Improving patient pre-screening for clinical trials with generative AI on AWS

Clinicians face several challenges when finding and enrolling patients in clinical trials. These include lack of awareness, limited access to trial information, especially for rare diseases, and time constraints. Typically, clinicians undertake a set of steps to find and evaluate clinical trials for their patients that include identification of patients who would benefit from a clinical trial, collecting relevant medical information about patient diagnosis, stage, and treatment history, searching clinical trial registries and institutional review boards, reviewing eligibility criteria, and more. One solution to streamline this process is to automate pre-screening by applying artificial intelligence.

In this post, we present a generative AI driven solution leveraging AWS services to facilitate an improved turnaround time for matching cancer patients to relevant clinical trials. A single cancer trial may have numerous inclusion and exclusion criteria, each of which needs to be checked against the patient’s specific medical history and current condition. By automating this process with generative AI, the screening time was reduced by ~ 40% in a real-life clinical trial matching task.

This solution leverages Amazon Bedrock knowledge base and Amazon Bedrock agents to find relevant trials using a patient’s medical condition and demographic data like age, gender, country, etc. By establishing matched trials based on inclusion and exclusion criteria, the solution incorporates chain of thought reasoning and explains the reason for selecting or not selecting a clinical trial. This solution also highlights the key differences between the recommended matches, enabling clinicians to compare and contrast the available options more effectively.

Amazon Bedrock is a fully managed service that offers a choice of high-performing foundation models (FMs) through a single API. It provides a broad set of capabilities needed to build generative AI applications with security, privacy, and responsible AI. Amazon Bedrock agents act as orchestrators, managing interactions between foundational models, data sources, knowledge bases, software applications, and user conversations. Additionally, they automate API calls to execute actions and access knowledge bases to enrich information relevant to these actions.

Clinical trial collection and matching workflow diagram

Architecture diagram

The architecture diagram illustrates the end-to-end workflow for matching cancer patients to relevant clinical trials using AWS services like Amazon Bedrock Agents, Knowledge Base, Amazon OpenSearch Service, and AWS Lambda functions. The clinician initiates the process by querying the clinical trials agent with the patient’s details, then fetches the medical history, leverages knowledge bases, and invokes different APIs to filter and recommend suitable trials based on the inclusion and exclusion criteria.

An agent-based framework on Amazon Bedrock allows developers to model and simulate complex systems by executing multistep tasks across different systems and data sources. The entire workflow spans multiple tasks starting from 1. Fetching patient relevant historical information 2. Querying clinical trials registry to fetch clinical trials relevant to patient current condition and matching the trials related inclusion & exclusion criteria to patient historical attributes to narrow down to specific clinical trials. This process enables automation of the clinical trial pre-screening process, which typically takes hours to a few minutes.

Next Steps

With the power of AWS services like Amazon Bedrock, healthcare providers can unlock new possibilities for streamlining clinical trial enrollment and accelerating the development of life-saving treatments. As generative AI continues to advance, its potential applications in the healthcare domain are vast, promising to revolutionize various aspects of patient care and medical research. We encourage you to explore the solution, adapt it to your specific needs, and unlock new possibilities in patient care and clinical processes. We value community collaboration and welcome your contributions – whether through submitting a pull request to enhance the solution’s functionality or reporting any issues you encounter via GitHub issues. Your feedback and involvement will help us continue to improve and strengthen this solution to better serve the healthcare community’s needs.