AI-Model Enhance Clinical Trial Searches by 50%

Coronavirus clinical studies seek participants for vaccines and therapeutics
web network in an AI environment

In response to the global COVID-19 disease emergency, about 500 clinical studies have been registered at the various clinical trial registry sites. 

And, over 300 studies are currently enrolling participants, reported The Lancet on April 24, 2020. 

Most of these studies offer comparative efficacy data versus standard of care, according to local COVID-19 treatment guidelines. Trials for COVID-19 are then mapped according to geographical, trial, patient, and intervention characteristics.

Given the accelerated rate at which trial information and findings are emerging, an urgent need exists to track clinical trials, avoid unnecessary duplication of efforts, and understand what trials are being done and where. 

In response, we have developed a COVID-19 clinical trial registry to collate all trials. Data are pulled from the International Clinical Trials Registry Platform, including those from the Chinese Clinical Trial Registry, ClinicalTrials.gov, Clinical Research Information Service - Republic of Korea, EU Clinical Trials Register, ISRCTN, Iranian Registry of Clinical Trials, Japan Primary Registries Network, and German Clinical Trials Register. 

Both automated and manual searches are done to ensure the minimization of duplicated entries and for appropriateness to the research questions. 

Identified studies are then manually reviewed by two separate reviewers before being entered into the registry. 

Concurrently, we have developed an artificial intelligence (AI)-based method for data searches to identify potential clinical studies, not captured in trial registries. 

These methods provide estimates of the likelihood of the importance of a study being included in our database, such that the study can then be reviewed manually for inclusion. 

The use of AI-based methods saves 50–80% of the time required to manually review all entries without loss of accuracy. 

Finally, we will use content aggregator services, such as LitCovid, to ensure our data acquisition strategy is complete. 

With this three-step process, the probability of missing important publications is greatly diminished and so the resulting data are representative of global COVID-19 research efforts.

These researchers declared no competing interests.

Precision Vaccinations publishes vaccine development news related to the coronavirus pandemic.