Algorithm May Improve Success of Clinical Trials
New method may help overcome barriers in patient recruitment to clinical trials.
A newly developed automated algorithm was found to predict patient participation in clinical trials significantly better than current recruitment practices, a recent study found.
The lack of participation in clinical trials is an ongoing challenge in the medical research field and can lead to comprised results or a discontinuation of the studies altogether.
A study published in the American Medical Informatics Association used machine learning technologies that predict whether or not a participant will enroll in a study.
“Challenges with patient recruitment for clinical trials are a major barrier to timely and efficient translational research,” said lead study author Yizhao Ni, PhD. “The ultimate goal of our research is to impact patient recruitment strategies to increase participation in clinical trials, and to help ensure that studies can be completed and the data are meaningful.”
Information from prior studies was incorporated into their research that identified objective and subjective factors that influence patient recruitment. Objective factors included age, race, education, socioeconomic level, financial resources, and required time commitment.
Subjective factors included attitudes towards medical research, family influence, seasonality, or whether a person's health condition has suddenly deteriorated.
During the study, researchers used the machine-learning algorithm to analyze, compare, and interpret different data sources to help predict specific patient decision making. Data was collected between 2010 through 2012 from clinical trial recruitment in the Emergency Department of Cincinnati Children’s Hospital Medical Center.
Since patient recruitment is now done on a per-patient-visit basis, researchers tried to match patients with appropriate clinical studies based on the study’s specific guidelines and goals before approaching patients for enrollment.
For the study, researchers collected data on the Emergency Department’s process and used scoring for each patient. Each patient invited to join the clinical trial was labeled as an “encounter.” Those who accepted the invitations were labeled as +1 and those who declined were given a -1.
The data included 3345 patient encounters for a set of 18 different clinical trials. Simultaneously, researchers collected demographic and socioeconomic data with clinical information from various sources to help build patient profiles.
The information was then fed into the automated algorithm machine and was processed through programs for predictive modeling, comparison, analysis, and prediction.
The effectiveness of the algorithm was compared to a random-response-prediction program that was developed to simulate the current recruiting method used in the Emergency Department. The results were confirmed by comparing them to the acceptance or decline responses that were recorded.
The results of the study found that approximately 60% of patients approached with the traditional recruitment process agreed to participate in the trial. Additionally, patients were less likely to participate in randomized studies, multi-center trials, more complex trials, and trials that required follow-up visits. However, researchers predict that the acceptance rates could rise to about 72% with the new automated algorithm.
As the algorithm continues to develop and tighten, researchers hope to reach their goal of an acceptance rate that surpasses 72%. Despite increasing information on why people accept or decline trial invitations, it is still difficult to manually process the information in busy medical environments, the researchers said.
The current practices are considered somewhat random and have a small amount of time and ability to account for individual patient preferences or biases. With an automated system that can both analyze and interpret these factors, it could lead to the development of precise patient-directed recruitment strategies that would help improve participation numbers.