463 - Facing Implicit Bias: Factors beyond Race and Ethnicity Impact Research Enrollment and Retention

Poster Number: 463
Publication Number: 463.367

Background: Retrospective studies demonstrate under- and over-representation of some groups of children in research.

Objective: This study sought to characterize participation of children at risk for health disparities (CRHD) in emergency and critical care (ECC) research. We hypothesized CRHD would have lower enrollment.

Design/Methods: We collected enrollment and demographic data for eligible patients in 17 ECC studies from 3/21-11/21. We defined CRHD as children meeting ≥1 National Institute for Minority Health and Disparities criteria (non-White race, Hispanic ethnicity, non-English speaking, public insurance, low child opportunity index (COI), rural ZIP code). We compared approach and study enrollment for CRHDs vs. non-CRHDs. Study-level factors including consent mechanism, intervention, and risk level were obtained from principal investigator surveys. Demographic data and potential for implicit bias (measured by presence of bias on any of three administered Harvard Implicit Association Tests-IATs) were collected from 11 clinical research coordinators (CRCs) conducting study approaches. Bivariate and multivariable generalized linear mixed model (GLMM) analyses were performed to identify predictors of approach and enrollment.

Results: Of 899 eligible patients, 708 (79%) were CRHD. Overall approach and enrollment rates were 77% and 55% respectively. APPROACH: In bivariate analyses (Table 1), higher approach rates were associated with private insurance vs. public, written consent v. waiver and female vs male CRC. In GLMM analyses, higher approach (Table 2) was associated with female CRC (OR=1.54, p=0.033). ENROLLMENT: In bivariate analyses, lower enrollment was associated with CRHD status, minoritized R/E, public insurance , low COI, consent waiver, interventional studies, and greater than minimal risk. In GLMM analyses we observed lower enrollment (Table 3) with potential for implicit bias favoring Whites, although this did not reach statistical significance. Likelihood of enrollment was associated with female CRC (OR=2.93, p < 0.001) and verbal consent (OR=5.08, p=0.02).

Conclusion(s): In adjusted models, approaches to, and enrollment of, children in ECC research were not associated with CRHD status, R/E, or study team implicit bias as measured by performance on the Harvard IAT. Rather, approach and enrollment were associated with factors such as consent mechanism and CRC gender. Each factor may be related to patient bias as preference for a certain type of consent and female CRC may reflect patient/family bias. Interventions exploring patient biases towards research may help to advance pediatric research equity.