Abstract
Background. Maternal vaccine coverage is suboptimal, and a substantial proportion of parents have concerns about vaccines. Most parents seek out vaccine information during and immediately after their first pregnancy. No study to our knowledge has analyzed survey data to identify homogeneous groups of pregnant women based on their vaccine attitudes and beliefs. Aims. To identify homogeneity among groups of pregnant women based on their vaccine attitudes and beliefs to facilitate audience segmentation and targeting of tailored educational interventions. Method. Between June 2017 and July 2018, we surveyed 2,196 pregnant women recruited from geographically and sociodemographically diverse prenatal care practices in Georgia and Colorado. We then performed a latent class analysis to identify homogeneity among groups of pregnant women. Results. Our latent class analysis produced three groups of pregnant women: vaccine supporters (36% of women), vaccine acceptors (41%), and vaccine skeptics (23%). Discussion. The major difference between the supporters and the acceptors were whether they mostly “strongly agreed” or just “agreed” to Likert-type scale survey items assessing their vaccine attitudes and beliefs. The skeptics most frequently chose “disagree” or “don’t know” for items assessing attitudinal constructs such as confidence in vaccine safety and efficacy and disease susceptibility. However, even skeptics often chose “agree” for items assessing constructs such as disease severity and self-efficacy. Conclusions. This article provides useful insight into the homogeneity among groups of pregnant women based on their vaccine attitudes and beliefs. This knowledge should help facilitate audience segmentation and targeting of tailored educational interventions among this population.
The Advisory Committee on Immunization Practices (ACIP) recommends that pregnant women receive two vaccines during pregnancy: tetanus toxoid, reduced diphtheria toxoid, and acellular pertussis vaccine (Tdap); and seasonal inactivated influenza vaccine (IIV; Centers for Disease Control and Prevention, 2013; Grohskopf et al., 2017). Unfortunately, only about half of pregnant women in the United States receive these vaccines (Kahn et al., 2018), which is substantially lower than the relatively high vaccine coverage among children (Hill et al., 2018).
Vaccine hesitancy, defined as “concerns about the decision to vaccinate oneself or one’s children” has emerged in recent decades as a threat to high vaccine coverage in all age ranges (Salmon et al., 2015). Clustering of vaccine refusal has led to associated outbreaks of vaccine-preventable diseases (VPDs; Atwell et al., 2013; Glanz, Narwaney, et al., 2013; Omer et al., 2018). The most recent outbreak of measles in the United States, in which more cases have been reported so far in 2019 than in any year since 1994, is an example of this threat (Patel et al., 2019).
Many parents seek out information about vaccines primarily during and immediately after their first pregnancy (Betsch et al., 2018; Glanz, Wagner, et al., 2013; O’Leary et al., 2018; Vannice et al., 2011). The first pregnancy may thus be a “teachable moment”—an important opportunity to provide accurate information pertaining to both maternal and infant vaccinations—as the parents’ vaccine attitudes and beliefs may not yet be solidified (Frew, Owens, et al., 2014; Frew, Saint-Victor, et al., 2014; Salmon et al., 2015). This is especially true because a vast majority of parents (Chung et al., 2017; Freed et al., 2011; Smith et al., 2006) and pregnant women (Beel et al., 2013; Dlugacz et al., 2012; Gorman et al., 2012; Healy et al., 2015; Henninger et al., 2015; MacDougall et al., 2016; MacDougall & Halperin, 2016) consider health care providers their most trusted source of vaccine information, even among those who express vaccine hesitancy. However, much lower levels of trust are reported in government and its associated agencies as well as in the pharmaceutical industry among those who express vaccine hesitancy compared with those who do not (Freed et al., 2011; Lee et al., 2016; Smith et al., 2006).
Although most parents believe vaccines to be safe, effective, and important, some also have concerns regarding vaccines, such as young infants receiving too many vaccines or potential serious side effects of vaccines (Chung et al., 2017; Freed et al., 2010; Frew et al., 2016; Kennedy et al., 2011). Substantial variance in vaccine knowledge, attitudes, and beliefs exist among those of differing gender, education, socioeconomic status, ethnicity, and race (Freed et al., 2010). Educational interventions should be tailored to account for these differences, especially considering the potential for nontailored messaging to be ineffective or even backfire among certain subgroups (Nyhan & Reifler, 2015).
Market segmentation and message targeting are strategies used to tailor messages to the specific needs, concerns, and barriers common to a certain number of relatively homogeneous groups or segments, instead of relying on generally less effective one-size-fits-all or highly resource-intensive individually tailored approaches. Such strategies are widely used in market/customer experience research in a variety of fields and are often based on behavioral theories such as the stages of change (Prochaska & DiClemente, 2005) and the consumer decision journey (Court et al., 2009). A good example of segmentation used in a vaccine context is WHO’s (World Health Organization) Tailoring Immunization Programmes, which is based on the COM-B (capability, opportunity, and motivation behavior system) framework (Michie et al., 2011) and provides countries a process to diagnose motivators and barriers to vaccination and design interventions tailored to these factors (Dube et al., 2018).
Very few statistical analyses of survey data have been conducted to identify homogeneous groups of parents based on their vaccine attitudes and beliefs. Gellin et al. (2000) analyzed data from a 1999 national telephone survey, of which three main (yet overlapping) subgroups of parents were identified: those who rated immunization as “extremely important” (87%), those who considered government or school requirements as the principal motivation for immunization (8%), and those who would choose to opt out of at least one immunization for their child (14%). Gust et al. (2005) performed a K-means clusters technique on data from 2002 HealthStyles and ConsumerStyles surveys and identified five homogeneous groups of parents: Immunization Advocates (33.0%), Go Along to Get Alongs (26.4%), Health Advocates (24.8%), Fencesitters (13.2%), and Worrieds (2.6%). Smith et al. (2011) analyzed data of more than 11,000 parents of young children from the 2009 National Immunization Survey and categorized parents as neither delaying nor refusing vaccines (60%), only delaying one or more vaccines (26%), only refusing one or more vaccines (8%), and both delaying and refusing vaccines (6%). Leask et al. (2012) used a literature review to identify five distinct parental groups as part of a framework for health professionals: the “unquestioning acceptor” (30%–40%), the “cautious acceptor” (25%–35%), the “hesitant” (20%–30%), the “late or selective vaccinator” (2%–27%), and the “refuser” of all vaccines (<2%). Weiss et al. (2016) performed a latent class analysis (LCA) on 14 five-point Likert-type scale belief statements from 189 questionnaires completed by Swiss mothers living in the Aargau region and identified three latent classes: positive attitudes (58%), fearful/uncertain attitudes (28%), and critical attitudes (14%).
Although the aforementioned studies used statistical analysis to identify homogeneous groups of parents based on their vaccine attitudes and beliefs, no study to our knowledge has done the same for pregnant women, a crucial population for early vaccine intervention. Thus, the objective of this study was to identify homogeneity among groups of pregnant women based on their vaccine attitudes and beliefs to facilitate audience segmentation and targeting of tailored educational interventions.
Method
Study Design
This analysis was possible due to the existing infrastructure of the P3+ study, a large randomized controlled trial of a three-level prenatal intervention to increase uptake of maternal and infant vaccines. Pregnant women were recruited by study staff from waiting rooms of a geographically and sociodemographically diverse set of prenatal care settings (including both obstetric and midwife practices) in Georgia and Colorado between June 2017 and July 2018. Women were eligible for participation if they were 8 to 26 weeks pregnant, 18 to 50 years old, and had not yet received Tdap vaccine during their current pregnancy. The study was approved by the institutional review board.
Data Collection
A survey was administered immediately on enrollment via tablets in the waiting rooms, and a $20 incentive was provided for its completion (Salmon et al., 2019). This survey included multiple-choice questions to assess the number of prior children and the baseline intention of participating women to receive recommended vaccines for themselves and their infants. The survey also included 58 Likert-type scale statements assessing latent attitudinal constructs related to maternal and infant vaccination—namely, confidence in vaccine safety and efficacy, perceived susceptibility to and severity of VPDs, descriptive (what people typically do) and injunctive (what people typically approve or disapprove) norms (Cialdini, 2003), self-efficacy (an individual’s belief in their capacity to execute behaviors necessary to produce specific performance attainments; Bandura, 1977), perception of knowledge, and trust in information sources. These constructs were chosen based on the review of other relevant behavioral models, theories, and scales (Gorman et al., 2012; Opel et al., 2011). Likert-type scale response options were strongly agree, agree, disagree, and strongly disagree; statements regarding knowledge and trust included a “don’t know” option; and trust statements regarding pediatricians and naturopathic/chiropractic doctors included options for “I don’t have a pediatrician yet” and “I don’t see this type of doctor,” respectively. Specific vaccine safety concern statements were administered only to participants who expressed a lack of confidence in the safety of a particular vaccine using automated survey skip logic. Sociodemographic information such as education and ethnicity was collected.
Data Analysis
LCA was performed to identify homogeneous groups of women based on their responses to the aforementioned survey items, all of which were included in the analysis (except for the 16 specific vaccine safety concern statements which were condensed into one ordinal variable representing the number of concerns identified to avoid missing data). Starting at two, the number of clusters was sequentially increased by one, and the Lo–Mendell–Rubin likelihood ratio test (chosen among other model fit criteria prior to analysis so as to avoid the potential for a biased decision) was performed at each iteration to determine the number of classes that best fit the model while remaining as parsimonious as possible.
LCA was performed using Mplus Version 8 (Muthén & Muthén, Los Angeles, CA, USA). Probabilities of latent class based on responses to survey items was then calculated using Stata/IC 12.1 (STATA Corp., College Station, TX, USA).
Results
Study Population
Of the 3,904 pregnant women found to be eligible for participation after screening, 2,196 (56%) agreed to participate and took the survey (Table 1). Reasons for eligible women declining study participation included being too busy to screen (18%), not being interested in the study (40%), being wary of the study (5%), and not being able to communicate or read in English (13%).
Latent Classes of Pregnant Women Stratified by State, Education, Ethnicity, and Prior Children.
p Value for the Pearson chi-squared proportion test at significance level of (α) 5%; boldfaced if significant.
Graduate degree includes master’s, doctoral, and professional degrees; undergraduate degree includes bachelor’s and associate’s degrees.
Roughly half of participants were from each state, and 46% were first-time pregnant women. Of women who provided education information (n = 1,812), 27% had a graduate (master’s, doctoral, or professional) degree, and 45% had an undergraduate (associate’s or bachelor’s) degree. Of women who provided their ethnicity (n = 1,862), 63% were White, 17% were Black, and 11% were Hispanic.
Intentions to Vaccinate
Sixty-three percent of pregnant women intended to receive an influenza vaccine and 65% intended to receive a Tdap vaccine during their pregnancy. Fifty-six percent of women intended to receive both vaccines, 15% intended to receive neither vaccine, and 13% were unsure. Sixty-eight percent of women intended their baby to receive all recommended vaccines on time. Twelve percent of women intended their baby to receive all recommended infant vaccines but intended to spread out the vaccine schedule past the recommended ages. Five percent of women intended their baby to receive only some vaccines on time, and 3% intended their baby to receive only some vaccines spread out past the recommended ages. Two percent intended their baby to receive no vaccines, and 9% were still unsure.
Latent Class Analysis
Per the Lo–Mendell–Rubin likelihood ratio test, the three-class model fit the data statistically significantly better than a two-class model (−2 times the log-likelihood difference: 10,817, degrees of freedom: 169, p value: <.01), and the four-class model did not fit the data statistically significantly better than a three-class model (−2 times the log-likelihood difference: 5,344, degrees of freedom: 169, p value: .79). Thus, three latent classes of pregnant women were established based on their vaccine intentions, confidence in vaccine safety, and trust in vaccine information sources (Table 1), as well as their vaccine knowledge, attitudes, and beliefs (Table 2). Although not used in the analysis to generate these latent classes, significant differences in latent class were found by sociodemographic characteristics such as education, race/ethnicity, and prior children (Table 3). The probability of being in each latent class based on each survey item response is outlined in Tables 4 and 5.
Vaccine Intentions, Confidence in Vaccine Safety, and Trust in Sources of Vaccine Information Within Latent Classes of Pregnant Women.
Vaccine Knowledge, Attitudes, and Beliefs Within Latent Classes of Pregnant Women.
Twenty questions deemed nonessential were included in surveys for only ~75% of the sample (based on randomly assigned groups) to keep surveys short enough to obtain high completion rates.
Don’t know; provided as an answer option only for knowledge- and trust-based questions.
Probability of Latent Class Based on Pregnant Women’s Vaccine Intentions, Confidence in Vaccine Safety, and Trust in Sources of Vaccine Information.
Probability of Latent Class Based on Pregnant Women’s Vaccine Knowledge, Attitudes, and Beliefs.
Twenty questions deemed nonessential were included in surveys for only ~75% of the sample (based on randomly assigned groups) to keep surveys short enough to obtain high completion rates.
Don’t know; provided as an answer option only for knowledge- and trust-based questions.
The first class identified by this model, containing 36% of pregnant women, consisted of women with strong positive attitudes toward vaccination and was characterized as “vaccine supporters.” Eighty-one percent of women in this class intended to get both recommended maternal vaccines, and 91% intended to get all recommended vaccines for their baby on time. The majority of vaccine supporters chose “strongly agree” for statements assessing attitudinal constructs, such as confidence in vaccine safety and efficacy, VPD severity, self-efficacy, pro-vaccine norms, and having enough vaccine information, and for statements assessing trust in vaccine information from obstetricians or midwives, pediatricians, the Centers for Disease Control and Prevention, and universities.
The largest class (41%) consisted of women with mostly positive attitudes toward vaccination but stated with less conviction and more variability and was characterized as “vaccine acceptors.” Sixty-two percent of women in this class intended to get both recommended maternal vaccines, 9% intended to get influenza vaccine but not Tdap vaccine, 11% intended to get Tdap vaccine but not influenza vaccine, 5% intended to get neither maternal vaccine, and 14% were unsure. Seventy-three percent intended to get all recommended vaccines for their baby on time, 16% intended to get their baby all recommended vaccines but some spread out past the recommended ages, and 4% were unsure. The majority of vaccine acceptors chose “agree” for statements assessing attitudinal constructs, such as confidence in vaccine safety and efficacy, VPD severity, self-efficacy, pro-vaccine descriptive and injunctive norms, and having enough vaccine information, and for statements assessing trust in vaccine information from obstetricians or midwives, pediatricians, the Centers for Disease Control and Prevention, and universities.
The smallest class (23%) consisted of women with more common negative attitudes toward vaccination and was characterized as “vaccine skeptics.” Only 7% of women in this class intended to get both recommended maternal vaccines, whereas 49% intended to get no maternal vaccines, and 26% were unsure. Likewise, only 26% intended to get all recommended vaccines for their baby on time, whereas 9% intended to get no vaccines for their baby, and 29% were unsure. Vaccine skeptics most frequently chose “disagree” or “don’t know” for statements assessing attitudinal constructs, such as confidence in vaccine safety and efficacy and VPD susceptibility, although many also chose “agree” for statements assessing attitudinal constructs, such as VPD severity and self-efficacy. Vaccine skeptics also demonstrated a variety of levels of trust in sources of vaccine information.
Discussion
Our LCA produced three groups of pregnant women: “vaccine supporters,” or those with strong positive vaccine attitudes (36% of women); “vaccine acceptors,” or those with moderately positive vaccine attitudes (41%); and “vaccine skeptics,” or those with negative vaccine attitudes (23%). The major difference between the supporters and the acceptors were whether they mostly “strongly agreed” or just “agreed” to Likert-type scale survey items assessing their vaccine attitudes and beliefs. The skeptics most frequently chose “disagree” or “don’t know” for items assessing attitudinal constructs such as confidence in vaccine safety and efficacy and VPD susceptibility. However, even skeptics often chose “agree” for items assessing VPD severity and self-efficacy. This may indicate that these constructs are less important to address when trying to increase vaccine acceptance; that even those who recognize the dangers of VPDs and who are confident in their ability to get vaccinated may choose not to vaccinate due to other factors, such as believing themselves unlikely to contract VPDs and/or concerns about the vaccines themselves. On the other hand, perhaps this existing recognition of VPD severity among many vaccine skeptics could serve as a starting point for changing other attitudes and beliefs by elaborating on the connections between these constructs and rooting the conversation in established common ground.
The findings of our analysis are similar in some ways to previous analyses but different in others. Similar to the LCA among Swiss mothers performed by Weiss et al. (2016), our LCA identified three classes. However, the description of the three classes and the percentages in each differ dramatically. The classes in the Swiss LCA were positive attitudes (58%), fearful/uncertain attitudes (28%), and critical attitudes (14%). Only one of their groups were described as having positive attitudes, whereas two of ours were strong positive and moderately positive. Combining our two positive groups leads to 77% with moderately to strongly positive attitudes, which is much greater than their 58%; similarly, combining their fearful/uncertain and negative attitude groups leads to 42%, much greater than our 23% with negative attitudes. These discrepancies could be due to many factors, including differences in nationality and culture, differences between current pregnant women and mothers of young children, and differences in the nature of the survey questions analyzed.
Comparing with the non-LCA analyses of parents performed by Gust et al. (2005) and Leask et al. (2012), our analysis identified fewer homogeneous groups (three vs. five). However, when combining similar groups, all three analyses depicted about one third as those with strong positive immunization attitudes, less than a third with serious reservations, and over a third without strong feelings in either direction.
A one-size-fits-all approach is unlikely to be useful for educational interventions about vaccines, especially considering the potential for nontailored messaging to be ineffective or even backfire among certain subgroups (Nyhan & Reifler, 2015). Many people are already fully supportive of vaccines (Chung et al., 2017; Freed et al., 2010; Frew et al., 2016; Kennedy et al., 2011), and interventions need to be cautious to not unnecessarily create uncertainty among this population. Among those with concerns, there is wide variability in the amount and specifics of those concerns, as well as varied source credibility (Freed et al., 2010). Audience segmentation would allow for tailoring vaccine messages to address only those concerns that already exist without oversaturating or creating new concerns. For vaccine supporters, short messages reinforcing the importance of vaccines and the risk of disease along with cues to action are likely to be more than sufficient. Vaccine acceptors may need one or two specific concerns addressed and may benefit from more detailed information or stronger cues to action than vaccine supporters. Vaccine skeptics may have multiple specific concerns and want more detailed information regarding vaccine safety and efficacy and VPD susceptibility but may also have less trust in typical sources of vaccine information, such as health care providers, federal agencies, and academic institutions, increasing the importance of tailoring based on source credibility.
Although both our analysis and the analyses mentioned above find that those with negative vaccine attitudes compose the smallest of the groups, this group is by no means insignificant. Nearly 25% (one in four) pregnant women in our sample were described as “vaccine skeptics.” Considering the increased demand for detailed information and need for carefully tailored messaging among this group, the amount of time and effort required from health care providers to adequately discuss vaccines with these patients is substantial, especially for prenatal care providers who may not have as much experience with such discussions as practitioners like pediatricians. Nevertheless, it is worth recognizing that compared with the other groups, the vaccine skeptics group had the greatest relative proportions of women stating uncertainty about receiving maternal vaccines and recommended childhood vaccines. Over a quarter said they were still unsure about maternal vaccines, and nearly 30% said they were unsure about vaccines for their baby. Similar to the “fence-sitters” from Gust’s research (Gust et al., 2005), these women’s mind-sets may still be malleable, leaving room for the possibility of becoming more open to vaccines, or at least to learning more about them. Despite having substantially lower levels of trust in doctors and scientists than the other two groups, about three quarters of vaccine skeptics said they trusted their doctors and more than half said they trusted university scientists (Table 1), illustrating the potential of providing them evidence-based vaccine information from trusted clinicians and reputable academic institutions.
There are several limitations of this article. First, these data are not nationally generalizable. Although the study sites were chosen to capture as wide a range of demographics and vaccine hesitancy as possible, the sample consists solely of pregnant women from two states who were willing to participate in our study; participating pregnant women may be different than those who chose not to participate and pregnant women in general. Our study population contained a higher proportion of educated, White women than indicated by the Centers for Disease Control and Prevention data on the demographics of U.S. births in 2016 in Colorado and Georgia (Martin et al., 2018). Some women in the sample did not complete the survey and thus questions near the end of the survey had slightly lower response rates than questions toward the beginning. It is also possible that some women answered survey questions dishonestly just to finish the survey faster, although we aimed to avoid that as much as possible by presenting the survey in practice waiting rooms while women were otherwise waiting for their appointment. Finally, this analysis does not link the latent classes with vaccination outcomes, so we cannot say whether these classes affect vaccine behavior despite differences in vaccine attitudes and beliefs. We plan to perform further research to determine the impact of these classes on vaccination outcomes as these data become available through the larger project.
Educational interventions for pregnant women regarding vaccination should be tailored to account for the differences described in this analysis. More analyses such as the one in this article are needed to support these efforts across different populations and settings.
Footnotes
Acknowledgements
We would like to thank everyone who contributed to survey design and/or participant recruitment in this study.
Authors’ Note
Clinical Trial Registration: clinicaltrials.gov registration number NCT02898688.
Declaration of Conflicting Interests
The author(s) declared the following potential conflicts of interest with respect to the research, authorship, and/or publication of this article: R. J. Limaye, S. B. Omer, S. T. O’Leary, M. K. Ellingson, C. I. Spina, S. E. Brewer, R. A. Bednarczyk, F. Malik, P. M. Frew have no conflicts and report no financial disclosures. M. Z. Dudley received some support from Walgreens. A. T. Chamberlain received paid consultancy with the American College of Obstetricians and Gynecologists regarding provider-to-patient communications. D. A. Salmon received consulting and/or research support form Merck, Walgreens, and Pfizer.
Funding
The author(s) disclosed receipt of the following financial support for the research, authorship, and/or publication of this article: This work was supported by the National Institutes of Health (Grant number R01AI110482). The funder had no role in the design and conduct of the study; collection, management, analysis, or interpretation of the data; or preparation, review, or approval of the manuscript.
