ACTIVATE: AI-driven Clinical-trial Trial-Information and Viability Assessment Tool for EHRs

AI-driven Clinical-trial Trial-Information and Viability Assessment Tool for EHRs (ACTIVATE)

This study aims to develop and evaluate ACTIVATE, an AI-driven tool for clinical trial information and viability assessment using electronic health records (EHRs). The project will leverage retrospective and prospective EHR data to build and validate algorithms that identify potentially eligible participants for clinical trials and facilitate trial matching.

ACTIVATE is a pragmatic health system intervention designed to improve clinical trial matching and accrual using AI-driven tools integrated with EHR data. The study will first retrospectively analyze data from approximately 70,000 participants who initiated new systemic therapy at Dana-Farber Cancer Institute since 2016 to develop and validate the MatchMiner-AI pipeline.

For the prospective evaluation, all DFCI patients' medical record numbers (MRNs) will be randomized into control and intervention groups. The intervention group will receive proactive notifications to treating oncologists when AI models detect progressive disease and a high probability of starting new treatment, including a ranked list of potential clinical trial options. The control group will continue with standard MatchMiner-AI workflows.

Inclusion Criteria:

• 3.1 The potentially eligible patient population includes any adult (≥18 years old) with a cancer diagnosis receiving care at DFCI. No direct patient recruitment will occur as part of this protocol; all data will be obtained retrospectively or prospectively from routine clinical documentation and electronic health records. TrialForecast will involve aggregate queries of this dataset for cohort size estimation. The randomized interventional component (TrialMatch) is a health system level email "nudge" to treating oncologists providing a list of clinical trial options for patients who have progressive disease based on their imaging reports as detected using our previously developed, validated, and deployed AI model for that purpose. 23-25 Secondary outcomes in our study will include oncologist satisfaction with information delivered via these pipelines. All DFCI oncologists at any DFCI-owned/operated site (Longwood, Chestnut Hill, and regional campus sites) will be eligible to use our pipeline and may receive notifications about clinical trial options for their patients. In 2024, there were approximately 593 such oncologists who had outpatient appointments with at least one patient. Clinicians will constitute study participants as well, since they will have the opportunity to provide feedback on our pipeline to be analyzed by the study team.
• 3.2 Our project will focus on adults with cancer treated at DFCI, as above. We will not have any mechanism for identifying, targeting, or excluding pregnant women or prisoners.

Exclusion Criteria:

• 3.2 Our project will focus on adults with cancer treated at DFCI, as above. We will not have any mechanism for identifying, targeting, or excluding pregnant women or prisoners.