Algorithms, data help ID clinical trial candidates

To increase participation in clinical drug trials, researchers at Cincinnati Children's Hospital Medical Center are looking to electronic health records.

The researchers created algorithms to comb through 3,000 clinical trial announcements and 1,655 clinical notes for medications and their attributes, such as dosage, form, and frequency, to identify patients eligible for a clinical trial.

Compared with the human-generated baseline, their two machine learning methods, binary classification and sequence labeling, were found significantly more accurate, according to a study published in Journal of the American Medical Informatics Association.

Though they're subject to alert fatigue, clinical trial alerts generated by electronic health records can be helpful in recruiting participants in clinical trials, a JAMIA article back in May concluded. Initially, physicians responded about 50 percent of the time. Though that rate declined over the 36-week study period, it remained between 30 and 40 percent--higher than expected.

In their efforts to improve ties between research and clinical practice, Keith Marsolo of the biomedical informatics division of Cincinnati Children's Hospital Medical Center previously wrote that the clinical informatics community must be involved in a redesign of electronic health record systems to encourage data-sharing between healthcare research and clinical practices. That article describes his many frustrations with having to develop workarounds to the EHR system's limitations.

Calling for better integration, he said that users must be able to create new data-collection forms and other interfaces that fit easily into clinical workflow.

A joint U.S. and British research group has developed an information architecture to support electronic practice-based research networks. The goal of that work has been to facilitate participation in clinical trials and other clinical research opportunities, including collection of aggregated data and the exchange of information between investigators and "clusters of independent data sites."

One of the challenges of that work has been the need to develop a common language allowing different EHR systems to share information for clinical studies.

To learn more:
- read the study


Fuel Top Line Growth and Increase Membership

In this webinar, payers can learn how to accelerate the sales pipeline and grow membership across all health insurance market segments using comprehensive sales technologies.