Article type
Year
Abstract
Background:
The Epistemonikos Database of Systematic Reviews is the largest repository of health-related systematic reviews. It combines a systematic approach, large-scale human collaboration and diverse technologies. Recognizing the significance of randomized trials in health decision-making, we are employing a similar approach to establish an extensive, freely accessible trial database.
Objectives:
To describe the methods used to identify relevant trials and to present the number of references retrieved by manual identification of trials included in systematic reviews. Additionally, to provide the percentage of records annotated by humans and by automated classifiers.
Methods:
To develop and maintain the database, we regularly screen multiple electronic sources, including electronic databases, trial registries, and preprint servers on a regular basis.
Additionally, we incorporate randomized trials included in the systematic reviews identified in the primary Epistemonikos database, among other methodologies.
Retrieved records undergo an automatic deduplication process. Utilizing various technologies, such as artificial intelligence and other machine-based methods, we identify potentially relevant trials. Subsequently, these trials are validated by a network of human collaborators.
Results:
The database is still in the development phase, with an expected release in 2024. As of September 2023, we have identified 862,392 trials through manual or automated methods. Out of these, 96,064 records correspond to trials included in systematic reviews in the Epistemonikos database.
Most records identified until now have only been classified by automated methods (97.9%). The validation is undergoing and up to now 18,156 records have been validated by a human.
Conclusions:
The Epistemonikos Database of Trials is an ongoing project with the goal of creating a comprehensive, reliable, and freely accessible database of randomized trials pertinent to healthcare. The expected release date is 2024.
Patient, public, and/or healthcare consumer involvement: none.
The Epistemonikos Database of Systematic Reviews is the largest repository of health-related systematic reviews. It combines a systematic approach, large-scale human collaboration and diverse technologies. Recognizing the significance of randomized trials in health decision-making, we are employing a similar approach to establish an extensive, freely accessible trial database.
Objectives:
To describe the methods used to identify relevant trials and to present the number of references retrieved by manual identification of trials included in systematic reviews. Additionally, to provide the percentage of records annotated by humans and by automated classifiers.
Methods:
To develop and maintain the database, we regularly screen multiple electronic sources, including electronic databases, trial registries, and preprint servers on a regular basis.
Additionally, we incorporate randomized trials included in the systematic reviews identified in the primary Epistemonikos database, among other methodologies.
Retrieved records undergo an automatic deduplication process. Utilizing various technologies, such as artificial intelligence and other machine-based methods, we identify potentially relevant trials. Subsequently, these trials are validated by a network of human collaborators.
Results:
The database is still in the development phase, with an expected release in 2024. As of September 2023, we have identified 862,392 trials through manual or automated methods. Out of these, 96,064 records correspond to trials included in systematic reviews in the Epistemonikos database.
Most records identified until now have only been classified by automated methods (97.9%). The validation is undergoing and up to now 18,156 records have been validated by a human.
Conclusions:
The Epistemonikos Database of Trials is an ongoing project with the goal of creating a comprehensive, reliable, and freely accessible database of randomized trials pertinent to healthcare. The expected release date is 2024.
Patient, public, and/or healthcare consumer involvement: none.