Oscar Perez-Concha
Current Organizations / Institutes
The organizations/institutes from the scholar's latest article
Previous Organizations / Institutes
The organizations / institutes from the scholar's other articles in the past ten years
Research Subjects
Oscar Perez-Concha has a total of 111 co-authors. Here are the top co-authors by number of shared publications:
- Jesús García
Computer Science and Engineering Department, Universidad Carlos III de Madrid, 28270 Colmenarejo, Spain
10 shared publications - Massimo Piccardi
Faculty of Engineering and IT, University of Technology Sydney
10 shared publications - Blanca Gallego
Centre for Big Data Research in Health, University of New South Wales , Randwick, NSW 2031
8 shared publications - Enrico Coiera
Centre for Health Informatics, Australian Institute of Health Innovation, Macquarie University, New South Wales, Australia
8 shared publications - Louisa Jorm
Centre for Big Data Research in Health, University of New South Wales, Sydney, New South Wales, Australia
7 shared publications - Jose Manuel Molina
Universidad Carlos III de Madrid
5 shared publications
Latest Publications
- Open Access
- 11 November 2026
- Published by JMIR Publications Inc. in JMIR Preprints
Protocol for a Feasibility Study of Computational Text Analysis of Child Protection Intake Reports in South Australia (Preprint)
UNSTRUCTURED Background: Child maltreatment statistics collected using surveys, in-depth interviews, and manual case reviews are limited in scope due to resource constraints and labour-intensive, non-scalable approaches. The advent of computational advancements such as natural language processing, text mining, and large language models opens up the possibility of rapidly analysing vast numbers of documents from child protection client information systems to generate useful information. However, often due to strict data access protocols, few research efforts have focused on the application of computational text analysis in the child maltreatment area. Objective: This study presents a protocol for the application and evaluation of a text mining method on 10,000 child protection intake reports from South Australia’s Department for Child Protection. Methods: The protocol is split into five phases: 1) Ethics, governance, and data access approvals; 2A) Establish gold standard for the automatic extraction of information from 100 redacted reports using manual expert annotation; 2B) Contextual de-identification to preserve relationships; 3) Text mining method development, validation and evaluation based on 300 reports; 4) Text mining application on large sample; and 5) Evaluation and data analysis. Discussion: This protocol describes a feasibility study applying computational text analysis to highly sensitive child protection reports to automatically extract key information. Key considerations and challenges included governance, ethics, security, and technical requirements across multiple stakeholders involved in the study and adapting the text mining methodology to address the loss of contextual information caused by redaction. Conclusion: Whilst a high level of trust between academics and the government sector and stringent data safety protocols are necessary, text mining child protection intake reports in Australia is possible with the potential to unlock valuable insights.
- 11 November 2026
- Published by JMIR Publications Inc. in JMIR Preprints
Open AccessProtocol for a Feasibility Study of Computational Text Analysis of Child Protection Intake Reports in South Australia (Preprint)
UNSTRUCTURED Background: Child maltreatment statistics collected using surveys, in-depth interviews, and manual case reviews are limited in scope due to resource constraints and labour-intensive, non-scalable approaches. The advent of computational advancements such as natural language processing, text mining, and large language models opens up the possibility of rapidly analysing vast numbers of documents from child protection client information systems to generate useful information. However, often due to strict data access protocols, few research efforts have focused on the application of computational text analysis in the child maltreatment area. Objective: This study presents a protocol for the application and evaluation of a text mining method on 10,000 child protection intake reports from South Australia’s Department for Child Protection. Methods: The protocol is split into five phases: 1) Ethics, governance, and data access approvals; 2A) Establish gold standard for the automatic extraction of information from 100 redacted reports using manual expert annotation; 2B) Contextual de-identification to preserve relationships; 3) Text mining method development, validation and evaluation based on 300 reports; 4) Text mining application on large sample; and 5) Evaluation and data analysis. Discussion: This protocol describes a feasibility study applying computational text analysis to highly sensitive child protection reports to automatically extract key information. Key considerations and challenges included governance, ethics, security, and technical requirements across multiple stakeholders involved in the study and adapting the text mining methodology to address the loss of contextual information caused by redaction. Conclusion: Whilst a high level of trust between academics and the government sector and stringent data safety protocols are necessary, text mining child protection intake reports in Australia is possible with the potential to unlock valuable insights.
- Open Access
- 24 July 2025
- Published by JMIR Publications Inc. in JMIR Pediatrics and Parenting
- Vol. 8, e73579
Leveraging AI to Investigate Child Maltreatment Text Narratives: Promising Benefits and Addressable Risks
The trove of information contained in child maltreatment narratives represents an opportunity to strengthen the evidence base for policy reform in this area, yet it remains underutilized by researchers and policy makers. Current research into child maltreatment often involves the use of qualitative methodologies or structured survey data that are either too broad or not representative, thereby limiting the development of effective policy responses and intervention strategies. Artificial intelligence (AI) approaches such as large language models (AI models that understand and generate language) can analyze large volumes of child maltreatment narratives by extracting population-level insights on factors of interest such as mental health and treatment needs. However, when applying such methods, it is useful to have a framework on which to base approaches to the data. We propose a seven step framework: (1) data governance; (2) researcher vetting; (3) data deidentification; (4) data access; (5) feasibility testing of baseline methods; (6) large-scale implementation of black box algorithms; and (7) domain expert result validation for such exercises to ensure careful execution and limit the risk of privacy and security breaches, bias, and unreliable conclusions.
- 24 July 2025
- Published by JMIR Publications Inc. in JMIR Pediatrics and Parenting
- Vol. 8, e73579
Open AccessLeveraging AI to Investigate Child Maltreatment Text Narratives: Promising Benefits and Addressable Risks
The trove of information contained in child maltreatment narratives represents an opportunity to strengthen the evidence base for policy reform in this area, yet it remains underutilized by researchers and policy makers. Current research into child maltreatment often involves the use of qualitative methodologies or structured survey data that are either too broad or not representative, thereby limiting the development of effective policy responses and intervention strategies. Artificial intelligence (AI) approaches such as large language models (AI models that understand and generate language) can analyze large volumes of child maltreatment narratives by extracting population-level insights on factors of interest such as mental health and treatment needs. However, when applying such methods, it is useful to have a framework on which to base approaches to the data. We propose a seven step framework: (1) data governance; (2) researcher vetting; (3) data deidentification; (4) data access; (5) feasibility testing of baseline methods; (6) large-scale implementation of black box algorithms; and (7) domain expert result validation for such exercises to ensure careful execution and limit the risk of privacy and security breaches, bias, and unreliable conclusions.
- Open Access
- 17 June 2025
- Published by Universitat Politecnica de Valencia
Teaching Health Data Science through Experiential Learning: The Datathon Approach
This paper describes the design and implementation of an annual health data science datathon aimed at teaching practical skills in health analytics to undergraduate and postgraduate university students. Datathons provide an experiential learning environment where students work in teams to analyse real-world health data and propose solutions to current challenges in the field. The steps to design and execute the datathon are outlined, covering data, partnerships, task design, and rubric development. We discuss the educational benefits of formalising research questions, working in groups, programming, and presenting findings. Growing access to electronic health data combined with the availability of free and opensource analysis platforms and libraries, and cheap cloud compute power make datathons a universally accessible learning experience.
- 17 June 2025
- Published by Universitat Politecnica de Valencia
Open AccessTeaching Health Data Science through Experiential Learning: The Datathon Approach
This paper describes the design and implementation of an annual health data science datathon aimed at teaching practical skills in health analytics to undergraduate and postgraduate university students. Datathons provide an experiential learning environment where students work in teams to analyse real-world health data and propose solutions to current challenges in the field. The steps to design and execute the datathon are outlined, covering data, partnerships, task design, and rubric development. We discuss the educational benefits of formalising research questions, working in groups, programming, and presenting findings. Growing access to electronic health data combined with the availability of free and opensource analysis platforms and libraries, and cheap cloud compute power make datathons a universally accessible learning experience.
- Open Access
- 6 March 2025
- Published by JMIR Publications Inc. in JMIR Preprints
Leveraging AI to Investigate Child Maltreatment Text Narratives: Promising Benefits and Addressable Risks (Preprint)
UNSTRUCTURED The use of child maltreatment text narratives represents an opportunity to strengthen the evidence base for child policy reform, yet they remain underutilized. Current research into child maltreatment often involves the use of structured data, potentially overlooking key details within the narratives that can be crucial to assist in developing effective interventions. Artificial Intelligence approaches such as large language models can assist in the large-scale analysis of child maltreatment narratives extracting useful population-level insights on indicators of mental health, and treatment needs. However, when applying such methods, a framework of five key steps can assist in their careful implementation to reduce risks of privacy and bias.
- 6 March 2025
- Published by JMIR Publications Inc. in JMIR Preprints
Open AccessLeveraging AI to Investigate Child Maltreatment Text Narratives: Promising Benefits and Addressable Risks (Preprint)
UNSTRUCTURED The use of child maltreatment text narratives represents an opportunity to strengthen the evidence base for child policy reform, yet they remain underutilized. Current research into child maltreatment often involves the use of structured data, potentially overlooking key details within the narratives that can be crucial to assist in developing effective interventions. Artificial Intelligence approaches such as large language models can assist in the large-scale analysis of child maltreatment narratives extracting useful population-level insights on indicators of mental health, and treatment needs. However, when applying such methods, a framework of five key steps can assist in their careful implementation to reduce risks of privacy and bias.
- Open Access
- 24 September 2024
- Published by IOS Press
- Vol. 318, 24-29
Adapting Large Language Models for Automated Summarisation of Electronic Medical Records in Clinical Coding
Encapsulating a patient’s clinical narrative into a condensed, informative summary is indispensable to clinical coding. The intricate nature of the clinical text makes the summarisation process challenging for clinical coders. Recent developments in large language models (LLMs) have shown promising performance in clinical text summarisation, particularly in radiology and echocardiographic reports, after adaptation to the clinical domain. To explore the summarisation potential of clinical domain adaptation of LLMs, a clinical text dataset, consisting of electronic medical records paired with “Brief Hospital Course” from the MIMIC-III database, was curated. Two open-source LLMs were then fine-tuned, one pre-trained on biomedical datasets and another on a general-content domain on the curated clinical dataset. The performance of the fine-tuned models against their base models were evaluated. The model pre-trained on biomedical data demonstrated superior performance after clinical domain adaptation. This finding highlights the potential benefits of adapting LLMs pre-trained on a related domain over a more generalised domain and suggests the possible role of clinically adapted LLMs as an assistive tool for clinical coders. Future work should explore adapting more advanced models to enhance model performance in higher-quality clinical datasets.
- 24 September 2024
- Published by IOS Press
- Vol. 318, 24-29
Open AccessAdapting Large Language Models for Automated Summarisation of Electronic Medical Records in Clinical Coding
Encapsulating a patient’s clinical narrative into a condensed, informative summary is indispensable to clinical coding. The intricate nature of the clinical text makes the summarisation process challenging for clinical coders. Recent developments in large language models (LLMs) have shown promising performance in clinical text summarisation, particularly in radiology and echocardiographic reports, after adaptation to the clinical domain. To explore the summarisation potential of clinical domain adaptation of LLMs, a clinical text dataset, consisting of electronic medical records paired with “Brief Hospital Course” from the MIMIC-III database, was curated. Two open-source LLMs were then fine-tuned, one pre-trained on biomedical datasets and another on a general-content domain on the curated clinical dataset. The performance of the fine-tuned models against their base models were evaluated. The model pre-trained on biomedical data demonstrated superior performance after clinical domain adaptation. This finding highlights the potential benefits of adapting LLMs pre-trained on a related domain over a more generalised domain and suggests the possible role of clinically adapted LLMs as an assistive tool for clinical coders. Future work should explore adapting more advanced models to enhance model performance in higher-quality clinical datasets.
- Open Access
- 16 January 2024
- Published by JMIR Publications Inc. in JMIR Medical Education
- Vol. 10, e51388
Enriching Data Science and Health Care Education: Application and Impact of Synthetic Data Sets Through the Health Gym Project
Large-scale medical data sets are vital for hands-on education in health data science but are often inaccessible due to privacy concerns. Addressing this gap, we developed the Health Gym project, a free and open-source platform designed to generate synthetic health data sets applicable to various areas of data science education, including machine learning, data visualization, and traditional statistical models. Initially, we generated 3 synthetic data sets for sepsis, acute hypotension, and antiretroviral therapy for HIV infection. This paper discusses the educational applications of Health Gym’s synthetic data sets. We illustrate this through their use in postgraduate health data science courses delivered by the University of New South Wales, Australia, and a Datathon event, involving academics, students, clinicians, and local health district professionals. We also include adaptable worked examples using our synthetic data sets, designed to enrich hands-on tutorial and workshop experiences. Although we highlight the potential of these data sets in advancing data science education and health care artificial intelligence, we also emphasize the need for continued research into the inherent limitations of synthetic data.
- 16 January 2024
- Published by JMIR Publications Inc. in JMIR Medical Education
- Vol. 10, e51388
Open AccessEnriching Data Science and Health Care Education: Application and Impact of Synthetic Data Sets Through the Health Gym Project
Large-scale medical data sets are vital for hands-on education in health data science but are often inaccessible due to privacy concerns. Addressing this gap, we developed the Health Gym project, a free and open-source platform designed to generate synthetic health data sets applicable to various areas of data science education, including machine learning, data visualization, and traditional statistical models. Initially, we generated 3 synthetic data sets for sepsis, acute hypotension, and antiretroviral therapy for HIV infection. This paper discusses the educational applications of Health Gym’s synthetic data sets. We illustrate this through their use in postgraduate health data science courses delivered by the University of New South Wales, Australia, and a Datathon event, involving academics, students, clinicians, and local health district professionals. We also include adaptable worked examples using our synthetic data sets, designed to enrich hands-on tutorial and workshop experiences. Although we highlight the potential of these data sets in advancing data science education and health care artificial intelligence, we also emphasize the need for continued research into the inherent limitations of synthetic data.
- Open Access
- 25 August 2023
- Published by JMIR Publications Inc. in Interactive Journal of Medical Research
- Vol. 12 (1) , e46322
Web-Based Application Based on Human-in-the-Loop Deep Learning for Deidentifying Free-Text Data in Electronic Medical Records: Development and Usability Study
Background: The narrative free-text data in electronic medical records (EMRs) contain valuable clinical information for analysis and research to inform better patient care. However, the release of free text for secondary use is hindered by concerns surrounding personally identifiable information (PII), as protecting individuals' privacy is paramount. Therefore, it is necessary to deidentify free text to remove PII. Manual deidentification is a time-consuming and labor-intensive process. Numerous automated deidentification approaches and systems have been attempted to overcome this challenge over the past decade. Objective: We sought to develop an accurate, web-based system deidentifying free text (DEFT), which can be readily and easily adopted in real-world settings for deidentification of free text in EMRs. The system has several key features including a simple and task-focused web user interface, customized PII types, use of a state-of-the-art deep learning model for tagging PII from free text, preannotation by an interactive learning loop, rapid manual annotation with autosave, support for project management and team collaboration, user access control, and central data storage. Methods: DEFT comprises frontend and backend modules and communicates with central data storage through a filesystem path access. The frontend web user interface provides end users with a user-friendly workspace for managing and annotating free text. The backend module processes the requests from the frontend and performs relevant persistence operations. DEFT manages the deidentification workflow as a project, which can contain one or more data sets. Customized PII types and user access control can also be configured. The deep learning model is based on a Bidirectional Long Short-Term Memory-Conditional Random Field (BiLSTM-CRF) with RoBERTa as the word embedding layer. The interactive learning loop is further integrated into DEFT to speed up the deidentification process and increase its performance over time. Results: DEFT has many advantages over existing deidentification systems in terms of its support for project management, user access control, data management, and an interactive learning process. Experimental results from DEFT on the 2014 i2b2 data set obtained the highest performance compared to 5 benchmark models in terms of microaverage strict entity–level recall and F1-scores of 0.9563 and 0.9627, respectively. In a real-world use case of deidentifying clinical notes, extracted from 1 referral hospital in Sydney, New South Wales, Australia, DEFT achieved a high microaverage strict entity–level F1-score of 0.9507 on a corpus of 600 annotated clinical notes. Moreover, the manual annotation process with preannotation demonstrated a 43% increase in work efficiency compared to the process without preannotation. Conclusions: DEFT is designed for health domain researchers and data custodians to easily deidentify free text in EMRs. DEFT supports an interactive learning loop and end users with minimal technical knowledge can perform the deidentification work with only a shallow learning curve.
- 25 August 2023
- Published by JMIR Publications Inc. in Interactive Journal of Medical Research
- Vol. 12 (1) , e46322
Open AccessWeb-Based Application Based on Human-in-the-Loop Deep Learning for Deidentifying Free-Text Data in Electronic Medical Records: Development and Usability Study
Background: The narrative free-text data in electronic medical records (EMRs) contain valuable clinical information for analysis and research to inform better patient care. However, the release of free text for secondary use is hindered by concerns surrounding personally identifiable information (PII), as protecting individuals' privacy is paramount. Therefore, it is necessary to deidentify free text to remove PII. Manual deidentification is a time-consuming and labor-intensive process. Numerous automated deidentification approaches and systems have been attempted to overcome this challenge over the past decade. Objective: We sought to develop an accurate, web-based system deidentifying free text (DEFT), which can be readily and easily adopted in real-world settings for deidentification of free text in EMRs. The system has several key features including a simple and task-focused web user interface, customized PII types, use of a state-of-the-art deep learning model for tagging PII from free text, preannotation by an interactive learning loop, rapid manual annotation with autosave, support for project management and team collaboration, user access control, and central data storage. Methods: DEFT comprises frontend and backend modules and communicates with central data storage through a filesystem path access. The frontend web user interface provides end users with a user-friendly workspace for managing and annotating free text. The backend module processes the requests from the frontend and performs relevant persistence operations. DEFT manages the deidentification workflow as a project, which can contain one or more data sets. Customized PII types and user access control can also be configured. The deep learning model is based on a Bidirectional Long Short-Term Memory-Conditional Random Field (BiLSTM-CRF) with RoBERTa as the word embedding layer. The interactive learning loop is further integrated into DEFT to speed up the deidentification process and increase its performance over time. Results: DEFT has many advantages over existing deidentification systems in terms of its support for project management, user access control, data management, and an interactive learning process. Experimental results from DEFT on the 2014 i2b2 data set obtained the highest performance compared to 5 benchmark models in terms of microaverage strict entity–level recall and F1-scores of 0.9563 and 0.9627, respectively. In a real-world use case of deidentifying clinical notes, extracted from 1 referral hospital in Sydney, New South Wales, Australia, DEFT achieved a high microaverage strict entity–level F1-score of 0.9507 on a corpus of 600 annotated clinical notes. Moreover, the manual annotation process with preannotation demonstrated a 43% increase in work efficiency compared to the process without preannotation. Conclusions: DEFT is designed for health domain researchers and data custodians to easily deidentify free text in EMRs. DEFT supports an interactive learning loop and end users with minimal technical knowledge can perform the deidentification work with only a shallow learning curve.
- Open Access
- 30 July 2023
- Published by JMIR Publications Inc. in JMIR Preprints
Enriching Data Science and Health Care Education: Application and Impact of Synthetic Data Sets Through the Health Gym Project (Preprint)
UNSTRUCTURED Large-scale medical data sets are vital for hands-on education in health data science but are often inaccessible due to privacy concerns. Addressing this gap, we developed the Health Gym project, a free and open-source platform designed to generate synthetic health data sets applicable to various areas of data science education, including machine learning, data visualization, and traditional statistical models. Initially, we generated 3 synthetic data sets for sepsis, acute hypotension, and antiretroviral therapy for HIV infection. This paper discusses the educational applications of Health Gym’s synthetic data sets. We illustrate this through their use in postgraduate health data science courses delivered by the University of New South Wales, Australia, and a Datathon event, involving academics, students, clinicians, and local health district professionals. We also include adaptable worked examples using our synthetic data sets, designed to enrich hands-on tutorial and workshop experiences. Although we highlight the potential of these data sets in advancing data science education and health care artificial intelligence, we also emphasize the need for continued research into the inherent limitations of synthetic data.
- 30 July 2023
- Published by JMIR Publications Inc. in JMIR Preprints
Open AccessEnriching Data Science and Health Care Education: Application and Impact of Synthetic Data Sets Through the Health Gym Project (Preprint)
UNSTRUCTURED Large-scale medical data sets are vital for hands-on education in health data science but are often inaccessible due to privacy concerns. Addressing this gap, we developed the Health Gym project, a free and open-source platform designed to generate synthetic health data sets applicable to various areas of data science education, including machine learning, data visualization, and traditional statistical models. Initially, we generated 3 synthetic data sets for sepsis, acute hypotension, and antiretroviral therapy for HIV infection. This paper discusses the educational applications of Health Gym’s synthetic data sets. We illustrate this through their use in postgraduate health data science courses delivered by the University of New South Wales, Australia, and a Datathon event, involving academics, students, clinicians, and local health district professionals. We also include adaptable worked examples using our synthetic data sets, designed to enrich hands-on tutorial and workshop experiences. Although we highlight the potential of these data sets in advancing data science education and health care artificial intelligence, we also emphasize the need for continued research into the inherent limitations of synthetic data.
- Open Access
The relationship between hyperglycaemia on admission and patient outcome is modified by hyperlactatemia and diabetic status: a retrospective analysis of the eICU collaborative research database
Both blood glucose and lactate are well-known predictors of organ dysfunction and mortality in critically ill patients. Previous research has shown that concurrent adjustment for glucose and lactate modifies the relationship between these variables and patient outcomes, including blunting of the association between blood glucose and patient outcome. We aim to investigate the relationship between ICU admission blood glucose and hospital mortality while accounting for lactate and diabetic status. Across 43,250 ICU admissions, weighted to account for missing data, we assessed the predictive ability of several logistic regression and generalised additive models that included blood glucose, blood lactate and diabetic status. We found that inclusion of blood glucose marginally improved predictive performance in all patients: AUC-ROC 0.665 vs 0.659, with a greater degree of improvement seen in non- diabetics: AUC-ROC 0.675 vs 0.663. Inspection of the estimated risk profiles revealed the standard U-shaped risk profile for blood glucose was only present in non-diabetic patients after controlling for blood lactate levels. Future research should aim to utilise observational data to estimate whether interventions such as insulin further modify this effect, with the goal of informing future RCTs of interventions targeting glycaemic control in the ICU.
Open AccessThe relationship between hyperglycaemia on admission and patient outcome is modified by hyperlactatemia and diabetic status: a retrospective analysis of the eICU collaborative research database
Both blood glucose and lactate are well-known predictors of organ dysfunction and mortality in critically ill patients. Previous research has shown that concurrent adjustment for glucose and lactate modifies the relationship between these variables and patient outcomes, including blunting of the association between blood glucose and patient outcome. We aim to investigate the relationship between ICU admission blood glucose and hospital mortality while accounting for lactate and diabetic status. Across 43,250 ICU admissions, weighted to account for missing data, we assessed the predictive ability of several logistic regression and generalised additive models that included blood glucose, blood lactate and diabetic status. We found that inclusion of blood glucose marginally improved predictive performance in all patients: AUC-ROC 0.665 vs 0.659, with a greater degree of improvement seen in non- diabetics: AUC-ROC 0.675 vs 0.663. Inspection of the estimated risk profiles revealed the standard U-shaped risk profile for blood glucose was only present in non-diabetic patients after controlling for blood lactate levels. Future research should aim to utilise observational data to estimate whether interventions such as insulin further modify this effect, with the goal of informing future RCTs of interventions targeting glycaemic control in the ICU.
- Open Access
Curation and description of a blood glucose management and nutritional support cohort using the eICU collaborative research database
Freely available electronic medical record (EMR) data collections have transformed data science and observational research in critical care medicine. Descriptive characterisation of these data collections can aid in highlighting variation in clinical practice and patient outcomes across Intensive Care Units (ICUs). Glycaemic control and nutritional management are important aspects of patient management in the ICU. Blood glucose on admission has a well-known U-shaped relationship with mortality and morbidity, with both hypo- and hyper-glycemia being associated with poor patient outcomes. The importance of nutritional support has been highlighted in critical care guidelines. However, both areas have open research questions and highly variable clinical practices that observational data may help highlight and inform. To aid in this research, we curated a database of patients using the eICU collaborative research data (eICU-CRD), which we describe in the current paper, focusing on patient blood glucose, insulin therapy and enteral nutrition. The eICU-CRD is derived from a telehealth EMR covering 208 United States hospitals from 2014-2015. In addition to descriptive statistics and graphical analysis, we highlight any limitations in data quality. Our results are in line with previous research suggesting the eICU-CRD cohort is of lower illness severity than the average ICU patient cohort and so receive less invasive interventions. Examinations of data missingness revealed issues with medication orders and non-reporting of nutrition by several hospitals. Overall, with care around missingness we believe the eICU-CRD to be a valuable resource in evidence generation for critical care research.
Open AccessCuration and description of a blood glucose management and nutritional support cohort using the eICU collaborative research database
Freely available electronic medical record (EMR) data collections have transformed data science and observational research in critical care medicine. Descriptive characterisation of these data collections can aid in highlighting variation in clinical practice and patient outcomes across Intensive Care Units (ICUs). Glycaemic control and nutritional management are important aspects of patient management in the ICU. Blood glucose on admission has a well-known U-shaped relationship with mortality and morbidity, with both hypo- and hyper-glycemia being associated with poor patient outcomes. The importance of nutritional support has been highlighted in critical care guidelines. However, both areas have open research questions and highly variable clinical practices that observational data may help highlight and inform. To aid in this research, we curated a database of patients using the eICU collaborative research data (eICU-CRD), which we describe in the current paper, focusing on patient blood glucose, insulin therapy and enteral nutrition. The eICU-CRD is derived from a telehealth EMR covering 208 United States hospitals from 2014-2015. In addition to descriptive statistics and graphical analysis, we highlight any limitations in data quality. Our results are in line with previous research suggesting the eICU-CRD cohort is of lower illness severity than the average ICU patient cohort and so receive less invasive interventions. Examinations of data missingness revealed issues with medication orders and non-reporting of nutrition by several hospitals. Overall, with care around missingness we believe the eICU-CRD to be a valuable resource in evidence generation for critical care research.