Accelerating Clinical Trials Through Automation and AI
The recruitment phase for clinical trials takes an average of 18 months, and nearly 20% of cancer trials fail because of low accrual rates. Automation and AI can improve this process — and help bring lifesaving treatments to patients faster — by identifying and recruiting eligible participants.
Robotic process automation tools can be especially beneficial in this area by assessing patient records and matching them to appropriate trials.
“Medical abstraction can be tedious and expensive. In clinical trial matching, structuring trial eligibility is easy, whereas structuring patient records is the real bottleneck,” says Poon. He cites Microsoft’s Healthcare Agent Orchestrator as an example of “how RPA can potentially unlock massive productivity gains by introducing agents to automate information gathering, normalization, integration and clinical trial matching scenarios.”
Intelligent document processing tools are also proving beneficial. IDP can help research teams avoid manual errors, improve patient data accuracy, and more efficiently analyze massive volume sets. Amazon Web Services points out that when powered by large language models, IDPs can generate reports and uncover actionable insights.
The TrialGPT algorithm, developed at the National Institutes of Health, is an example of this type of technology. In a pilot study, researchers found that when assessing patients for trial eligibility, TrialGPT spent 40% less time on screening but achieved the same level of accuracy as human clinicians. TrialGPT also created summaries explaining why a patient was a good fit for a trial.
EXPLORE: Here are 13 ways AI enhances healthcare operations, patient care and treatments.
The Role of Cloud and Advanced Analytics in Drug Discovery
“We and others have already used AI systems to generate promising drug candidates, and I expect such successes to rapidly accumulate in the next few years,” Poon says. “We can shrink the time for target identification, lead compound identification and optimization.”
Researchers say the drug discovery phase, which typically takes three to six years and accounts for about 35% of the total cost of developing a new treatment, can be shortened by one or two years with AI. That’s because AI can identify and test the effects of different compounds faster than a human can.
Advanced data analytics are critical for this type of analysis. AI-powered algorithms can analyze and compare massive amounts of information across multiple databases to identify which combinations will be most effective in creating a new drug.
This type of work wouldn’t be possible without cloud computing and storage. While on-premises data centers have a defined amount of space, the cloud gives life sciences organizations the unlimited scale they may need to manage and analyze these large data sets.
The cloud also allows organizations to adjust their storage capabilities — and therefore, better control costs — by partnering with vendors for access to powerful graphics processing units and CPUs.
“Let’s say you’re working in a Google Cloud environment and using their high-performance computing to run protein folding scenarios,” says Joe Miles, industry director of life sciences at UiPath. “You can then take that information and route it to appropriate repositories as it pertains to an individual trial.”
