Disparities by Ethnicity in Enrollment of a Clinical Trial.

Abstract

Previously identified barriers to diverse inclusion in research include participant-level factors, such as transportation, child care, and marginalization by academic medicine.1 Efforts to overcome these barriers have generally focused on participant-level levers, such as participation incentives and efforts to optimize subsequent contact.2 Another barrier could be personally mediated racism, defined as differential assumptions about and actions toward others according to their race, which can occur intentionally or unintentionally.3We conducted a secondary analysis of data from the Bronchiolitis Follow-up Intervention Trial (BeneFIT), in which researchers randomly assigned previously healthy children hospitalized for bronchiolitis to an automatic (the control group) versus as-needed (PRN) (the intervention group) posthospitalization follow-up visit.4 Enrollment was limited to children whose parents spoke English or Spanish; however, parent language preference was only recorded for enrolled patients. The electronic medical record was used to collect demographic data, including insurance type as a proxy for poverty. Before enrollment, the hospitalists and/or primary care providers of eligible children were asked to confirm that the intervention group, PRN follow-up, could be appropriate for the child (physician-determined appropriateness). If a patient was assessed as not appropriate for the intervention, physicians provided their rationale. These rationales were previously reported and included clinical (ie, illness severity or chronic medical needs) and social (ie, perceived low parent health literacy or social risk) justifications.4 Next, parents of physician-determined appropriate children were offered enrollment if they were willing to accept the possibility of PRN follow-up (parental acceptance of the intervention). Among patients randomly assigned to PRN follow-up, parents reported their compliance with this recommendation, namely, not automatically scheduling a follow-up visit at the time of discharge (parental adoption of the intervention). These measures for uptake of the trial intervention (PRN follow-up) were operationalized from previously developed implementation outcomes.5We assessed the association between patient demographic and clinical factors with the 3 intervention uptake measures using χ2 tests. We constructed multivariable logistic regression models that included factors with P values <0.20 from bivariable analyses.Of 548 eligible patients for BeneFIT, 506 (92%) were physician-determined appropriate for the intervention (Fig 1). Of physician-determined appropriate patients, 29% were of Hispanic ethnicity, compared with 52% Hispanic ethnicity among patients considered not appropriate (Table 1). In the multivariable model, only Hispanic ethnicity remained significantly associated with decreased odds of physician-determined appropriateness (odds ratio 0.4, 95% confidence interval 0.2–0.8).Of the 506 physician-determined appropriate patients, 99 (20%) were excluded for reasons unrelated to parental acceptance of the intervention (Fig 1). The parents of 372 (91%) of the remaining 407 patients demonstrated parental acceptance of the intervention. The only factor associated with parental acceptance in the multivariable model was mean duration of illness (6.0 days among accepting parents versus 8.2 days among nonaccepting parents [odds ratio 0.8, 95% confidence interval 0.7–0.9]).Of patients randomly assigned to the intervention, parental adoption occurred for 112 of 138 (81%) patients. Parental adoption was not associated with any examined demographic or clinical factors.Physicians more often determined that Hispanic patients were not appropriate for PRN follow-up and excluded them from trial participation, compared with non-Hispanic patients. However, among those approached for enrollment, Hispanic participants demonstrated equivalent acceptance and adoption of PRN follow-up compared with non-Hispanic participants.Physicians may have recognized true barriers to accessing primary care and appropriately screened out families for whom PRN follow-up would be challenging to adopt; this interpretation points to the need to enhance equitable access to care. The alternative explanation is personally mediated racism (eg, assuming low health literacy or increased social risk for Hispanic families), which has been previously described in medical settings.6–8 Differential trial access based on biased assumptions can deprive historically excluded populations from the direct benefits of research and decrease generalizability of findings. Relevant to this trial’s specific intervention, previous research suggests that clinic visits involve more time burden and financial sacrifices for Black and Hispanic patients.9,10 Therefore, our findings indicate that families with potentially the most to gain from PRN follow-up were the least likely to be offered the opportunity to enroll.This secondary analysis of trial enrollment data is limited by the likely existence of additional, unmeasured variables that influence uptake of PRN follow-up (ie, financial or social hardship), and the applicability of these findings to other trial settings is unknown. Nevertheless, our findings suggest that physician biases may have reduced the generalizability of trial results and perpetuated disparities by depriving Hispanic families of an opportunity to reduce their time and financial burdens after hospital discharge through trial participation. In future trials, researchers should consider interventions addressing provider and researcher biases at the time of participant screening to maximize diverse and inclusive participation.