AI Solution Reduces Clinical Trial Screening Time by 34%

A computerized solution that used artificial intelligence (AI) reduced the time for clinical trial patient screening by 34%, according to the findings of a study published in JMIR Medical Informatics.

The system, called the Automated Clinical Trial Eligibility Scanner (ACTES), improved enrollment by 11.1%. What’s more, use of ACTES improved the number of patients screened by 14.7% and those approached by 11.1%.

The research team designed ACTES to streamline the inefficient clinical trial recruiting process, which does not always find enough qualified candidates.

“Because of the large volume of data documented in EHRs, the recruiting processes used now to find relevant information are very labor-intensive within the short time frame needed,” said lead study investigator Yizhao Ni, Ph.D., from the biomedical informatics division at Cincinnati Children’s Hospital Medical Center. “By leveraging natural language processing and machine-learning technologies, ACTES was able to quickly analyze different types of data and automatically determine patients’ suitability for clinical trials.”

Researchers aimed to evaluate the impact of ACTES on institutional workflow. The research team believed using EHR-based automated screening would improve efficiency of patient identification, streamline recruitment workflow and increase clinical trial enrollment.

During the study period, the Cincinnati Children’s Hospital Medical Center emergency department had six clinical trials actively recruiting patients. Trials required review of structured data like demographics, vital signs and medications, or patients’ clinical conditions from unstructured narrative notes, including signs and symptoms.

The EHR recorded patient information as structured entries and unstructured clinical notes. ACTES identified relevant information from unstructured clinical notes using natural language processing technologies. The notes were tokenized and grouped. The system identified relevant phrases, like symptom-related keywords from the text, and extracted their medical concepts from clinical terminologies. Information retrieval algorithms then matched extracted terms and ranked patient candidates.

If the system deemed a patient eligible, the clinical research coordinator documented the patient’s eligibility and approached him or her before discharge. If a patient was ineligible, the clinical research coordinator documented why. The documentation was fed into ACTES in real time, so it could continue learning and adjust trial criteria.

To evaluate the effects of ACTES on improving patient screening efficiency, the research team tracked how a clinical research coordinator allocated their time during a 120-minute observation section at 30-second increments. The investigator also shadowed the research coordinator to observe the patient recruitment workflow.

The research coordinators completed post-evaluation usability surveys, which included the system usability scale and open-ended questions. The usability scale had 10 statements on a five-point agreement scale between strongly disagree and strongly agree.

Clinical research coordinators spent 38.5% of time on electronic screening without using ACTES. When ACTES was in place, the time reduced to 25.6%. Overall screening time decreased from 47.6% to 32.5% when the research coordinators used ACTES. The research coordinators used the saved time for other work-related activities, including waiting for sample collection and study-related administrative tasks.

Using ACTES resulted in more screened patients over all trials.

When research coordinators first started using ACTES, the total system usability scale score was a 67.9, which suggested an acceptable computerized solution. At the end of the study period, the score improved to 80, which suggested ACTES was a good computerized solution.

The researchers believe ACTES will provide significant benefits to nationwide research networks and healthcare institutions in executing research by harnessing real-time EHR data.

Get the best insights in digital health directly to your inbox.

Related

Risk Prediction Model Helps Identify Risk of Developing Lung Cancer

Digital Program and mHealth App Could Increase Smoking Cessation

Deep Learning Model Outperforms Clinical Model in Lung Cancer Survival Prediction