Data-driven trends in critical care informatics: a bibliometric analysis of global collaborations using the MIMIC database (2004-2024)

Abstract

Background: The Medical Information Mart for Intensive Care (MIMIC) database has become a cornerstone resource for critical care research, enabling advances in outcome prediction, machine learning, and patient management. However, comprehensive bibliometric understanding of MIMIC-related research evolution, global collaboration, and thematic trends remains limited.

Objective: This study aimed to perform a comprehensive bibliometric analysis of MIMIC-related publications (2004-2024), identifying thematic evolution, global research trends, and emerging areas, to guide future research directions.

Methods: We conducted a bibliometric analysis of 2769 MIMIC-related publications indexed in the Web of Science Core Collection. Eligible peer-reviewed articles and reviews in English were screened through a dual-blinded process. Bibliometric analyses were performed using multiple software tools: R (v4.4.3) with RStudio (v2024.12.1 + 563) for data cleaning, disambiguation, and visualization; Bibliometrix (v4.3.2) for metadata extraction, descriptive statistics, and science mapping; VOSviewer (v1.6.20) for keyword co-occurrence, clustering, and citation network analyses; and Pajek (v6.01) for large-scale network visualization and layout optimization. A multi-step disambiguation strategy was applied to ensure data consistency in author and institution names. Citation metrics, thematic clustering, temporal keyword trends, and collaboration networks were comprehensively assessed to elucidate research dynamics.

Results: Among the 2769 analyzed publications, 2747 (99.2 %) were peer-reviewed original research articles. The average annual publication growth rate was 40.6 %, with an average citation rate of 11.29 per article. Publication trends showed three phases: slow growth (2004-2015), rapid expansion (2016-2020) following updated MIMIC dataset releases, and sustained momentum (2021-2024). Major journals publishing MIMIC research included Scientific Reports, Frontiers in Medicine, Frontiers in Cardiovascular, and others. China was the most productive country with 1998 publications, led by institutions such as Zhejiang University, Jinan University, and Wenzhou Medical University; however, its international collaboration rate was relatively low. In contrast, the United States demonstrated strong global influence, dominating highly cited publications and fostering extensive international collaborations. Thematic clustering and keyword co-occurrence analysis revealed an evolution in MIMIC-based research, transitioning from early descriptive studies to increasingly sophisticated applications of machine learning (ML) and artificial intelligence (AI). Foundational highly cited articles from US institutions highlighted the pivotal role of deep learning models and open-access ICU databases in critical care informatics.

Conclusions: MIMIC-based research has grown substantially, with China and the U.S. leading in output and impact. Studies have evolved from descriptive analyses to advanced AI applications, yet real-world integration remains limited by issues like single-center data and model opacity. Addressing these gaps will require transparent, clinically relevant models and stronger cross-national collaboration. Aligning technical innovation with ethical and practical considerations will enhance the translational value of MIMIC research in critical care.

Keywords: Bibliometric analysis; Critical care informatics; Data visualization; Health informatic; MIMIC database.