Racial/Ethnic Differences in the Predictive Utility of Psychosocial Determinants of COVID-19 Vaccination Intentions

Design

An anonymous Qualtrics survey was conducted online at the University of Houston May-August 2021. Eligible participants were unvaccinated college students recruited through a cloud-based participant pool management system (ie, SONA). All participants provided electronic consent. The protocol was approved by University of Houston’s Institutional Review Board.

Sample

Participants were 314 unvaccinated college students. We excluded 12 participants because they failed to provide their vaccination intentions (n = 6) or had received the COVID-19 vaccine (n = 6). The final sample included 302 participants (77.2% female) with a mean age of 22.17 (SD = 4.10). The sample was racially/ethnically diverse, with 25.8% Hispanic, 22.2% European American, 20.9% Asian American, 12.6% African American, 11.9% multiracial/multiethnic, and 6.6% ‘other.’ Most participants were employed (55.0%) and had never tested positive for COVID-19 (81.1%). See Supplemental Table 1 for additional demographics.

Measures

Complete measures can be found in Supplemental Tables 2-4. The primary outcome variable, COVID-19 vaccination intentions, was adapted from a single-item measure: “All things considered, how likely are you to get a coronavirus (ie, COVID-19) vaccine?” Participants responded on a 5-point Likert-type scale, with higher scores indicating stronger intentions.

Analysis

Analyses were performed using Stata 15.1. First, we conducted a series of one-way ANOVAs to test for race/ethnicity differences across continuous variables. Second, we used block-sequential multiple regression to test hypothesized predictors and examine race/ethnicity moderation effects. We used Tobit regression because a relatively high percentage (∼35%) of intention scores were right censored. Continuous predictors were mean-centered prior to analysis. Demographics were entered into Block 1. TPB and HBM constructs were entered into Block 2. Hypothesized predictors were entered into Block 3. We tested for moderation in Block 4 by adding interactions between race/ethnicity and each hypothesized predictor. Simple slopes analyses were performed to identify significant race/ethnicity slopes and pairwise differences in slopes across racial-ethnic groups. Tobit coefficients were decomposed into two components: (1) linear effects (LE)—below-threshold change in intentions with a one-unit change in the independent variable, and (2) probability-below-threshold effects (PBT)—probability that the intentions score is below the upper threshold (ie, 4) with a one-unit change in the independent variable. Minimum sample size (250 participants) was calculated using G*Power 3.1, assuming an R² increase in multiple regression (Block 4: 4 categorical variable X continuous variable interactions above Blocks 1-3 [ie, 25 terms]) as the final and most complex model to be analyzed (46 terms; α = .05; 1-β = .80; f² = .10).

Primary Analyses

The final model (Supplemental Table 7) fit the data well, χ² (45) = 266.96, P < .001, and explained ∼77% of the variance in COVID-19 vaccination intentions. The TPB constructs of self-efficacy and attitudes toward COVID-19 vaccination were positively related to vaccination intentions. In contrast, behavioral control and anticipated regret for vaccinating and subsequently getting COVID-19 were negatively related to intentions.

Summary

In the current study, subjective norms, vaccine conspiracy beliefs, and prosocial concern predicted COVID-19 vaccination intentions. As expected, we found vaccine conspiracy beliefs to be a stronger predictor of intentions for people of color than European Americans. However, we also found race/ethnicity differences in the predictive utility of other hypothesized constructs, suggesting the importance of culturally tailoring vaccine messages.

Our findings suggest stronger subjective norms are associated with stronger vaccination intentions for European and African Americans but not for other racial-ethnic groups. These racial-ethnic differences could be attributed to acculturation levels. For instance, subjective norms tend to be more salient and influential for Hispanics who are less accultured. ¹⁰ Notably, our sample could have been comprised of more accultured individuals, especially since less accultured individuals have unique barriers that affect their ability to pursue higher education. ¹¹

Unexpectedly, race/ethnicity-specific descriptive norms did not predict intentions. It is possible that participants did not strongly identify with their racial-ethnic group. Regardless, all racial-ethnic minority groups’ descriptive norms were lower than their groups’ actual COVID-19 vaccination national percentage. ¹² Although our findings suggest normative corrections may not directly increase intentions, such corrections could improve other antecedents of intention (eg, attitudes).

Prosocial concern predicted intentions only for European Americans and multiracial/multiethnic Americans. This could be due to racial-ethnic differences in perceived susceptibility. Specifically, prosocial motivation predicts vaccination intentions more strongly when perceived susceptibility is low. ¹³ Thus, messages that emphasize the prosocial benefits of vaccination may be particularly impactful for groups with low perceived susceptibility levels (eg, European Americans).

Limitations

There are a few limitations. First, causal interpretations are not possible. Second, unmeasured factors could differentially impact intentions in a racially/ethnically diverse population. For instance, medical mistrust is considered to be an important component of vaccination hesitancy among people of color. ¹⁴ Finally, our sample was relatively small and comprised primarily of female students. There are also several strengths. First, the sample was racially/ethnically diverse. Second, this research extended common theoretical frameworks. Finally, Tobit regression accounts for threshold/censored values and does not dichotomize intentions, which reduces information loss and chances of spurious findings.

Significance

Our findings underscore the importance of emphasizing the prosocial benefits of vaccination for racial-ethnic groups with relatively low perceived susceptability levels (eg, European Americans); (2) signaling a positive subjective norm for vaccination, especially among European Americans and African Americans; and (3) correcting vaccination conspiracy beliefs. Our findings have implications for narrative persuasion messages, which use stories tied to audiences’ sociocultural experiences to present factual information and empower informed health choices. ¹⁵ Future narrative messages could have story protagonists acquire vaccine information from important others in their racial-ethnic group who express vaccine approval and emphasize family/community needs and expectations. These messages may spark conversations with close others about the importance of COVID-19 vaccination.

So What?