Feasibility of EHR-Based Prescreening for an Adjuvant Breast Cancer Trial in a Safety-Net Health System

Racial and ethnic inequities persist in oncology clinical trial participation, while safety-net health systems may face resource and data limitations that complicate candidate identification. We evaluated the feasibility of an electronic health record (EHR)-assisted prescreening workflow for EMBER-4, a phase 3 adjuvant breast cancer trial, at a safety-net health system.

Structured Query Language (SQL) queries applied to Epic Clarity identified potential candidates using diagnosis codes, receptor-status codes, endocrine therapy medication records, recent oncology follow-up, and appointment data. Research staff then manually reviewed pathology reports, treatment history, comorbidities, and other protocol criteria to determine full eligibility.

The source population with a diagnosis of breast cancer included 829 patients: 29% Hispanic, 30% Black, 31% non-Hispanic White and 9.7% others. Primary language was English in 75% and Spanish in 18%. The SQL query identified 123 potential candidates; manual review confirmed 18 eligible patients comprising 67% Hispanic and 44% Spanish language, yielding a positive predictive value of 15%. Clinicians identified 3 additional eligible patients missed by the query, for 21 known eligible patients overall. Eleven declined participation, one did not attend the scheduled clinic visit, and 9 consented. One patient withdrew before randomization and 8 were enrolled. Six enrolled patients were Hispanic, 1 was non-Hispanic White and 1 was Black.

In a safety-net oncology setting, EHR-assisted prescreening was feasible but required manual verification and clinician involvement. The workflow identified most known eligible patients, although incomplete structured data and low positive predictive value limited its utility as a standalone screening method.