hospital supplies, from the admission of a patient until their discharge. In this process, decision-support tools are invaluable, as detailed analyses can identify bottlenecks throughout the patient cycle. It is notable that optimizing hospital beds directly impacts patient mortality rates and clinical outcomes, particularly in emergencies such as pandemics or natural disasters (8).

Context

An ICU is a sector characterized by the distribution of critical resources, as the admission rate in intensive care is associated with patient recovery rates. Specifically, in critically ill patients, there is a 1.5% increase in mortality for each hour of delay in admission to the ICU (9).

Source of evidence

This study will consider research employing quantitative, qualitative, and mixed methods, as well as conference abstracts, experience reports, manuals, books, gray literature and guidelines.

Search strategy

The search will be conducted in electronic databases such as PubMed, Scopus, Web of Science, and Embase, as well as in gray literature sources to include unpublished or non-indexed studies. Keywords and controlled vocabulary (MeSH terms) were combined using Boolean operators, as presented in Table 1. The strategy was refined through preliminary searches and expert consultation. Additionally, a manual search of the reference lists of included studies will be conducted to enhance literature coverage.

Table 1

Search strategy and database

Database Search key words
Brazilian Digital Library of Theses and Dissertations (“Triagem” OR “Triagens” OR “Classificação de paciente” OR “Classificação de pacientes” OR “Triaje” OR “Clasificación de Pacientes” OR “Clasificación de laPrioridadAsistencial” OR “Triado Asistencial de Pacientes” OR “Triado de Pacientes” OR “Triage” OR “Triages” OR “triaged” OR “triaging” OR “PatientClassification” OR “ClassificationofCarePriority”) AND (“Unidades de terapia intensiva” OR “UTI” OR “CTI” OR “terapia intensiva” OR “centro intensivo” OR “centros intensivos” OR “cuidados intensivos” OR “cuidado intensivo” OR “Cuidados Críticos” OR “Cuidado Crítico” OR “Unidades de Cuidados Intensivos” OR “Unidad de Cuidados Intensivos” OR “Unidad de Vigilancia Intensiva” OR “UCI” OR “UVI” OR “IntensiveCareUnits” OR “Intensivecare” OR “ICU” OR “Intensivecares”) AND (“Previsões” OR “Predição” OR “Predições” OR “Alocação de Recursos” OR “Aprendizado de Máquina” OR “Aprendizado Automático” OR “Aprendizagem Automática” OR “Aprendizagem de Máquina” OR Algoritmo* OR “Predicción” OR Predictivo* OR “Predicciones” OR “Asignación de Recursos” OR “Aprendizaje Automático” OR “aprendizaje por transferencia” OR “Forecasting” OR Prediction* OR “ResourceAllocation” OR “Machine Learning” OR “Transfer Learning” OR Algorithm*)
Catalog of Theses and Dissertations triage* AND “terapia intensiva”
CINAHL (EBSCO) (“Triage” OR “Triages” OR “triaged” OR “triaging” OR “Patient Classification” OR “Classification of Care Priority”) AND (“Intensive Care Units” OR “Intensive care” OR “ICU” OR “Intensive cares”) AND (“Forecasting” OR Prediction* OR “Resource Allocation” OR “Machine Learning” OR “Transfer Learning” OR Algorithm*)
Embase (Elsevier) (“Triage” OR “Triages” OR “triaged” OR “triaging” OR “Patient Classification” OR “Classification of Care Priority”) AND (“Intensive Care Units” OR “Intensive care” OR “ICU” OR “Intensive cares”) AND (“Forecasting” OR Prediction* OR “Resource Allocation” OR “Machine Learning” OR “Transfer Learning” OR Algorithm*)
IEEE Xplore (“Triage” OR “Triages” OR “triaged” OR “triaging” OR “Patient Classification” OR “Classification of Care Priority”) AND (“Intensive Care Units” OR “Intensive care” OR “ICU” OR “Intensive cares”) AND (“Forecasting” OR Prediction* OR “Resource Allocation” OR “Machine Learning” OR “Transfer Learning” OR Algorithm*)
LILACS/BDENF (“Triagem” OR “Triagens” OR “Classificação de paciente” OR “Classificação de pacientes” OR “Triaje” OR “Clasificación de Pacientes” OR “Clasificación de laPrioridadAsistencial” OR “Triado Asistencial de Pacientes” OR “Triado de Pacientes” OR “Triage” OR “Triages” OR “triaged” OR “triaging” OR “PatientClassification” OR “ClassificationofCarePriority”) AND (“Unidades de terapia intensiva” OR “UTI” OR “CTI” OR “terapia intensiva” OR “centro intensivo” OR “centros intensivos” OR “cuidados intensivos” OR “cuidado intensivo” OR “Cuidados Críticos” OR “Cuidado Crítico” OR “Unidades de Cuidados Intensivos” OR “Unidad de Cuidados Intensivos” OR “Unidad de Vigilancia Intensiva” OR “UCI” OR “UVI” OR “IntensiveCareUnits” OR “Intensivecare” OR “ICU” OR “Intensivecares”) AND (“Previsões” OR “Predição” OR “Predições” OR “Alocação de Recursos” OR “Aprendizado de Máquina” OR “Aprendizado Automático” OR “Aprendizagem Automática” OR “Aprendizagem de Máquina” OR Algoritmo* OR “Predicción” OR Predictivo* OR “Predicciones” OR “Asignación de Recursos” OR “Aprendizaje Automático” OR “aprendizaje por transferencia” OR “Forecasting” OR Prediction* OR “ResourceAllocation” OR “Machine Learning” OR “Transfer Learning” OR Algorithm*)
ProQuest Dissertations & Theses Global NOFT (“Triage” OR “Triages” OR “triaged” OR “triaging” OR “Patient Classification” OR “Classification of Care Priority”) AND (“Intensive Care Units” OR “Intensive care” OR “ICU” OR “Intensive cares”) AND (“Forecasting” OR Prediction* OR “Resource Allocation” OR “Machine Learning” OR “Transfer Learning” OR Algorithm*)
PubMed/Medline (“Triage” OR “Triages” OR “triaged” OR “triaging” OR “Patient Classification” OR “Classification of Care Priority”) AND (“Intensive Care Units” OR “Intensive care” OR “ICU” OR “Intensive cares”) AND (“Forecasting” OR Prediction* OR “Resource Allocation” OR “Machine Learning” OR “Transfer Learning” OR Algorithm*)
SciELO (“Triagem” OR “Triagens” OR “Classificação de paciente” OR “Classificação de pacientes” OR “Triaje” OR “Clasificación de Pacientes” OR “Clasificación de laPrioridadAsistencial” OR “Triado Asistencial de Pacientes” OR “Triado de Pacientes” OR “Triage” OR “Triages” OR “triaged” OR “triaging” OR “PatientClassification” OR “ClassificationofCarePriority”) AND (“Unidades de terapia intensiva” OR “UTI” OR “CTI” OR “terapia intensiva” OR “centro intensivo” OR “centros intensivos” OR “cuidados intensivos” OR “cuidado intensivo” OR “Cuidados Críticos” OR “Cuidado Crítico” OR “Unidades de Cuidados Intensivos” OR “Unidad de Cuidados Intensivos” OR “Unidad de Vigilancia Intensiva” OR “UCI” OR “UVI” OR “IntensiveCareUnits” OR “Intensivecare” OR “ICU” OR “Intensivecares”) AND (“Previsões” OR “Predição” OR “Predições” OR “Alocação de Recursos” OR “Aprendizado de Máquina” OR “Aprendizado Automático” OR “Aprendizagem Automática” OR “Aprendizagem de Máquina” OR Algoritmo* OR “Predicción” OR Predictivo* OR “Predicciones” OR “Asignación de Recursos” OR “Aprendizaje Automático” OR “aprendizaje por transferencia” OR “Forecasting” OR Prediction* OR “ResourceAllocation” OR “Machine Learning” OR “Transfer Learning” OR Algorithm*)
Scopus (Elsevier) (“Triage” OR “Triages” OR “triaged” OR “triaging” OR “Patient Classification” OR “Classification of Care Priority”) AND (“Intensive Care Units” OR “Intensive care” OR “ICU” OR “Intensive cares”) AND (“Forecasting” OR Prediction* OR “Resource Allocation” OR “Machine Learning” OR “Transfer Learning” OR Algorithm*)
Web of Science (Clarivate Analytics) (“Triage” OR “Triages” OR “triaged” OR “triaging” OR “Patient Classification” OR “Classification of Care Priority”) AND (“Intensive Care Units” OR “Intensive care” OR “ICU” OR “Intensive cares”) AND (“Forecasting” OR Prediction* OR “Resource Allocation” OR “Machine Learning” OR “Transfer Learning” OR Algorithm*)
ACM Digital Library (“Triage” OR “Triages” OR “triaged” OR “triaging” OR “Patient Classification” OR “Classification of Care Priority”) AND (“Intensive Care Units” OR “Intensive care” OR “ICU” OR “Intensive cares”) AND (“Forecasting” OR Prediction* OR “Resource Allocation” OR “Machine Learning” OR “Transfer Learning” OR Algorithm*)

ACM, Association for Computing Machinery; BDENF, Brazilian Nursing Database; CINAHL, Cumulative Index to Nursing and Allied Health Literature; IEEE, Institute of Electrical and Electronics Engineers; LILACS, Latin American and Caribbean Center on Health Sciences Information.

Screening and selection of evidence

The articles collected for analysis will be imported into the reference management software Rayyan, where duplicate entries will be screened. Subsequently, two independent reviewers will evaluate the titles and abstracts of the articles according to the inclusion and exclusion criteria to determine which studies will proceed to full-text reading and subsequent data extraction.

Analysis and presentation of results

The results will be presented in tables, which will be refined during the data extraction phase, and categorized according to the predefined research categories to address the questions outlined in the protocol. Alongside the tabulated data, a narrative synthesis will be provided, highlighting the main characteristics of the selected literature, with a focus on the predictive models used in ICU bed management.

Ethics and dissemination

This study does not require submission to an ethics committee, as it utilizes publicly available materials. However, it will strictly adhere to ethical guidelines and scientific writing standards, respecting and appropriately citing the ideas and knowledge of other authors based on published materials accessible in public databases. Regarding dissemination, the findings of this scoping review will serve as a theoretical foundation for the development of a predictive model aimed at enhancing decision-making related to ICU bed management.

Different countries exhibit distinct characteristics in their healthcare systems, with some adopting prevention-focused approaches, such as the National Health Service (NHS) in England, while others lean towards more medical-hospital-centered models, like that of the United States. Brazil, on the other hand, features the Unified Health System (SUS), which emphasizes prevention while also incorporating medical and hospital resources as necessary. Despite geographic, cultural, and funding differences, many countries face challenges in managing healthcare resources, particularly those that are high-cost and specialized (10).

Discussion

The difficulty in resource management becomes particularly evident during natural disasters and pandemics, as seen in the case of coronavirus disease 2019 (COVID-19). With the exponential increase in cases and the growing demand for intensive care in ICUs, various countries experienced the collapse of their healthcare systems. In Brazil, research indicates that the pandemic impacted all sectors of care, with the number of deaths from COVID-19 paralleling the mortality rates from other causes. These peaks coincide with periods of reduced healthcare service delivery, suggesting an overload of the hospital network caused by the pandemic (11).

In this context, techniques for processing large volumes of data have emerged as valuable tools for healthcare management (12,13). It is estimated that approximately 30% of the data generated and stored globally pertains to the health sector (14). Given this backdrop, this protocol aims to explore studies that employ advanced data processing techniques to predict ICU bed occupancy.

Predictive models are widely utilized across various healthcare domains. Recent research highlights these models as essential tools for supporting decision-making in environments that require rapid responses and have limited resources (2). This utility arises from the ability of these techniques to extract hidden insights, resulting in models that assist not only in decision-making but also in process automation (15). Predictive models do not focus on inferential relationships; they aim to identify variables capable of predicting potential outcomes. These characteristics render predictive models particularly relevant to healthcare, contributing to improved clinical practice at the care level (16) and in management (17).

However, implementing predictive models in healthcare, particularly in ICU bed management, presents several challenges. First, there is the issue of data quality and completeness. Missing, outdated, or inconsistent data can compromise model accuracy. Second, the integration of predictive tools into clinical workflows requires significant adjustments, including training staff and ensuring technological infrastructure. Ethical considerations, such as data privacy and the potential for algorithmic biases, also pose challenges. Potential solutions include establishing standardized data collection protocols, investing in robust infrastructure, and employing explainable artificial intelligence (AI) models to enhance trust and transparency.

This protocol has notable strengths. By considering a broad range of predictive models, whether mathematical or machine learning-based, it ensures a comprehensive analysis of their applicability in ICU resource management. Additionally, the focus on actionable insights supports decision-making processes, aiming to reduce waiting times and improve patient outcomes. Nonetheless, the protocol has its limitations. For instance, differences in healthcare systems and available resources across countries may limit the generalizability of findings. Furthermore, the reliance on existing studies means the results are contingent on the quality and scope of the included literature.

Upon completion, this study will offer valuable insights into the proactive forecasting of ICU bed needs, minimizing human error and subjective decision-making while grounding processes in data. These contributions are expected to enhance patient care, optimize resource allocation, and provide a foundation for future research in this critical area.

Acknowledgments

None.

Footnote

Peer Review File: Available at https://jmai.amegroups.com/article/view/10.21037/jmai-24-390/prf

Funding: None.

Conflicts of Interest: All authors have completed the ICMJE uniform disclosure form (available at https://jmai.amegroups.com/article/view/10.21037/jmai-24-390/coif). The authors have no conflicts of interest to declare.

Ethical Statement: The authors are accountable for all aspects of the work in ensuring that questions related to the accuracy or integrity of any part of the work are appropriately investigated and resolved.

Open Access Statement: This is an Open Access article distributed in accordance with the Creative Commons Attribution-NonCommercial-NoDerivs 4.0 International License (CC BY-NC-ND 4.0), which permits the non-commercial replication and distribution of the article with the strict proviso that no changes or edits are made and the original work is properly cited (including links to both the formal publication through the relevant DOI and the license). See: https://creativecommons.org/licenses/by-nc-nd/4.0/.

References

  1. Power BI Report. 2024 [cited 2024 Oct 7]. Available online: https://cieges.saude.sc.gov.br/
  2. Hou W, Zhao Z, Chen A, et al. Machining learning predicts the need for escalated care and mortality in COVID-19 patients from clinical variables. Int J Med Sci 2021;18:1739-45. [Crossref] [PubMed]
  3. Soares FM, da Rocha Carvalho Rosa LO, Cecatti JG, et al. Design, construction, and validation of obstetric risk classification systems to predict intensive care unit admission. Int J Gynaecol Obstet 2024;167:1243-54. [Crossref] [PubMed]
  4. Covino M, Sandroni C, Santoro M, et al. Predicting intensive care unit admission and death for COVID-19 patients in the emergency department using early warning scores. Resuscitation 2020;156:84-91. [Crossref] [PubMed]
  5. Alpaydın E. Introduction to Machine Learning. 4th edition. Cambridge, MA, USA: The MIT Press; 2020.
  6. O'Neill TJ, Xi Y, Stehel E, et al. Active Reprioritization of the Reading Worklist Using Artificial Intelligence Has a Beneficial Effect on the Turnaround Time for Interpretation of Head CT with Intracranial Hemorrhage. Radiol Artif Intell 2020;3:e200024. [Crossref] [PubMed]
  7. Peters MDJ, Marnie C, Tricco AC, et al. Updated methodological guidance for the conduct of scoping reviews. JBI Evid Synth 2020;18:2119-26. [Crossref] [PubMed]
  8. Vranas KC, Kerlin MP. Looking to the Past, Learning From the Present, and Preparing for the Future: Toward Understanding Critical Care Strain During a Global Pandemic. Chest 2021;159:469-70. [Crossref] [PubMed]
  9. Lepre RL, Mezzaroba AL, Cardoso LT, et al. Recusa de leitos e triagem de pacientes admitidos nas unidades de terapia intensiva do Brasil: estudo transversal do tipo survey nacional. Rev Bras Ter Intensiva 2022;34:484-91. [Crossref] [PubMed]
  10. Serapioni M, Tesser CD. O Sistema de Saúde brasileiro ante a tipologia internacional: uma discussão prospectiva e inevitável. Saúde debate 2019;43:44-57. [Crossref]
  11. Agência Fiocruz de Notícias. Estudo revela como a pandemia afetou os atendimentos no SUS. 2021 [cited 2024 Oct 7]. Available online: https://agencia.fiocruz.br/estudo-revela-como-pandemia-afetou-os-atendimentos-no-sus
  12. Araz OM, Olson D, Ramirez-Nafarrate A. Predictive analytics for hospital admissions from the emergency department using triage information. International Journal of Production Economics 2019;208:199-207. [Crossref]
  13. Izquierdo JL, Ancochea JSavana COVID-19 Research Group, et al. Clinical Characteristics and Prognostic Factors for Intensive Care Unit Admission of Patients With COVID-19: Retrospective Study Using Machine Learning and Natural Language Processing. J Med Internet Res 2020;22:e21801. [Crossref] [PubMed]
  14. Huesch MD, Mosher TJ. Using It or Losing It? The Case for Data Scientists Inside Health Care. Open Health News 2017 [cited 2024 Oct 7]. Available online: https://www.openhealthnews.com/news-clipping/2017-05-04/using-it-or-losing-it-case-data-scientists-inside-health-care
  15. Cuocolo R, Caruso M, Perillo T, et al. Machine Learning in oncology: A clinical appraisal. Cancer Lett 2020;481:55-62. [Crossref] [PubMed]
  16. Cichosz SL, Bender C, Hejlesen O. A Comparative Analysis of Machine Learning Models for the Detection of Undiagnosed Diabetes Patients. Diabetology 2024;5:1-11. [Crossref]
  17. Fabrizzio GC, Erdmann AL, Oliveira LM. Web App for prediction of hospitalisation in Intensive Care Unit by covid-19. Rev Bras Enferm 2023;76:e20220740. [Crossref] [PubMed]
doi: 10.21037/jmai-24-390
Cite this article as: Grolli RE, Fabrizzio GC, Cabral EC, Souza JMD, Vieira LDS, Zambeli AS, Silva GM, Machado M, Bornia LG, Werner SS, Carvalho JT, Fileto R, Zibetti AW. Predictive tools in intensive care unit management: a protocol of a scoping review. J Med Artif Intell 2025;8:64.

Article Options

Download Citation

Share