JMIR Publications

ABSTRACT

Background:

Clinical research in intensive care has been integral in helping us better understand the complexity of a myriad of diseases and treatments. Open-access databases have become increasingly popular and allow us to utilize machine learning (ML) techniques to gain insights.

Objective:

Our aim is to identify the knowledge gaps between the two and find ways to utilize them in a complementary fashion in future intensive care research by comparing traditional intensive care research to those using open-access databases.

Methods:

This study uses ML for the analysis of publications in the Web of Science (WoS) database. Articles were categorized into "Open-Access Database" (OAD) and "Traditional Intensive Care" (TIC) studies. OAD studies included MIMIC, eICU Collaborative Research Database, AmsterdamUMCdb, HiRID, and Pediatric Intensive Care database. TIC studies included all other intensive care studies that did not fall under OAD. Uniform Manifold Approximation and Projection was used to visualize the corpus distribution. The BERTopic model was used to generate 30 topic unique identification numbers and subsequently categorized into 22 topic families. These were analyzed for the differences in distribution between OAD and TIC studies.

Results:

A total of 145,426 TIC studies and 1,301 OAD studies were identified. TIC studies have been exponentially increasing in publication volume over the last two decades, while OAD studies have shown a steady rise since their introduction in 2003. The TIC studies have more even coverage of topics while OAD studies have highly skewed topic distributions such as predictive model and sepsis.

Conclusions:

This study demonstrated that while OADs offer an abundance of information that can be used to create predictive models with larger cohorts, certain aspects of research, such as ethics and human psychology, cannot be easily done with OADs. While OAD research is expected to continue to grow, it will unlikely replace TIC studies but rather should be used to complement and supplement them.