Development and validation of an EHR data-based program to identify eligible RCT participants.

Abstract

338 Background: Identifying participants for time-sensitive clinical trials such as those with newly diagnosed cancer presents challenges. Manual identification methods, the current standard, can be resource intensive and present challenges with data collection, quality assurance and can introduce bias to the selection process. Electronic health record (EHR) data, designed to optimize clinical care, quality, and billing, may also streamline participant identification and facilitate recruitment. We aim to develop an EHR-based study eligibility program (SEP) to identify participants for recruitment to the Cancer Financial Experience (CAFÉ) randomized trial. Methods: CAFÉ is a NIH-funded randomized trial to test the impact of a novel financial navigation intervention on financial hardship and quality of life for cancer patients. Inclusion criteria: currently enrolled at Kaiser Permanente (KP) Northwest or KP Washington, age ≥18, had a visit to a CAFÉ-participating oncology clinic within last 14 days (defined as qualifying visit), diagnosed with invasive cancer (new or recurrent) within 120 days of qualifying visit, 6 months continuous enrollment prior to qualifying visit, English or Spanish speaker. Exclusion criteria: diagnosis of non-melanoma skin cancer, diagnosis of benign or in-situ tumors, hospice care within last 12 months, unable to complete baseline survey, household member enrolled in CAFÉ study. To assess the validity of the SEP for use in CAFÉ, we extracted EHR data on study inclusion/exclusion criteria for n = 125 patients across both study sites between 3/1/19–2/29/20 to determine eligibility status (eligible, ineligible). Manual EHR review was then conducted on the n = 125 patients to assess inclusion/exclusion criteria and determine eligibility. Results of SEP vs. manual review were compared, noting reasons for discrepancies. Positive predictive value (PPV), negative predictive value (NPV), sensitivity and specificity values were calculated. We herein report results of one site (n = 75), with next steps to complete analysis of the total n = 125 at both sites. Results: The SEP had a PPV of 76%, NPV of 96%, sensitivity of 97% and specificity of 67%, based on n = 75 patients. We identified areas of improvement for the SEP and groups to conduct manual validation during the CAFÉ trial. Potential reasons for lower specificity and lower PPV include: the SEP was identifying patients with a hematologic malignancy (e.g., lymphoma) and a recent visit to oncology, but whose diagnosis date was outside the eligible period; additionally, the SEP was identifying patients with a recent oncology visit, but who had a non-cancer diagnosis (e.g., monoclonal gammopathy of undetermined significance [MGUS]). Conclusions: Development of a validated EHR-based tool to rapidly identify cancer clinical trial participants offers a low resource option that may overcome challenges associated with manual chart review.