A self-affirmation exercise does not improve intentions to vaccinate among parents with negative vaccine attitudes (and may decrease intentions to vaccinate)

Rachael D. Reavis; Jacob B. Ebbs; Adaobi K. Onunkwo; L. Mariah Sage

doi:10.1371/journal.pone.0181368

Abstract

Two studies investigated the effectiveness of a self-affirmation exercise on vaccine safety beliefs and intent to vaccinate future children. In Study 1, a sample of 585 parents with at least one child under the age of 18 in the home participated through Amazon’s MTurk. Participants were randomly assigned to one of four conditions in a 2 x 2 design. Participants read either correcting information refuting a link between the measles, mumps, and rubella (MMR) vaccine and autism or a control passage about bird feeding. Additionally, participants either completed a self-affirmation exercise where they reflected on their personal values or in a control condition in which they reflected on least-personally-important values that might be important to others. Participants exposed to the correcting information were less likely to believe that vaccines cause serious side effects, but no less likely to believe that the MMR vaccine causes autism. For parents with initially positive vaccine attitudes, there was no effect of condition on intent to vaccinate a future child. For parents with initially negative vaccine attitudes, self-affirmation was ineffective in the presence of correcting information and resulted in less intention to vaccinate in the absence of correcting information. This effect was partially replicated in Study 2 (N = 576), which provided no correcting information but otherwise followed the same procedure as Study 1.

Citation: Reavis RD, Ebbs JB, Onunkwo AK, Sage LM (2017) A self-affirmation exercise does not improve intentions to vaccinate among parents with negative vaccine attitudes (and may decrease intentions to vaccinate). PLoS ONE 12(7): e0181368. https://doi.org/10.1371/journal.pone.0181368

Editor: Chris T. Bauch, University of Waterloo, CANADA

Received: March 21, 2017; Accepted: May 24, 2017; Published: July 13, 2017

Copyright: © 2017 Reavis et al. This is an open access article distributed under the terms of the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original author and source are credited.

Data Availability: Data are available in two files (one for Study 1; one for Study 2). Both are accessible via the Open Science Framework (doi: 10.17605/OSF.IO/5KBSE).

Funding: The authors received no specific funding for this work.

Competing interests: The authors have declared that no competing interests exist.

Introduction

The Center for Disease Control and Prevention (CDC) recommends administration of 13 different vaccines for children ages 0 to 18 [1,2]. Vaccines are largely responsible for the impressive decline in childhood illnesses [3], many of which are now rare in the United States (e.g., polio, measles). However, lower-than-ideal vaccination rates in the U.S. are a growing public health concern [4], and have been implicated in several measles and pertussis outbreaks [3]. Despite the scientific evidence that vaccines are safe for the large majority of children and prevent childhood disease and mortality [5,6], some parents remain resistant to vaccinating their children. Researchers have tested various messaging techniques to decrease negative vaccine attitudes and to increase vaccination rates, but have found it difficult to change parents’ negative attitudes (e.g., [7,8]). The present study attempted to apply a psychological intervention—a self-affirmation exercise—which has proven helpful in health messaging acceptance in other domains (e.g., smoking, healthy eating) [9,10] to vaccine safety messaging for parents.

Fifteen U.S. states are below the Healthy People 2020 target vaccination rate of 90% and 36 states are below the World Health Organization (WHO) recommendation rates of 93% vaccination rates to prevent epidemics [11,12]. From 2001 to 2010, there were eight years with fewer than 100 cases of measles in the U.S. and two years with fewer than 150 cases [13]. In 2011 and 2013 there were approximately 200 cases each year, and in 2014, there were over 600 cases in 23 outbreaks. In all outbreaks, the majority of infected individuals had not been vaccinated [13]. The largest outbreak in 2014 originated in California at Disneyland [14], and there are schools in California where fewer than 50% of kindergartners are up-to-date on their vaccines [15].

The rise in parents choosing not to vaccinate their children may be partially due to parental fears that vaccines cause serious side effects, including autism [16]. Although fears about vaccines have been present since their introduction [17], the wave of current fears is due in part to a discredited study by Andrew Wakefield, which purported to show a link between the MMR vaccine and autism [18]. Despite that fact that this study was eventually withdrawn and determined to be fraudulent [19] and that the scientific consensus is that vaccines are not related to the development of autism (e.g., [20,21]), vaccine safety concerns have persisted [22].

Nyhan and colleagues [7] tested the effect of several messaging techniques on negative attitudes toward vaccines (beliefs that they cause autism or other severe side effects) and likelihood to vaccinate future children. One technique was to expose parents to scientific evidence, similar to what is provided by the CDC, which refutes a link between the MMR vaccine and autism. Interestingly, although exposure to such information decreased agreement with statements that the MMR vaccine is unsafe, exposure also decreased intention to vaccinate future children, particularly among those with already negative attitudes toward vaccines. Thus, although the researchers found evidence of surface acceptance of the health messaging (in that beliefs about vaccine dangers decreased), the health message had a negative effect on intent to vaccinate. Parents with negative attitudes were paradoxically less likely to say they would vaccinate a future child than if they had been exposed to no messaging at all.

One possibility for this paradoxical effect is that the correction of misinformation may have threatened participants’ sense of self. We are generally motivated to maintain a positive sense of self [23], and for parents with negative attitudes toward vaccines, being exposed to conflicting scientific information could be perceived as threatening. In other areas of health messaging, participants who read health messages that are particularly relevant for their lives (e.g., obese people who are unhealthy eaters reading about the importance of healthy eating) may react defensively as if the health message is a personal attack on their self-worth [24]. For example, an obese person who is an unhealthy eater may interpret health messages about healthy eating as implying that they lack self-control or are ignorant, which would threaten their self-worth. Thus, they are more likely to dismiss the message.

When people affirm their sense of self, they are less sensitive to threats to their self-worth and may be more willing to accept health messages [24]. One method to help participants affirm and strengthen their sense of self is to have individuals reflect on their personal values and why they view them as important [25]. Individuals who self-affirm tend to see potentially threatening health messages as less threatening [24]. Self-affirmation may help individuals maintain a positive self-image and seems to improve the acceptance of health messages in the domains of fruit and vegetable consumption and smoking [26,27].

The effects of self-affirmation on health message acceptance seem to be largest for relevant high-risk groups. For example, high-risk participants who completed a self-affirmation exercise showed the most willingness to take a Type-2 diabetes risk test following a health message, compared to low-risk participants or high-risk participants who did not self-affirm [28]. Similar results have been seen with consumption of high-mercury seafood and caffeine consumption [29,30].

Self-affirmation may be an effective tool at increasing acceptance of vaccine safety messaging, particularly for those who have more negative vaccine attitudes (e.g., those for whom the messages are intended and are the most relevant). Based on research in other health domains, we would expect self-affirmed participants—particularly those with initially negative vaccine attitudes—to be more willing to accept the safety messaging, and perhaps be more willing to vaccinate future children.

However, there are important differences between vaccine safety messaging and other types of health messages. The first is that in the case of vaccines, parents are making health decisions not for themselves, but for their children. The effect of self-affirmation for health decisions made for others rather than the self has not been studied. Another difference is that parents with negative vaccine attitudes may have deeply held beliefs that vaccines are unsafe, whereas it is unlikely that individuals believe that cigarettes are good for you (as opposed to believe cigarettes are not as bad as health messaging claims) or that eating vegetables is an unhealthy lifestyle. It may be that self-affirmation will actually strengthen these negative attitudes and make participants less likely to accept vaccine safety messages. Given these differences, we do not make directional hypotheses testing the effect of self-affirmation on acceptance of vaccine health messaging, particularly for those with initially negative vaccine attitudes. In the absence of self-affirmation, we predict results similar to those of Nyhan et al. [7]. We expect that participants exposed to scientific information refuting the connection between the MMR vaccine and autism will be less likely to agree with statements that suggest vaccines are unsafe, but that those with initially negative vaccine attitudes will be less likely to say they would vaccinate a future child if exposed to correcting information versus a control condition. Again, for participants with negative vaccine attitudes, self-affirmation may strengthen or reverse these results.

Study 1

Method

The Earlham College IRB approved this research (IRB Approval Number: 1314-e027). All participants provided written (digital) consent.

Participants.

Participants were 585 adults (47% male) with at least one child under the age of 18 living in the household. The median number of children in the home was one child (M = 1.65, SD = 0.88, range: 1–5 children). All participants were recruited through MTurk and were compensated $0.30, which was an effective hourly rate of $2.65. The consent form indicated that participation required having one child under the age of 18 living in the home. Originally, 680 participants opened the survey, but 15 did not advance past the consent form, and 37 did not advance past the first question where they indicated the number of children (28 of those indicated that they had no children at home). Six participants did not complete past the demographics. Fifteen participants indicated their initial attitudes and then stopped participating. Three participants completed the first experimental manipulation (self-affirmation), but did not advance to the second (correction information vs. control passage.) An additional 11 participants failed to agree with the statement, “I was honest in answering all questions (including about whether I have a child in the home),” with five disagreeing and six skipping the question. Six participants answered an attention check question incorrectly, and two participants did not follow instructions and failed to write about their values when prompted.

The sample was predominantly Caucasian (70%). Educational background was varied, ranging from participants who had not graduated from high school to participants with doctoral and professional degrees (47% had a 4-year college degree or above). The average age of parents was 32.76 years (SD = 7.9 years; range: 18–61 years). The majority of participants were married or cohabitating (76%). The majority of participants were from suburban areas (57%), followed by urban (25%), and rural (18%). Participants were from 48 U.S. states plus the District of Columbia. There were no participants from Montana or Vermont. There was a wide range of income, with 4% reporting annual incomes under $10,000 and 2% reporting incomes over $150,000, with a median income bracket of $40,000 to $49,999. Table 1 shows demographic comparisons between the sample of the current study and Nyhan and colleagues’ sample [7].

Download:

Table 1. Comparison of demographics between Nyhan et al. [7], Study 1, and Study 2.

https://doi.org/10.1371/journal.pone.0181368.t001

Procedure.

Participants completed the survey on Qualtrics, using a link on MTurk, in June 2014. A consent form indicated that the study was on memory and health decisions for children. Participation was restricted to parents/guardians of at least one child in the home under the age of 18. Following a demographic form, participants completed a measure of their vaccine attitudes [31], plus two questions about political affiliation. Statements were presented randomly. Participants were then randomly assigned to a self-affirmation or control condition using a standard values affirmation procedure (e.g., [25]). Participants were presented (in random order) with 11 qualities they might value (e.g., religious values, humor) and asked to either select the value that was most important (self-affirmation) or least important (control). They then either wrote about why they felt the value was important to them (self-affirmation) or to someone else (control). Finally, they answered four questions on a 6-point Likert scale (Strongly Disagree to Strongly Agree) about the importance of the value to themselves (self-affirmation) or to other people (control).

Following the values procedure, participants were randomly assigned to one of two reading passages, both of which were taken from Nyhan et al. [7]. They were told that they would answer questions about the passage later on, although they did not actually do this. The instructions were to increase the likelihood of close reading. In the Control condition, participants read about the costs and benefits of feeding birds. In the Autism Correction condition, participants read information from the CDC refuting the link between autism and the MMR vaccine. There were four total conditions: Autism Correction/Self-Affirmation (N = 156), Autism Correction/Values Control (N = 139), Control Passage/Self-Affirmation (N = 138), and Control Passage/Values Control (N = 152).

Participants were presented (in random order) three outcome questions measuring beliefs about the MMR vaccine and autism (also from Nyhan et al. [7]) and an attention check question, which asked participants to select “Neutral” on a “Strongly Disagree” to “Strongly Agree” scale to demonstrate that they had been paying attention. After entering their MTurk ID, participants were asked to agree or disagree with a statement about whether they had answered honestly and were told that their answer would not affect their payment (and it did not). Participants were then debriefed. The procedure with individual questions is in the appendix.

Measures.

Pre-intervention vaccine attitudes were assessed using 10 questions from Freed et al. [31], answered on a 5-point Likert scale from Strongly Disagree to Strongly Agree, such as “Getting vaccines is a good way to protect my child(ren) from disease,” and “Some vaccines cause autism in healthy children” (reverse scored). The scale was reliable, α = .89. Participants who discontinued after indicating their attitudes had more negative attitudes (M = 3.95, SD = 0.85) than participants who completed the study (M = 4.31, SD = 0.79; t(620) = -2.74, p = .006), suggesting that the eventual sample was biased toward those with more positive views of vaccines.

To examine the effect of condition based on initial vaccine attitudes, Nyhan and colleagues used a tercile split. Participants whose vaccine attitudes scores were less than 3.57 were classified as having “least favorable attitudes,” (and referred to as “negative” in the conclusion) even though their score is closer on the scale to “Agree” than to “Neither Agree or Disagree.” (Those with scores greater than or equal to 4.125 were classified as having “most favorable” attitudes.) These cut-offs resulted in those with “least favorable” attitudes having a group mean of 3.0 (“Neither Agree Nor Disagree”). We felt it was important to examine negative attitudes and to use an objective, rather than relative, cut-off for determining those negative attitudes. Thus, we divided participants based on their position on the scale, not based on their position relative to other participants. This approach resulted in participants being classified into a positive vaccine attitude group (individual mean scale score at or above 4 or “Agree,” N = 399, group M = 4.74, SD = 0.44), a negative vaccine attitude group (individual mean scale score below 3 “Neither Agree Nor Disagree,” N = 38, group M = 2.51, SD = 0.43), or being removed from analyses (N = 148). However, we have included analyses using both approaches.

Following Nyhan et al. [7], autism belief was measured with a single item, answered on a 5-point Likert-scale (Strongly Disagree to Strongly Agree): “Some vaccines cause autism in healthy children.” Side effects belief was measured with a single item, answered on a 6-point Likert-scale (Very Unlikely to Very Likely): “How likely do you think it is that a child will suffer from serious side effects from the measles, mumps, and rubella (MMR) vaccine?” Intent to vaccinate was measured with a single item on the same 6-point scale: “If you had another child, how likely is it that you would give that child the measles, mumps, and rubella vaccine, which is known as the MMR vaccine?”

Results

Table 2 displays the descriptive statistics for Study 1 variables. Using a chi-square, we tested whether participants who were excluded after the experimental manipulations were more likely to have been assigned to one condition over another. We saw no support for this, χ² = 1.96, p = .58. We examined the effect of conditions on each of three outcomes: autism belief, side effects belief, and intent to vaccinate. In each case, we first ran a 2 (passage condition: autism correction or control) x 2 (affirmation condition: self-affirmation or control) ANOVA with the full sample. Then we conducted a 2x2x2 ANOVA with the vaccine attitude group as a third variable, which involved a subset of the sample and did not include participants with moderate vaccine attitudes.

Download:

Table 2. Means & Standard deviations of Study 1 variables.

https://doi.org/10.1371/journal.pone.0181368.t002

Autism belief.

We found no main effect of passage condition on beliefs that the MMR vaccine causes autism, F(1,579) = 0.06, p > .250, which was in contrast to Nyhan et al. [7]. Likewise, there was no main effect of affirmation condition on autism belief, nor was there an interaction with passage condition, Fs(1, 579) < 0.10, ps > .250. Using the subsample based on vaccine attitudes, we found no significant three-way interaction between passage condition, affirmation condition, and initial vaccine attitudes on autism belief, F(1, 427) = 0.18, p > .250. We also tested the three-way interaction using Nyhan’s cut-offs for relative vaccine attitudes. This interaction was not statistically significant, F(1, 461) = 2.32, p = .128.

Side effects belief.

Unlike Nyhan et al. [7], we found a small main effect of passage condition on side effects belief, F(1, 580) = 4.95, p = .026, η²_p = 0.01, such that participants who read the autism correction passage agreed less with the statement that the MMR vaccine is likely to cause serious side effects (M = 2.15, SD = 1.14, 95% CI: 2.01–2.29) than did participants who read the control passage (M = 2.38, SD = 1.29, 95% CI: 2.23–2.52). There was no main effect of affirmation condition and there was no interaction with passage condition, Fs(1, 580) < 1.0, ps > .250. Using the subsample based on vaccine attitudes, we found no significant three-way interaction between passage condition, affirmation condition, and initial vaccine attitudes on side effects belief, F(1, 429) = 0.07, p > .250. Similarly, there was no statistically significant three-way interaction using Nyhan’s cut-offs for relative vaccine attitudes, F(1, 462) = .14, p > .250.

Intent to vaccinate.

Whereas Nyhan et al. [7] found that reading the autism correction passage decreased the intent to vaccinate a future child, we found no significant main effect of passage condition on intent to vaccinate, F(1, 578) = 0.03, p > .250. There was no main effect of affirmation condition and there was no interaction with passage condition, Fs(1, 580) < 1.0, ps > .250. Using the subsample based on vaccine attitudes, we found a significant three-way interaction between passage condition, affirmation condition, and initial vaccine attitudes on intent to vaccinate, F(1, 426) = 4.02, p = .046, η²_p = 0.01 (see Fig 1). For participants with initial positive vaccine attitudes, there was little difference between conditions. For participants with initial negative vaccine attitudes, values affirmation decreased intentions to vaccinate in the absence of correcting information (i.e., in the control passage condition) and had no effect in the presence of it (i.e., in the autism correction condition).

Download:

Fig 1. Depiction of three-way interaction between passage condition, affirmation condition, and initial vaccine attitudes.

Vaccine attitudes [31] were measured on a 5-point scale (Strongly Disagree to Strongly Agree). (1a) Participants with mean attitude score lower than 3. (1b) Participants with a mean attitude score of 4 or greater. Intent to vaccinate was answered on a 6-point scale (Very Unlikely to Very Likely). Reading passages were drawn from Nyhan et al. [7]. The values affirmation procedure was drawn from Sherman et al. [25]. Error bars represent ±1 SE.

https://doi.org/10.1371/journal.pone.0181368.g001

The analysis using Nyhan’s cut-offs for relative vaccine attitudes was not statistically significant (F(1, 460) = 1.31, p > .250), but the pattern was similar.

Study 2

We conducted a second study to attempt to replicate our result that in the absence of correcting information, self-affirmation may increase parental resistance to vaccines among those with initially negative attitudes. Because we were interested in parents’ responses in the absence of correcting information, we dropped the passage condition. Thus, all participants either self-affirmed or did not and no participants in Study 2 saw correcting information about the safety of vaccines.