A goals-plans-action model analysis of messages encouraging hesitant family members in the United States to get vaccinated for COVID-19

ABSTRACT

Drawing on the goals-plans-action (GPA) model, we explore how individuals encourage hesitant family members to get vaccinated for COVID-19. We test models analyzing how multiple goals mediate associations between (a) perceived threat to the family member’s health, (b) anticipated resistance, and (c) participants’ political ideology, with their (d) planning effort and (e) message features. Participants who perceive a greater threat, anticipate greater resistance, and are more liberal put greater emphasis on the primary goal, which in turn predicts greater effort, directness, and reason giving. Liberal participants also place less emphasis on secondary goals, and thus exert more pressure. Theoretical implications linking the GPA model with fear appeals and intergroup communication frameworks and practical implications for families and public health campaigns are discussed.

KEYWORDS:

• difficult conversations
• goals-plans-action model
• vaccine hesitancy
• family communication
• political ideology

Disclosure statement

No potential conflict of interest was reported by the author(s).

Notes

1 Data quality was assessed by reviewing IP addresses for duplicate responses, reviewing open-ended responses, examining response speeds (any participant > the 99th or < the 1st percentile was removed), and removing those who failed attention check questions or “straight-lined” responses (based on mean root of squares and scale point variation indices; Kim et al., 2019). Some potential participants were removed because their responses to open-ended questions were copied from the internet (i.e., bots). Using these procedures, we eliminated 142 responses.

2 Given evidence that perceived susceptibility and severity may play distinct roles in terms of message acceptance (e.g., Totzkay et al., 2022), we report supplemental analyses comparing the individual effects of these two variables. Although the strength of association sometimes varies, the overall findings are similar for perceived susceptibility and severity; hence, most analyses are based on an overall perceived threat score that averages the two variables.

3 Bootstrapped confidence intervals for omegas are based on 1000 resamples.

4 Pilot responses were collected via MTurk prior to the main survey experiment, to gather feedback on scenario realism and develop coding schemes for rating written messages.

5 Skewness and kurtosis for most study variables indicated that distributions were approximately normal (skew within ±1.0 and kurtosis within ±2.0). However, the three types of reasons were negatively skewed, as is common for “count” variables. Log transformations (log10 [X + 1]) were performed on protect yourself, protect others, and vaccine-related reasons. Raw scores are reported given similar findings with transformed variables.

6 Because Democrats are overrepresented in our sample (see Supplemental Table 1), we ran the 2 × 2 ANCOVA with anticipated resistance as the dependent variable while including political liberalism as a second covariate along with perceived scenario realism. Results were virtually identical even after controlling for liberalism.

7 We ran PROCESS analyses substituting either perceived susceptibility or severity for perceived threat, to see if findings differed when susceptibility or severity was used instead of threat. See Figures 5 and 6 in Supplemental Materials for findings when the dependent variable is planning effort. Given that findings were similar regardless of whether perceived susceptibility and severity were analyzed separately or were combined, we only report results for the combined perceived threat variable in the manuscript.