While the clinical research industry has developed innovative therapies and devices to serve patients all over the world, the industry has been more hesitant when adopting technology to improve research operations. Currently, machine learning and artificial intelligence is not widely applied to the research industry. However, Advarra is pushing to bring innovative approaches to the research community, and we’re starting with one of the costliest, most resource-consuming challenges facing research teams: zero and under-accruing protocols.
Using Machine Learning to Predict Accrual
Studies that do not accrue subjects at the site level, an estimated 20–50 % of all studies, waste valuable resources and staff time.¹ Studies ‘fail’ for a variety of reasons, whether it’s a poor catchment area for a particular disease, competing studies vying for the same population, the track record of a principal investigator, lack of site resources to support the study—the list goes on and on. When conducting a feasibility assessment and making the choice to activate a study at your site, it’s impossible to review all the information across all variables to determine if the study will accrue the estimated participants, if any at all.
While no one person can analyze and calculate all elements that may impact accrual, Advarra has leveraged machine learning and mathematical modeling to create a predictive model that can review hundreds of data points to inform a more confident decision about which studies to activate, and even predict how many participants each study is likely to enroll.
Introducing Advarra’s Insights Accrual Prediction Platform
Study feasibility, the process of reviewing logistical aspects of a research study prior to starting the activation process, is performed in many ways across the research community. Currently, feasibility assessments may include:
- Site questionnaires from sponsors,
- administrative reviews of technical expertise, equipment, and staffing,
- a budget analysis or,
- simply, “Dr. So-and-so wants to do a study.”
Committed to delivering innovative solutions to industry challenges, our experienced staff identified the potential value of a tool to support this process and sought to bring it to life. Throughout the course of her research career, Dr. Wendy Tate, Director of Research Operations at Advarra, set out to develop an objective method to predict clinical trial accrual based on previous experiences of the research organization. Last week, Advarra released the Insights Accrual Prediction platform to pilot research organizations. These pilot organizations will apply the tool to their typical feasibility assessment workflows for interventional treatment industry trials and will examine additional use cases to not only predict accrual, but to support difficult conversations with their investigators or sponsors, or to inform resource distribution to ensure a better accrual outcome.
To build trust and transparency in the model, each predictive output is accompanied by an accuracy table to show how the different models predicted a subset of test protocols and range of “confidence” in the prediction. Committed to optimizing research performance and empowering clinical sites, every output of the platform serves your organization’s decision-making and success.
Pushing the Expectations of Clinical Research Technology
At Advarra, we’re advancing clinical research to be safer, smarter, and faster™. We’re excited to bring advanced tactics like machine learning to our community to provide your organization the tools needed to conduct efficient and compliant research.
What’s even more exciting? This is only the beginning. We look forward to our collaboration with our pilot participants to refine this valuable and innovative platform and continue to look ahead towards applying advanced strategies like this to additional industry challenges. Return to our resource library next week to read more about the history behind the algorithm, and where Dr. Wendy Tate sees this technology headed next.
1. Tate WR, Cranmer LD. Developing a predictive model for cancer clinical trial accrual. J Clin Oncol. 2014;32(15S):6557. Tate WR, Abraham I, Cranmer LD.