Increasing Vaccination: Putting Psychological Science Into Action

One problem is inadequacy, which we define as vaccination coverage that falls below a stated vaccination goal. For example, the United States has established Healthy People 2020, a set of objectives for health behaviors, including coverage targets for various vaccines. Coverage can be problematically low for people who face substantial barriers or actively refuse vaccination (Omer, Salmon, Orenstein, DeHart, & Halsey, 2009; Rainey et al., 2011). The relative contribution of these barriers depends on the country and the health-care system in place there. Receipt of the complete set of recommended childhood vaccines is suboptimal in most countries. More than a quarter of 19- to 35-month olds in the United States do not have all recommended vaccines (Hill, Elam-Evans, Yankey, Singleton, & Dietz, 2016). Vaccine coverage for adolescents and adults is lower still (CDC, 2016). Older adults do the best of any age group on uptake of the seasonal-influenza vaccine, but less than a third get the new shingles vaccine in the United States (W. W. Williams et al., 2017), in part because insurance coverage for the vaccine is spotty and, although only doctors can prescribe it, doses are available primarily at pharmacies.

Health-care providers regularly come into contact with people who are old, sick, or have compromised immune systems. Carrying infectious diseases such as influenza or hepatitis to these groups can be deadly. Health-care workers in 11 European countries have seasonal-influenza vaccination coverage below 30% (P. R. Blank, Schwenkglenks, & Szucs, 2009). Coverage also varies by role and setting. In the United States, seasonal-influenza vaccination coverage is above 90% for workers in hospitals, 80% in outpatient care, and below 70% in long-term care (Black et al., 2016).

Explicit refusal of vaccines when offered can lead to low coverage, as can inaction, such as not going to a provider or ignoring vaccination campaigns. Some U.S. parents (6%–25%) have refused one or more vaccines for their children (Gilkey, McRee, et al., 2016; Gilkey, Reiter, et al., 2016; Gust, Darling, Kennedy, & Schwartz, 2008; P. J. Smith et al., 2011). A large minority of parents who initially refuse a vaccine agree to it later (Dempsey et al., 2011; Kornides, McRee, & Gilkey, in press). Actively refusing all vaccines is rare, typically around 1% to 2% in high-income countries (Beard, Hull, Leask, Dey, & McIntyre, 2016; Dempsey et al., 2011; Hill, Elam-Evans, Yankey, Singleton, & Dietz, 2016; Samad et al., 2006; P. J. Smith et al., 2011). Refusal of all vaccines is more common among White children with older, university-educated mothers living in higher income households (Samad et al., 2006; P. J. Smith et al., 2011). Vaccine refusal explains a substantial proportion of measles cases and some pertussis cases (Phadke, Bednarczyk, Salmon, & Omer, 2016), although this may vary by country, depending on access to vaccination.

Another problem is delay, which we define as receiving a vaccination after the recommended age. Low coverage reflects both individuals who will never be vaccinated and those for whom vaccination is delayed but eventually occurs. Like low coverage, delay can be the result of a deliberate choice, passive inaction, or forces external to the individual, such as a vaccine shortage. Over a third of U.S. infants remain undervaccinated for 6 months or longer past the recommended age (Luman et al., 2005). A quarter of U.S. parents report having delayed one or more vaccines for young children (Gilkey, McRee, et al., 2016). Some parents spread out doses according to alternative schedules for reasons that include concerns about pain or safety (Dempsey et al., 2011; Gust et al., 2008). Young children whose parents intentionally delay one or more vaccinations are more likely to remain undervaccinated (Smith, Humiston, Parnell, Vannice, & Salmon, 2010). Many physicians (40%) report that difficult interactions with parents about spreading out of vaccines decreases their job satisfaction (Kempe, O’Leary, et al., 2015). However, providers are also a source of delay. For example, some providers recommend delaying the HPV vaccine or other vaccines (Gilkey, Malo, Shah, Hall, & Brewer, 2015). Timeliness is also a problem in low- and middle-income countries, in part because of challenges related to supply (Akmatov & Mikolajczyk, 2012).

Claims of rising vaccine refusal or delay are commonplace in the research literature, but the data are largely indirect (Cooper, Larson, & Katz, 2008; Omer et al., 2009). The most direct example is that refusal of HPV vaccination for adolescents increased by 10 percentage points over 4 years (Gilkey, Calo, Marciniak, & Brewer, 2017). However, this conclusion relies on comparing coverage estimates made using different methods. Requested exemptions from school requirements for vaccination have also increased in the United States (Gowda & Dempsey, 2013; Omer, Richards, Ward, & Bednarczyk, 2012; P. J. Smith et al., 2017), although studies do not show contemporaneous drops in vaccination. In Australia, registered objection to vaccination remained stable between 1999 and 2015 (Beard et al., 2016). “Shot limiting” (having two or fewer vaccines at a doctor’s visit before the age of 9 months) increased over a 3-year period and was associated with more medical visits and fewer vaccine doses delivered (Robison, Groom, & Young, 2012). Undervaccination among 2-year-olds increased over a 4-year period in a study of eight U.S. managed care organizations (Glanz et al., 2013), although national data did not show corresponding decreases. The number of pediatricians who experienced a parent’s refusal of vaccines went up by 13 percentage points over 7 years in the United States (Hough-Telford et al., 2016). Refusal of required childhood vaccines experienced a small increase (0.5%) in the first 2 years of a national U.S. study but was flat in the final years of the study (Omer, Porter, Allen, Salmon, & Bednarczyk, 2017). In sum, we are unaware of compelling evidence of substantially increased refusal and delay, and recent U.S. data suggest that refusal may be stable.

A third problem is instability, which we define as a sudden drop in coverage, often as a result of a vaccine-safety scare. Although instability also has some overlap with low coverage, the emphasis is on reductions in coverage over a relatively short time. For example, in Japan, unconfirmed reports of safety issues facing the HPV vaccine surfaced in the media (Hanley, Yoshioka, Ito, & Kishi, 2015). Within 2 months, Japan’s government removed a proactive recommendation for the HPV vaccine, which caused uptake to plunge from around 70% to less than 1%. Denmark and Ireland have also seen large declines in HPV vaccination following the lobbying of groups of parents formed through a shared belief about the role of the vaccine in any health problems their children experienced. Italy had a substantial drop in seasonal-influenza vaccination after a safety scare (Rosselli, Martini, Bragazzi, & Watad, 2017). Given that safety scares emerge with regularity, national vaccination programs need practical advice about what they can do to prepare for specific, unanticipated vaccine-safety scares and to maximize the robustness of uptake.

In the context of vaccination, risk appraisals focus on the infectious agent and the harm it can cause. Widely studied risk appraisals include perceived likelihood (how likely a person is to get infected) and perceived severity (how bad the infection would be). Meta-analyses have shown that both these risk beliefs are associated with vaccination behavior (Brewer, Chapman, et al., 2007) and other heath behaviors (Floyd, Prentice-Dunn, & Rogers, 2000; Harrison, Mullen, & Green, 1992). Across 12 studies, people who perceived higher likelihood of harm (r = .26) were more likely to receive a vaccination. Likewise, among people with higher perceived severity of harm, vaccination was more likely, although the association was smaller (r = .16 across 32 studies). Some researchers suggest perceived vulnerability (feeling personally susceptible) is separate and distinct from beliefs about likelihood or severity. This belief is also associated with receiving a vaccination (r = .24 across 15 studies; Brewer, Chapman, et al., 2007).

Research on risk appraisals has also focused on anticipated regret, the expectation that an unpleasant outcome will lead a person to wish they had made a different decision (Sandberg & Conner, 2008). Researchers have proposed that anticipated regret is a primary motivator of receiving a vaccination. A recent meta-analysis largely supported this speculation (Brewer, DeFrank, & Gilkey, 2016). Anticipating that one would regret a decision not to get vaccinated was associated with receiving a vaccination (r = .27) in 18 studies. Anticipated regret stood out as a stronger predictor of intentions than other types of anticipated affect (e.g., anticipated guilt). Anticipated regret was also generally a stronger predictor of intentions and behavior than other risk appraisals, such as perceived likelihood, perceived severity, and worry (Brewer et al., 2016). This may be because the construct captures a combination of affect and cognition or because anticipated regret taps a natural process of imagining the consequences of a vaccination decision.

Another conceptualization of risk appraisal is a lack of concern about vaccine-preventable diseases (H. J. Larson et al., 2014). As advances in vaccine development make vaccinations more effective, fewer people have personal experiences with the diseases. It is unclear whether this lack of experience lowers perceived risk or mutes affective reactions to the disease. Furthermore, because some vaccines, such as the seasonal-influenza vaccine, are repeated periodically (Gierisch, Reiter, Rimer, & Brewer, 2010), people receive ongoing feedback from past vaccination behavior and its consequences (or lack thereof). Not receiving a vaccination and experiencing no repercussions can reinforce the decision not to get vaccinated in the future (Kahn & Luce, 2006). Low concern remains an understudied area in part because no standard measure exists and studies to capture the phenomenon are difficult and costly to conduct. Disease outbreaks due to low vaccination can increase uptake immediately thereafter (Oster, 2016).

An empirical approach is to look at how different risk appraisals intercorrelate. Such an approach led to the tri-risk model, which identifies three correlated risk components, each of which has unique predictive value with respect to health behaviors (Ferrer, Klein, Persoskie, Avishai-Yitshak, & Sheeran, 2016). The first component is deliberative risk, which includes likelihood beliefs, such as beliefs about chance, probability and risk. A second component is experiential risk, which includes susceptibility beliefs, vividness, and gut-level impressions. These first two components correspond loosely to the constructs of perceived likelihood and perceived vulnerability described previously. As reviewed above, these beliefs are associated with receipt of various vaccines among many different populations.

A third component of the tri-risk model is affective risk, which includes worry, anxiety, and fear. Worry motivates people to get vaccinated. People are more likely to get the seasonal-influenza vaccine if they worry about getting the flu or expect that getting vaccinated would make them worry less (Chapman & Coups, 2006; Weinstein et al., 2007). The finding holds in various populations such as healthy adults, asthmatic children, pregnant women, and gay and bisexual men (Gorman, Brewer, Wang, & Chambers, 2012; Reiter, McRee, Katz, & Paskett, 2015; Szilagyi, Rodewald, Savageau, Yoos, & Doane, 1992; Tucker Edmonds, Coleman, Armstrong, & Shea, 2011) and with various vaccines, such as the seasonal-influenza vaccine, the HPV vaccine, and the pandemic influenza vaccine (Bish, Yardley, Nicoll, & Michie, 2011; Setbon & Raude, 2010). Few studies have examined the role of fear of infection or disease as a motivator of vaccination, and there has yet to be a meta-analytic assessment of the role of affect in motivating vaccination.

A perennial question is whether new vaccines, by making people feel protected and thus less at risk, encourage people to “spend” that risk surplus on unhealthy patterns of behavior. This phenomenon, sometimes called risk compensation, disinhibition, or licensing, is especially problematic when vaccines are imperfectly effective and people overestimate the protection they receive. For example, getting the Lyme disease vaccine (which is no longer on the market but was ~70% effective) could confer a false sense of security and prompt a drop-off in protective behavior. A study with U.S. adults found that Lyme disease vaccination indeed reduced perceived likelihood of infection, but it did not reduce self-protective behavior such as checking for ticks after being outdoors (Brewer, Cuite, Herrington, & Weinstein, 2007). Likewise, many studies have found that girls and women who have received the HPV vaccine are not more likely to engage in sexual behavior, become pregnant, or get sexually transmitted infections compared with those who have not received the vaccine (Bednarczyk, Davis, Ault, Orenstein, & Omer, 2012; Liddon, Leichliter, & Markowitz, 2012). If anything, those who received the HPV vaccine were more likely to use condoms when having sex (Liddon et al., 2012). Evidence for licensing effects in moral behavior suggests the effect is small at best (Blanken, van de Ven, & Zeelenberg, 2015). Thus, disinhibition as a result of vaccination is unlikely.

Although ample evidence points to a correlation between risk appraisals and vaccination, few intervention studies provide evidence for a causal link. A handful of interventions that have successfully changed risk appraisals have increased vaccination coverage (Sheeran, Harris, & Epton, 2014). Across five studies of tetanus and influenza vaccination that successfully changed risk appraisals (including perceived likelihood, perceived severity, and negative affect), the pooled effect size was d = 0.33, a moderate-sized intervention effect. No intervention studies isolating the effect of changes in anticipated regret on vaccination have been published. However, a recent national television campaign in the United States played on anticipated regret to motivate parents to get the HPV vaccine for their adolescent children, and studies have used interventions theorized to harness anticipated regret to motivate other health behaviors, such as weight loss (Volpp et al., 2008).

Research on various health behaviors has found that fear-appeal interventions are generally effective in changing behavior and are especially effective when people experience both heightened fear and a feeling of efficacy for taking action to ameliorate the threat (Tannenbaum et al., 2015). Reviews have not identified studies showing that fear communication increases vaccination coverage (Tannenbaum et al., 2015). For example, in a classic study, eliciting fear increased intention to receive a tetanus vaccination but did not affect actual vaccination uptake (Leventhal, Singer, & Jones, 1965).

However, fear communication can elicit anger and other forms of message reactance (Brehm, 1966; Hall et al., 2016, 2017). Those reactions are associated with having less intention of and not getting vaccinated (Betsch & Böhm, 2016; Leventhal et al., 1965). Thus, fear communication may increase intentions to get vaccinated, whereas anger toward the message may undermine its impact.

Risk appraisal research has yielded strong evidence for correlates of vaccination. Of particular interest is that anticipated regret is likely the risk appraisal with the strongest correlation to vaccination behavior. However, the area has yielded few insights into how to increase vaccine uptake, in part because vaccination interventions based primarily on risk appraisals are few. More than 50 years have passed since the first small trial used a fear appeal to motivate vaccination (Leventhal et al., 1965), yet we are aware of few experiments that have pursued this question in the interim (e.g., Ordoñana, González-Javier, Espín-López, & Gómez-Amor, 2009).

Understanding how best to measure vaccine confidence is a priority globally. It is an important indicator of success in meeting objectives in the WHO (2013) Global Vaccine Action Plan. Their proposed indicators include “percentage of countries that have assessed (or measured) confidence in vaccination at subnational level” and “percentage of un- and under-vaccinated in whom lack of confidence was a factor that influenced their decision” (WHO, 2013, p. 92). The National Vaccine Advisory Committee in the United States has also recognized the importance of vaccine confidence surveillance (National Vaccine Advisory Committee, 2015). For this reason, we address the issue of measuring confidence in some depth.

One way to think about vaccine confidence is as a nested set of beliefs and attitudes. A United States government working group defined vaccine confidence as the trust that parents and providers have in recommended vaccines, providers who give vaccines, and processes that leads to vaccine licensure and national vaccination schedules (National Vaccine Advisory Committee, 2015). European researchers have similarly argued that confidence suggests “trust in the vaccine (the product), trust in the vaccinator or other health professional (the provider), and trust in those who make the decisions about vaccine provision (the policymaker)” (H. J. Larson, Schulz, Tucker, and Smith 2015, screen 2). We focus here mostly on thoughts and feelings about vaccines, leaving a more extensive discussion of trust in providers to Section 3.

Scales for measuring vaccine confidence typically have subscales for vaccine benefit (e.g., perceived effectiveness), vaccine harm (e.g., side effects, safety), and sometimes trust in providers. A large number of vaccine confidence measures exist for specific vaccines. For example, a confidence scale specific to the HPV vaccine is the Carolina HPV Immunization Belief and Attitude Scale (CHIAS; McRee, Brewer, Reiter, Gottlieb, & Smith, 2010). The 16-item scale includes subscales labeled “effectiveness,” “harms,” “uncertainty,” and “barriers.” A 17-item scale for influenza vaccine confidence, validated in three countries, has subscales for “influenza perceptions,” “influenza vaccine perceptions,” “trust in vaccine stakeholders,” and “social influence” (Wheelock, Miraldo, Thomson, Vincent, & Sevdalis, 2017).

Relatively few scales exist to measure attitudes toward vaccines in general. One scale, the Vaccination Confidence Scale, has eight items that assess vaccine “benefits,” “harms,” and “trust in healthcare provider” (Gilkey et al., 2014). A four-item short form consists of the perceived benefits subscale and performs as well as the entire scale, suggesting that harms and trust beliefs may be less important in these scales (Gilkey, Reiter, et al., 2016). The scale’s psychometric properties have been favorably evaluated with national probability samples containing a total of 18,977 U.S. parents of young children and adolescents. Another scale, the Parent Attitudes about Childhood Vaccines, has 15 items that assess “safety/efficacy,” “general attitudes,” and “behavior” (Opel et al., 2011, 2013). The scale’s psychometric properties have been evaluated with a convenience sample of 437 U.S. parents of young children.

Some researchers have used single-item measures or multiple items analyzed individually. A consortium based in the United Kingdom used four items (efficacy, safety, importance, and religious beliefs). The measures of perceived vaccine effectiveness, safety, and importance were highly correlated. In their survey of 65,819 individuals in 67 countries, vaccine confidence was quite high overall (H. J. Larson et al., 2016). The lowest endorsement of vaccines as being safe was in European countries, especially in France. Across the countries, confidence was higher in high-income countries and in countries with lower mean education; at the level of respondents, confidence was higher among those with higher incomes and education.

Finally, some researchers are working on ways to assess vaccine confidence using traditional and social-media indicators. The Vaccine Sentimeter uses natural-language processing to analyze data from mainstream media sources and Twitter (Bahk et al., 2016). Others have used machine learning to characterize vaccine sentiment of online posts, including Twitter (Dunn et al., 2015). In sum, (a) the field has many measures of confidence, (b) measuring positive and negative attitudes toward vaccination is necessary, and (c) measuring confidence in specific vaccines adds predictive power beyond general vaccine confidence.

Confidence is associated with vaccine uptake (Schmid, Rauber, Betsch, Lidolt, & Denker, 2017), including in numerous prospective studies, but the conditions that strengthen or weaken the association are not yet well understood. In the context of seasonal-influenza vaccination, a systematic review found that uptake was associated with higher perceived effectiveness in all 17 studies that assessed the construct (Chapman & Coups, 1999a). In the context of HPV vaccination, perceived effectiveness has been associated with uptake in some studies but not others (e.g., Brewer et al., 2011; Gilkey et al., 2017; Reiter, McRee, et al., 2013; Reiter, Brewer, Gottlieb, McRee, & Smith, 2009). Studies have shown mixed findings for an association between vaccination and perceived harms and relatively few associations with trust in providers (Gilkey, McRee, et al., 2016; Gilkey, Reiter, et al., 2016). Many confidence studies have used patient or parent self-report of vaccination behavior, but some have used large national probability samples with confirmation of vaccination by clinic records and have reached similar conclusions (Gilkey, McRee, et al., 2016; Gilkey, Reiter, et al., 2016).

Interventions appear to be able to increase vaccine confidence, but the impact of increased confidence on uptake is unknown. A randomized controlled trial (RCT) in Pakistan found that discussion groups increased positive attitudes toward childhood vaccination (R. Amin et al., 1997). The trial also found increased uptake for two of three children’s vaccines, although no mediation analysis was performed. A systematic review found that educational interventions (including brochures, pamphlets, and posters) increased vaccination confidence in 8 of 15 studies (Sadaf, Richards, Glanz, Salmon, & Omer, 2013); none of the studies assessed impact of confidence on vaccine uptake.

Other studies of educational interventions found little or no effect on vaccination. A Cochrane review concluded that face-to-face educational interventions had uncertain effects on vaccination (Kaufman et al., 2013). Systematic reviews for the CDC’s Community Guide to Preventive Services said evidence was insufficient to recommend education programs that were community wide (which may include small or mass media, person-to-person interactions, and community mobilization) or clinic-based client education. A systematic review of community-focused interventions identified two trials in low-income countries (Saeterdal, Lewin, Austvoll-Dahlgren, Glenton, & Munabi-Babigumira, 2014). An RCT in India found that public meetings and leaflet distribution increased the number of children who had received at least one vaccination and increased tetanus vaccination among pregnant women (Pandey, Sehgal, Riboud, Levine, & Goyal, 2007). Neither clinic-based nor community-wide education has been shown to be effective in increasing vaccination, according to systematic reviews (Community Preventive Services Task Force, 2015).

Several studies conducted since these reviews were published show the heterogeneity of study designs and findings. An educational effort using a video and written information in primary-care clinics targeted parents with low vaccine confidence (S. E. Williams et al., 2013). The intervention increased positive attitudes about vaccination but not did not increase vaccine uptake. An online experiment showed that providing information about vaccine safety and efficacy can have unpredictable effects on concerns about vaccination and may even increase such concerns in some circumstances (Nyhan, Reifler, Richey, & Freed, 2014). Another online experiment found that, in terms of increasing vaccine confidence, correcting autism myths was ineffective but describing the disease risk that vaccination can address was effective (Horne, Powell, Hummel, & Holyoak, 2015). Finally, a series of experiments showed that strong statements that vaccines do not cause risk can backfire and actually increase concerns about vaccine harm, but only when the source of the information is a pharmaceutical company (Betsch & Sachse, 2013).

Effective interventions targeting people low in confidence do not yet exist. For example, no evidence has established a best practice for providers to address parents’ concerns and questions about vaccination. Several suggested approaches based on communication skills taught in medical curricula have the general format of eliciting parents’ concerns, acknowledging them without judgment, providing information, and recommending vaccination (Silverman, Kurtz, & Draper, 2005; Table 3). Perhaps the earliest such approach is the CASE method (Table 3) developed by Singer, a vaccine advocate (Public Health Live!, 2010). Despite being widely disseminated, the CASE method has not been the subject of empirical research. The related EASE approach was one part of a communication training that increased vaccine uptake, but the evaluation did not isolate whether EASE itself was effective (Brewer et al., 2017). The ask-acknowledge-advise approach was ineffective in changing parent vaccine confidence in a trial that trained providers (Henrikson et al., 2015).

Table 3. Structured Approaches for Clinicians to Address Vaccination Concerns

Table 3. Structured Approaches for Clinicians to Address Vaccination Concerns

View larger version

Research from cognitive and social psychology suggests several important steps to counteract misinformation that may be useful when addressing vaccination myths (Lewandowsky, Ecker, Seifert, Schwarz, & Cook, 2012). First, understand people’s mental models of vaccination (Downs et al., 2008), identify how the rumor fills a gap in the model, and offer an alternative explanation. Second, use repeated statements to counteract the misinformation, but avoid repeating the incorrect statement, which can make it seem familiar and true. Third, emphasize what is true to avoid a familiarity backfire effect that further entrenches the myth in people’s minds. Fourth, warn listeners before sharing a myth to make them less likely to be influenced by it. Fifth, use materials that are engaging, simple, and in plain language. As Lewandowski and colleagues (2012) stated, “If the myth is simpler and more compelling than your debunking, it will be cognitively more attractive, and you will risk an overkill backfire effect” (p. 123). Sixth, consider whether your material might threaten closely held values of your audience. If it might, focus on potential benefits instead of risks and harms, consider using self-affirmation (Sherman, Nelson, & Steele, 2000), and consider using approaches that change behavior directly without challenging people’s values (see Section 4).

Perhaps the most important advice is to instill an accurate understanding of vaccination to before people develop false beliefs that, once established, can be difficult to correct (Jolley & Douglas, 2017). Useful insights come from fuzzy-trace theory (Blalock & Reyna, 2016; Reyna, 2008) which suggest that aiding the creation of meaning (i.e., gist) allows people to remember and later take action on the information (Reyna, 2012). Antivaccine messages often impart a memorable idea (or gist) and use emotion to add meaning to their “facts” (Reyna, 2012). The use of first-person narratives in antivaccine messaging may be another reason for its potency (Winterbottom, Bekker, Conner, & Mooney, 2008, but see Bekker et al., 2013). To have impact, official communications about vaccines should have a clear take-home message, tell a memorable story, and elicit feeling (Shelby & Ernst, 2013).

A great deal of research on vaccine confidence has accumulated without yielding insights for increasing vaccination coverage. Many dozens of confidence measures exist, and they often correlate with vaccine uptake. Yet we do not know how to reliably increase vaccine confidence. Interventions to increase confidence through persuasion and education have had no appreciable or reliable effect on vaccination coverage. Despite the international fascination with vaccine confidence, the direct impact of increased confidence on vaccine uptake is unknown.

Motivation is associated with a broad range of behaviors, both those related to health and those unrelated (for a review of systematic reviews, see Webb & Sheeran, 2006). The association is reliable and large (d = 1.47). For example, among parents who intended to get the HPV vaccine for their adolescent daughters, 38% did so in the next year compared with 10% who did not intend to get their daughters vaccinated (Brewer et al., 2011). However, a sizable gap exists between intentions and behavior (Sheeran, 2002). Although many parents in the example above intended to have their daughters vaccinated and did so, the majority of parents did not act on their favorable intentions (Brewer et al., 2011). The example illustrates a common finding of asymmetry in the link between intentions and behavior, sometimes called literal inconsistency (Fishbein & Ajzen, 2010). Among parents who did not intend to have their children vaccinated, 90% met those intentions; in contrast, among parents who did intend to have their children vaccinated, only 38% did so (Brewer et al., 2011; DiBonaventura & Chapman, 2005). Inaction is an easier goal to meet.

People have many ideas of actions they want to take to be healthier, but they often do not follow through. A large literature has examined conditions that undermine the association of intentions and behavior (Sheeran & Webb, 2016; Webb & Sheeran, 2006), including such barriers as vaccine shortages (Brewer et al., 2011; DiBonaventura & Chapman, 2005) and logistical obstacles (Witteman, Chipenda Dansokho, et al., 2015). For example, during the shortage of seasonal-influenza vaccine in 2004–2005, many people who attempted to get vaccinated were turned away. Intentions assessed in the fall of 2004 were associated with later vaccine uptake among those not turned away, but the two variables were unassociated among those who had been turned away at least once (Brewer et al., 2011). Other moderators include the stability of one’s intentions over time, with stronger associations for people with more stable intentions (DiBonaventura & Chapman, 2005).

Interventions can address hesitancy in several different ways. First, an intervention can be designed to increase motivation to be vaccinated. Interventions that successfully increase intentions reliably increase many health behaviors (Sheeran & Webb, 2016; Webb & Sheeran, 2006). Many studies have successfully increased people’s intentions to get vaccinated. We are aware of no studies demonstrating that increasing intentions or reducing hesitancy increases vaccination coverage (Jarrett et al., 2015; Webb & Sheeran, 2006). Researchers have proposed motivational interviewing as a way to address vaccination hesitancy (Leask et al., 2012). Communication skills that are part of this approach focus on working with resistance, identifying motivation to be vaccinated, and evoking and reinforcing change talk (Miller & Rollnick, 2013). Motivational interviewing is effective in many contexts (Rubak, Sandbæk, Lauritzen, & Christensen, 2005), but studies do not yet show that it increases vaccination coverage. An RCT of motivational interviewing in a pediatric clinic showed an increase in HPV vaccination that was not statistically significant (51% vs. 56%; Joseph et al., 2016), and an uncontrolled study of the technique did not increase vaccination uptake in a community pharmacy but did increase readiness for two of five vaccines examined (Brackett, Butler, & Chapman, 2015). Challenges to routine use of motivational interviewing include the length of training (a median of 9 hr), although briefer trainings are in development (Lundahl et al., 2013). In addition, vaccination is but one of many issues to cover in medical visits; such visits last only 18 min on average for adults in the United States (Olson et al., 2004), but even very brief motivational interviews take 10 to 20 min (Lundahl et al., 2013).

Second, an intervention can target a group of parents with a known motivation profile (e.g., hesitant). No published interventions designed to target hesitant parents have led to an increase in vaccination coverage (Sadaf et al., 2013; S. E. Williams, 2014), although some targeted interventions have changed intentions, which the authors interpret as a proxy for behavior. Third, interventions can focus on parents who intend to have their children vaccinated and try to close the intention-behavior gap (Sheeran & Webb, 2016; Webb & Sheeran, 2006). Interventions to increase the correspondence of intentions and behavior include minimizing logistical barriers and addressing implementation intentions, which we discuss in a later section on changing behavior directly (Section 4).

To help patients make informed health-care choices, decision aids describe treatment options’ benefits and harms, often including numerical estimates of their likelihood or magnitude; guide people to clarify the value they place on those benefits and harms; and encourage people to make a choice (Stacey et al., 2017). Decision aids were originally developed for situations with clinical equipoise— that is, situations in which options are available, but no single option is best, given the best evidence. The absence of a best action for a population makes a “best” decision one that is based solely on the preferences of the individual patient. Decision aids address the general call for more informed and participatory health-care interactions.

We suggest that it is inappropriate to offer vaccination decision aids as part of routine clinical practice with all patients. Vaccination is not a situation of clinical equipoise: Evidence strongly favors vaccination, national recommendations clearly indicate who should be vaccinated, and the option to not be vaccinated affects people beyond the individual decider. In the context of patients who have already expressed vaccination hesitancy, decision aids could be a helpful adjunct to communication from health-care providers.

Studies of decision aids have not tested them specifically in hesitant populations. However, RCTs with general populations show limited effects on vaccine uptake but positive effects on intentions and attitudes. In an RCT with pregnant women in New Zealand, a decision aid about the measles-mumps-rubella (MMR) vaccine did not increase women’s self-reported vaccination of their infants (Wroe, Turner, & Owens, 2005), but it did increase their intentions to vaccinate and vaccination timeliness; it also reduced anxiety about vaccination. In an RCT with parents in the United Kingdom, an online decision aid did not increase MMR vaccination; uptake was 100% in the intervention group and 99% in the usual-care group (Shourie et al., 2013). However, the decision aid reduced people’s internal conflict about the decision. In a pre-post uncontrolled study in Australia, an evaluation of an earlier version of the United Kingdom decision aid found increased positive parental attitudes toward MMR vaccination (Wallace, Leask, & Trevena, 2006). In an RCT with parents in the United States, giving risk information and an online values-clarification exercise did not increase parental self-report of having their children vaccinated against seasonal influenza in the 6 months after the exercise (Witteman, Chipenda Dansokho, et al., 2015). The intervention increased parents’ intentions to have their children vaccinated, especially among parents hesitant to do so. In an RCT in the United States, a decision-support tool increased the number of physicians who themselves got hepatitis B vaccine or screening (Clancy, Cebul, & Williams, 1988). In an RCT, a decision aid led to more positive HPV vaccination attitudes, but the study did not assess behavior (Kennedy, Sapsis, Stokley, Curtis, & Gust, 2011). In an RCT with health-care personnel in Canada, a decision aid improved “confidence” and reduced uncertainty but did not increase intention to get the seasonal-influenza vaccine (Chambers et al., 2012).

In summary, one trial has shown that decision aids increase vaccination uptake and several others have not; one showed increased vaccination intentions and another increased timeliness. Although studies have observed increased knowledge and satisfaction and reduced decisional conflict in the limited evaluation of vaccination decision aids to date, the importance of these outcomes for increasing vaccination coverage is unclear. Some of the trials had relatively small samples (~100) and thus were underpowered to detect even modest differences. Moreover, the limited samples precluded examining whether decision aids were more effective for people who are hesitant about vaccination (but see Witteman, Dansokho, et al., 2015). The many limitations of the studies make any conclusion tentative.

Having received a vaccination in the past is a reliable predictor of receiving a vaccination in the future. Having received an influenza vaccine strongly correlates with getting the vaccine in a subsequent year (Chapman & Coups, 1999a; Schmid et al., 2017). Receiving one recommended vaccine is associated with being more likely to receive other recommended vaccines (Kessels et al., 2012), an insight that is useful in the context of new vaccines and pandemics (Bish et al., 2011). The pattern holds across parents and children (Robison & Osborn, 2017). Although past behavior is an important correlate, the past behavior itself can be the result of various psychological mechanisms, several of which we discussed earlier, as well as structural factors. In other words, past behavior should reflect people’s thoughts and feelings about vaccination to the extent that these constructs explain why people got vaccinated in the past (Weinstein, 2007). This means that it may be more productive to study and intervene in the psychological antecedents of vaccination.

Much of the material in this section on thoughts and feelings about vaccination presumes that people are rational creatures who thoughtfully seek out information and then impartially and accurately make sense of it. However, people’s busy lives mean that they cannot spend limitless time seeking out all possible information; information is overly abundant even for the casual seeker, and making sense of it all can require more cognitive capacity than people have at any one time. One result is that people take mental shortcuts, called heuristics. These generally save time and effort but also lead to predictable irrational errors, called biases (Tversky & Kahneman, 1974).

One such bias is the omission bias, a preference for inaction even when taking action is substantially more beneficial (Ritov & Baron, 1990, 1999). In the context of vaccination, people give disproportionate weight to harms of receiving a vaccination and dismiss harms that accrue from not receiving a vaccination. For example, when a hypothetical scenario attributed symptoms to a vaccine adverse event rather than to a disease, parents said that symptoms would be more severe and last longer (Brown et al., 2010). Exhibiting the omission bias is associated with lower parent willingness to have their children get the pertussis vaccine (Asch et al., 1994) and with lower vaccination coverage (Meszaros et al., 1996). Likewise, seasonal-influenza vaccination is less common among healthy adults who showed the omission bias (DiBonaventura & Chapman, 2008). Some have hypothesized that the omission bias may be a marker for negative attitudes toward vaccination or anticipated regret (Connolly & Reb, 2003). To date, the omission bias has not been the target of interventions to increase vaccination uptake.

Another bias that may affect vaccination is the confirmation bias, the tendency to seek out or overweight information that is consistent with one’s original hypothesis (Klayman, 1995). This tendency to favor one’s original position and to find evidence consistent with that view may help to explain why those opposed to vaccination maintain that position despite the sound evidence presented to them by others. A classic study (Lord, Ross, & Lepper, 1979) demonstrated that presenting evidence does not cause individuals to bring their beliefs more in line with the evidence but instead causes them to hold their original beliefs more strongly. Presenting the same evidence to two groups with polarized views does not bring their viewpoints closer together but actually increases the polarization between the groups. Cultural cognitions that reflect strongly held moral values shape beliefs about some vaccines, making those beliefs difficult to change (Kahan, 2013; Kahan, Jenkins-Smith, & Braman, 2011). In the words of a vaccine-hesitant person in an online discussion who rejected evidence of vaccine safety, “The science is tainted by bias. Do some more reading” (Rodriguez, 2016).

This section reviews research on thoughts and feelings most commonly studied in the context of vaccination. We acknowledge that this focus on some constructs but not others is somewhat artificial, excluding, for example, values and self-efficacy (A. B. Amin et al., 2017). Many of these other constructs are correlated with vaccination (Bish et al., 2011). However, to our knowledge, none has been the basis of effective interventions and thus would not meaningfully change this section’s main conclusions.

A goal for many health-care organizations is full vaccine coverage of their staff, but few reach this goal without extensive interventions that often include requirements for vaccination (see Section 4). Health-care workers’ reasons for receiving or not receiving a vaccination are similar to those of the general population (Capolongo, Dibonaventura, & Chapman, 2006). Research suggests that the same psychology that motivates the general public to get vaccination also motivates health-care providers, including risk appraisals and vaccine confidence (Corace et al., 2016; Herzog et al., 2013).

Although health-care provider recommendations are influential, psychological science has not yet fully explained why they motivate vaccination uptake. Any of the mechanisms described in this section on thoughts and feelings can provide reasonable explanations. If a provider recommends HPV vaccination, it may cause people to see the hazards as more likely and severe or to see themselves as more susceptible. A recommendation could increase beliefs in the vaccine’s safety and reduce concerns about harm. Any of these changes would presumably increase motivation to get a vaccination. Presumptive announcements that a child is due for vaccination may be the most potent form of provider recommendation (Brewer et al., 2017; Opel et al., 2015), as we discuss in Section 4.

Providers must address parents’ vaccination questions and concerns many times a day, yet no evidence-based practice exists for doing so effectively. Most currently suggested approaches involve some element of acknowledging patients’ concerns as a first step (see Table 3). Self-affirmation and motivational interviewing techniques were developed specifically to address this sort of problem (Leask et al., 2012), and at least one intervention has used motivational interviewing as part of a multicomponent trial to increase vaccination uptake (A. F. Dempsey, personal communication, May 14, 2017). Psychological research on reactance, debunking myths, and resistance to new information offers additional insights for such interventions (Hall et al., 2016, 2017; Lewandowsky et al., 2012).

Up to this point, we have focused on the psychology of the individual. However, people are social creatures whose thoughts and feelings also bear on their social context. We next turn to understanding how such social processes affect vaccination.

Most vaccination decisions take place in the context of a relationship. Patients typically receive vaccinations from a physician, nurse, or pharmacist. People consistently rate these three professions among the highest for honesty and ethical standards (Gallup, 2016). Trust is a key component of vaccine confidence (Gilkey et al., 2014); however, variation in trust of health-care providers does not appear to explain variation in vaccination coverage (Gilkey, McRee, et al., 2016; Gilkey, Reiter, et al., 2016). Thus, it is not clear that trust in providers is the reason that provider recommendations of vaccination are so influential: Patients who feel less trust are not any less likely to get vaccinated. Trust in government bodies who set vaccination schedules and recommendations may be a more important determinant of vaccination uptake, or at least motivation. Indeed, trust in government predicts H1N1 vaccination (Freimuth, Musa, Hilyard, Quinn, & Kim, 2014), and trust in the U.S. Food and Drug Administration (FDA) predicts stated willingness to take antiviral medication (Quinn, Hilyard, Castaneda-Angarita, & Freimuth, 2015).

Parents and other surrogates decide, with providers’ advice, whether minors will receive vaccinations because laws often prevent minors from giving consent. As offspring enter adolescence, they become more involved in vaccination decisions. Half of teens play some role in decisions to get the HPV vaccine, and by age 17, a quarter make the decision on their own (McRee, Reiter, & Brewer, 2010). Even so, many teens are uncomfortable making vaccination decisions without consulting their parents (Kennedy, Stokley, Curtis, & Gust, 2012). Some parents see vaccination decisions as an opportunity to increase children’s involvement in their health-care decisions as they grow into adulthood. Although many interventions target parents or other caregivers, we know little about how interventions differently affect people who are deciding about vaccines for themselves or others.

In studies of families, adolescents and their parents have quite similar attitudes about HPV vaccination (Moss, Reiter, & Brewer, 2015; Vietri, Li, Galvani, & Chapman, 2012). It would be difficult to examine agreement in vaccination attitudes between parents and younger children, but the data from adolescents suggests that parents provide a good proxy for their children’s vaccination preferences.

Research on decision making in general has demonstrated that being in the surrogate role can alter preferences. People value the future consequences of health decisions (e.g., disease prevention) similarly whether they make the decisions for themselves or for someone else (Cairns & van der Pol, 1999). Some evidence shows that people are more risk averse when making decisions for others than when deciding for themselves (Roszkowski & Snelbecker, 1990). In addition, surrogates tend to choose more aggressive treatment for others than the patients would choose for themselves, an overtreatment bias (Fagerlin, Ditto, Danks, Houts, & Smucker, 2001). Surrogates’ greater risk aversion and greater interest in treatment compared with people choosing for themselves could explain why parents are more positive about vaccinating their adolescent children against HPV than the children are about getting vaccinated (Vietri et al., 2012). Thus, the many vaccination decisions that parents make on behalf of their children are of no small importance.

Multiple aspects of family relationships may shape vaccination behavior. For example, dynamics within a couple have important influences on health behavior; the behavior of one person influences the health behavior of the other (Lewis et al., 2006). Classic studies have shown that individuals imitate the behavior of those around them (Bandura, 1971), suggesting that individuals may imitate the vaccination behavior of others. Thus, one family member may encourage vaccination among others through example, recommendation, or by scheduling vaccination appointments for other family members.

Correlational studies demonstrate that individuals with similar health characteristics tend to cluster together in a network. Thus, for example, smokers tend to associate with other smokers and nonsmokers with other nonsmokers. Furthermore, when one smoker quits, other network contacts are more likely to quit than they otherwise would have been (Christakis & Fowler, 2008). However, with correlational data, it is not possible to fully disentangle whether homophily or contagion is the source of these effects (Shalizi & Thomas, 2011). Few observational studies have examined homophily and contagion related to vaccination in social networks, although some have demonstrated geographical clustering of undervaccinated children (Lieu, Ray, Klein, Chung, & Kulldorff, 2015).

Modeling simulations, often built using observational data, support the conclusion that clustering exacerbates disease spread. Nonrandom mixing of individuals in a network can explain why disease outbreaks occur even when the level of immunity exceeds the herd-immunity threshold that would be needed to prevent progression of an epidemic under the assumption of random mixing (Glasser, Feng, Omer, Smith, & Rodewald, 2016). One modeling analysis of social networks demonstrated that homophily in social networks—individuals with similar vaccination views clustering together—results in pockets of disease susceptibility and increased likelihood of disease outbreak (Salathé & Bonhoeffer, 2008). A similar analysis (Eames, 2009) demonstrated that clustering of parents’ vaccination opinions results in clustering of unvaccinated children and hence more disease outbreaks. This result is sensitive to both the extent of parents’ opinion clustering and the overlap between the parents’ and children’s networks (e.g., parents who share opinions have children who interact). A final modeling article (F. Fu, Rosenbloom, Wang, & Nowak, 2011) demonstrates that imitating others within a social network can result in low vaccination coverage, under certain assumptions, because individuals end up imitating successful free riders. The role of free riding is discussed in a later section.

The spread of infectious disease requires physical contact or proximity, often a result of ongoing geographical clustering of people but also a result of travel and migration. In contrast, the spread of ideas does not require physical proximity and can occur via in-person conversation, distance communication, media, and social media. Thus, the networks that spread disease may only partially overlap the networks that spread vaccination information. Studies have found marked clustering of sentiment about vaccination in social-media conversations about vaccination (Dunn et al., 2015). Indeed, such conversations can be tracked as an indicator of social-network support for vaccination (Bahk et al., 2016; Centola, 2013).

Social contagion and disease contagion can interact when the social network and geographical cluster overlap (Bauch & Galvani, 2013). For example, a highly connected node in the social network (e.g., a celebrity) could spread suggestions that vaccines are risky, resulting in the spread of beliefs and an associated decrease in the number of people who get vaccinated (social contagion). Consequently, an infectious disease could spread easily through the network (disease contagion), which may in turn prompt individuals to get vaccinated (Bauch & Galvani, 2013).

Correlational evidence from real-world behaviors such as smoking leaves open the question of whether clustering of similar people is solely the result of homophily or also results from contagion. Experimental evidence indicates that health behavior does indeed spread through social networks via contagion (Centola, 2010). In an ambitious online experiment, more than 1,500 participants were randomly assigned to one of two network structures that allowed them to interact virtually with other participants. Investigators observed the spread across the network of a specific health-relevant behavior: registering for an online health forum. Specified seed nodes in the network sent messages to their neighbors encouraging them to register, which initiated the behavior in the network. The investigators then observed how the actual participants in the study continued to spread the behavioral recommendation (and the actual behavior) across the network. One network structure—the clustered lattice, which had a high level of clustering in which each node’s neighbors were linked to one another—led to wider diffusion than the other network structure (random or “unstructured” network; Centola, 2010). A similar study demonstrated that homophily increases adoption of the new health behavior (Centola, 2011). In this study, 700 online participants were randomly assigned to either a homophilous network or an unstructured network. In the homophilous network, participants were clustered by individual characteristics (e.g., gender, age, body mass index). A few seed nodes in each network initiated use of a diet diary, and investigators observed the spread of this behavior across the network. The spread was faster in the homophilous network. These studies indicate that social networks do indeed have a causal effect on health behavior via contagion and that homophily amplifies that contagion effect.

Although carefully controlled network experiments have not examined vaccination behavior, Centola’s work suggests that vaccination attitude and behavior spread causally through social networks. Network homophily and contagion help to explain why negative attitudes toward vaccination cluster geographically. Although parents with strong principled objections to childhood vaccinations make up a tiny fraction of the population, they tend to live near other like-minded families and associate with them socially (Beard et al., 2016; Onnela et al., 2016). The geographical clustering is important because it can contribute to the spread of infectious disease (Glasser et al., 2016; Ndeffo-Mbah et al., 2012). Communities with a sizable number of vaccine refusers will lack herd immunity; consequently, if an infectious disease is introduced, it can spread easily among the community.

Studies on health social networks face some challenging methodological issues. In some studies of social networks, the participants self-report on the vaccination attitudes or other characteristics of their network members (e.g., Brunson, 2013; Nyhan, Reifler, & Richey, 2012), whereas in others, the behavior of each network member is objectively assessed (Christakis & Fowler, 2008). In other studies, the structure of the social network is experimentally controlled, and the behavior of each network node is objectively measured (Centola, 2010, 2011). The latter method provides more experimental control but less realism (for a discussion of methodological strengths and weaknesses in social-network studies, see Centola, 2013).

The evidence supports the existence of homophily and contagion as two distinct sources of clustering in networks, although data are spotty in the specific area of vaccination. Because homophily is a correlational phenomenon whereas contagion is a causal process, evidence for contagion is difficult to obtain outside of experiments in which the behavior of target nodes is manipulated (e.g., Centola, 2010). Little is known about the psychological mechanisms for contagion, but social norms are one candidate explanation.

A large number of correlational studies have linked vaccination to norms. Many of these correlational studies use the theory-of-reasoned-action and theory-of-planned-behavior paradigm to examine the association between subjective norms and vaccination intentions. In the theory of planned behavior, subjective norms are assessed as ratings of the extent to which people important to the participant want him or her to get vaccinated and thus correspond to injunctive norms. Several studies show cross-sectional correlations between injunctive norms and vaccination intentions or behavior. For example, Juraskova et al. (2012) demonstrated that subjective norms were correlated with intentions to get vaccinated against HPV. Gerend and Shepherd (2012) found an association between subjective norms and HPV vaccination behavior.

Studies also show a correlation between vaccination and perceived injunctive norms. In one, college students’ support for vaccination in their social networks predicted their intentions to receive an H1N1 vaccine (Nyhan et al., 2012). In a study by Brunson (2013), parents self-reported their vaccination decisions and the vaccination attitudes of members of their social networks. Parents who did not meet the recommended vaccination schedule believed that a larger fraction of their social network recommended nonvaccination. This predicted decisions better than demographic characteristics of the parents themselves. Health-care workers were more likely to get vaccinated against H1N1 if they reported being encouraged by family or coworkers to do so (Stokes & Ismail, 2011).

Unlike the field studies employing social-norm interventions to modify hotel towel-reuse behavior or petrified-wood collection, no intervention studies in vaccination have tested the effects of a theoretically pure social-norm manipulation on uptake. Nevertheless, the lessons from findings outside vaccination are applicable to the design of vaccination campaigns. For example, in fall 2015, Rite Aid pharmacies launched an advertising campaign to promote influenza vaccination that said “Get your flu shot today because 63% of your friends didn’t.” This message was presumably designed to encourage vaccination by emphasizing that the infection risk was high because of the large number of unvaccinated individuals with which one has contact. In contrast, the descriptive norm research, such as the petrified-wood study, would suggest that this message might actually drive vaccination coverage down (“if very few others are getting vaccinated, then I won’t either”). Indeed, a laboratory experiment varied the characteristics of a hypothetical vaccine (Hershey, Asch, Thumasathit, Meszaros, & Waters, 1994) to allow examination of the impact of perceived descriptive norms, what the authors called “bandwagoning.” Participants were more likely to say they would get vaccinated when many others were already getting vaccinated. The Rite Aid campaign was designed to counter the motive to free ride (“Because so few others are vaccinated, I can’t free ride and thus need to get vaccinated”). However, the Hershey et al. (1994) experiment findings suggest that bandwagoning is the primary behavioral factor. Consequently, the Rite Aid message may have been ill-conceived and ineffective. Their campaign was short-lived.

A few vaccination field studies have used multifaceted interventions that have social-norm components, suggesting that social norms could have a powerful influence on vaccination behavior. Health-care workers at a Swiss hospital who received the influenza vaccine were given a badge to wear that read “I am vaccinated against influenza to protect you.” (Iten, Bonfillon, Bouvard, Siegrist, & Pittet, 2013). Nonvaccinated health-care workers were required to wear a face-mask during the seasonal influenza epidemic period and to wear a badge that read “I wear a mask to protect you.” In the year that this policy was introduced, vaccination coverage increased to 37%, compared with 21% to 29% in the decade before the intervention.

Note that this study did not use a traditional descriptive or injunctive social-norm message. Instead, the badges had the potential to make salient the descriptive social norm, because health-care workers could easily see how many of their coworkers had been vaccinated. However, this could backfire and reduce vaccination coverage if the norm revealed that few people got vaccinated. In addition, it is likely that the badges set up an injunctive social norm. Finally, because the badge messages were visible to patients, they took on the role of norm-enforcers. Any health-care worker who was not wearing a vaccination badge could be negatively evaluated by patients.

In another study (Riphagen-Dalhuisen et al., 2013), similar badges with prosocial messages were used as part of a large multimodal intervention to promote influenza vaccination among hospital health-care workers. Those in the intervention condition received a multifaceted program that included education and testimonials from role models. In addition, vaccinated health-care workers in the intervention group received pins that said “deliberately vaccinated for you.” The vaccination rate was 20% in the control group compared with 32% in the intervention group. The multifaceted nature of the intervention likely contributed to its efficacy but also limits conclusions about which aspects of the intervention were effective. It is possible to speculate, however, that the use of the pins created a descriptive and injunctive norm, as described above. Note that the pins used in this intervention also highlight the prosocial motivation for getting vaccinated (protecting patients). As discussed in the next subsection, prosocial motives on their own, apart from norms, may drive people to get vaccinated. Indeed, previous research has indicated that prosocial messages affect other behaviors of health-care workers, such as hand washing (Grant & Hofmann, 2011).

Because individuals learn about descriptive social norms, in part, by observing the behavior of others, it should be possible to change social norms by sufficiently changing the behavior of the group. Thus, group behavior changed by another intervention strategy (e.g., incentives; see Section 4) could set a new group norm and thus serve to perpetuate the effect of the initial intervention. Examining the potential interdependence of different intervention strategies, in particular, as they unfold over time, is an important topic for future research.

Are people motivated to get vaccinated because of the benefit that vaccination confers to others? They sometimes say that they are. Patients and health-care workers both sometimes give prosocial reasons as a retrospective rationale when asked why they got vaccinated. A systematic review found that 30% to 60% of parents agree that community benefit was an important reason for getting their children vaccinated (Quadri-Sheriff et al., 2012). Qualitative analyses of a parents’ online chat room (Skea, Entwistle, Watt, & Russell, 2008) and parents’ focus groups (Leask, Chapman, Hawe, & Burgess, 2006) revealed that some parents do spontaneously name herd immunity and protection of others as a motivation to vaccinate their children. In a questionnaire study, health-care workers self-reported that self-interest and protecting patients were their first and second most important reasons, respectively, for getting vaccinated against influenza (Hakim, Gaur, & McCullers, 2011). Other studies with health-care workers show similar findings of altruistic motives (Bautista, Vila, Uso, Tellez, & Zanon, 2006; Christini, Shutt, & Byers, 2007).

Hypothetical scenarios provide experimental control, allowing a test of whether benefit to others affects judgments about vaccination. Several hypothetical-scenario studies have explored the respective roles of altruism and free riding as motivations for vaccination. A classic study described earlier varied whether the vaccine hypothetically protected only the self or also prevented transmission to others. The study also varied the proportion of the population that was vaccinated (Hershey et al., 1994). Although this design allowed examination of the impact of perceived descriptive norms, it also examined the impact of opportunities to free ride (the more people who get vaccinated, the less benefit I get from getting vaccinated), and the goal to act altruistically (the fewer people who get vaccinated, the more I can benefit others by getting vaccinated myself). Participants’ vaccination-intention ratings showed evidence of both altruism and free riding. As mentioned above, however, the largest effect was bandwagoning: Participants were more likely to get vaccinated when many others were already doing so.

In a hypothetical scenario study somewhat similar to the Hershey et al. study (Betsch, Böhm, & Korn, 2013), participants read scenarios that varied the information given about herd immunity. When the individual benefit of herd immunity was communicated, intentions to get vaccinated were lower. However, communicating the social benefit of vaccination reduced this tendency. These results, which are similar to findings outside the vaccination literature (Grant & Hofmann, 2011), suggest that the tendency to free ride on the vaccination of others can be reduced by communicating the social benefit that vaccination provides.

Increasing people’s understanding of herd immunity could increase vaccination coverage (because people understand the opportunity for altruistic vaccination) or decrease vaccination coverage (because the opportunity for free riding becomes clear). In a scenario study, Betsch, Böhm, Korn, and Holtmann (2017) found that explicitly communicating the role of herd immunity increased people’s intention to get vaccinated, especially among participants from Western, individualist cultures. In Eastern, collectivist cultures, intention to get vaccinated was high even without the herd-immunity communication. In another scenario study that showed an altruistic motivation for vaccination (Vietri et al., 2012), participants responded to scenarios that varied the percentage of the population that was already immune to the infectious disease. As long as participants learned that they themselves were at no risk of infection, then their stated likelihood of getting vaccinated increased as the percentage of the population that was immune decreased—indicating that more individuals would benefit from the protection provided by participants’ own vaccination.

Böhm, Betsch, Korn, and Holtmann (2016) explored prosocial motivation for health-care-worker vaccination in a hypothetical scenario study. American and Korean community members responded to a scenario in which they played the role of a health-care worker who could get vaccinated at a cost to themselves; however, if enough other health-care workers got vaccinated, patients would benefit. Vaccination responses in this prosocial scenario were higher among participants in Korea (a collectivist culture) than among participants in the United States (an individualist culture), and this difference was mediated by viewing vaccination as a social act. This study suggests that vaccination among health-care workers might be increased by promoting collectivist values and the view that vaccination is a social rather than individual behavior.

These scenario studies indicate that social preferences for altruism or for free riding may influence vaccination decisions. Furthermore, reinforcing altruistic motives may offset free riding. Although the hypothetical scenario studies in this subsection provide experimental control, they do not allow observation of actual vaccination behavior. Thus, it is unclear how these results would generalize to real-world settings.

In hypothetical-scenario studies, the main limitation is that participants’ responses have no real consequences (although research shows that preferences are frequently similar with and without consequences). Laboratory games offer a solution by providing real consequences for decisions—albeit monetary and not health consequences. In one such laboratory experiment (Chapman et al., 2012), participants engaged in an interactive strategic task in which they had to decide whether to spend points to get vaccinated. Participants were randomly assigned to age roles: For those in the “young” role, vaccination was more effective, in terms of both reducing individual risk of infection and contributing to herd immunity; for those in the “old” role, getting the flu resulted in a more severe point penalty. When the individual’s point total determined payment for being in the study, participants behaved in accordance with the self-interested prediction from game theory, with fewer “young” players than “old” players getting vaccinated. In contrast, when participants were paid according to the group’s point total, they behaved in accordance with the group optimum solution, with more “young” players than “old” players getting vaccinated. The results from this laboratory task suggest that, if the incentives are set correctly, young people may be willing to get vaccinated prosocially to protect older adults. Thus, for example, young people may be willing to get vaccinated if it is clear that their vaccination will benefit their grandparents and this is an outcome they value.

Böhm, Betsch, and Korn (2017) used a similar laboratory vaccination-game paradigm and found that participants with prosocial (rather than proself) preferences were more likely to get vaccinated in the laboratory game, and this effect was strongest among those with a positive attitude toward vaccination. The hypothetical scenarios and laboratory-game studies described thus far in this section did not assess actual vaccination behavior. Consequently, they can examine altruism and free riding in hypothetical vaccination judgments or in laboratory choices that have monetary consequences, but these results may not translate to actual vaccination decisions.

Several studies have examined whether measures of altruistic motives correlated with self-reports of actual vaccination behavior or intentions to vaccinate. Patients’ perception of the likelihood of becoming infected and their perception of the likelihood of infecting others each predict unique variance in actual influenza vaccination decisions. However risk to self is much more strongly predictive of receiving a vaccination than is risk to others, at a 3:1 ratio (Shim, Chapman, Townsend, & Galvani, 2012). Thus, to the extent that the relationship between vaccination behavior and perceived likelihood of infecting others captures altruistic motives, we can say that altruism is associated with actually getting vaccinated but plays a minor role relative to self-interest motives. This study was correlational and did not demonstrate a causal link between changes in altruistic motives and getting vaccinated.

Vaccinating boys against HPV has a notable prosocial benefit: It protects the vaccinated boys themselves from HPV-related cancers such as anal and penile cancer, and it also protects the boys’ future sexual partners from six HPV-related cancers. Polonijo, Carpiano, Reiter, and Brewer (2016) found that male teenagers were more willing to get the HPV vaccine if they found it important that the vaccine could protect their future romantic partners. Parents of the teens who held the same belief were also more willing to get them HPV vaccine. Furthermore, these prosocial attitudes mediated the relationship between race and willingness to get vaccinated, such that Blacks and Hispanics (vs. Whites) had more prosocial attitudes, and these attitudes were associated, in turn, with higher willingness to get vaccinated. Additional analyses found that perceived benefit to the son, benefit to his future sexual partners, and benefit to the community were all associated with parents’ willingness to allow the vaccination and the sons’ willingness to get vaccinated (Moss et al., 2015). However, in analyses adjusting for all three beliefs, only protecting the partner was associated with willingness.

In another study, young men gave higher ratings of hypothetical willingness to receive HPV vaccine if they received information on both the benefits to self of vaccination as well as the altruistic benefits to future sexual partners compared with (a) receiving either of these pieces of information in isolation or (b) receiving neither piece of information (a control condition; Bonafide & Vanable, 2015). This study shows a benefit of two pieces versus one piece of information but not necessarily a special role for information on the altruistic benefits of vaccination per se. In contrast, messages emphasizing the social benefits of the MMR vaccine (relative to comparison messages that omit this information) do not increase the intentions of parents to get their infant children vaccinated (Hendrix et al., 2014).

One might expect that health-care workers would be particularly motivated by the prosocial benefits of vaccination. For one thing, health-care workers’ contact patterns increase the risk of contracting infectious diseases such as influenza or hepatitis A from patients and of transmitting the infection unknowingly to other patients. In addition, health-care workers presumably have a professional duty to care for patients and protect them from adverse health outcomes, and vaccination provides one route to do that. The evidence indicates, however, that health-care workers’ vaccination decisions are not any more motivated by prosocial concerns than are those of other decision makers (Betsch, 2014). A survey study found that health-care workers had higher influenza vaccination coverage than non–health-care workers, but not because of prosocial concerns for patients (Capolongo et al., 2006). The health-care workers perceived a higher risk of infection; had easier access to the vaccine, more knowledge about the vaccine, and stronger pro-Western medical beliefs; and perceived stronger social influence to get vaccinated.

One study examined whether measures of free-riding motives correlated with self-reports of previous vaccination behavior. Free-riding motives were higher among parents who had not gotten the pertussis vaccine for their children (Meszaros et al., 1996).

In an online study, participants read a prosocial message (about someone who died from the flu because his contacts had not been vaccinated), a control message, or no message (Li, Taylor, Atkins, Chapman, & Galvani, 2016). The prosocial message increased prosocial motives, as measured by willingness to donate to an unrelated cause. Prosocial motives in turn were associated with stated intention to get vaccinated against influenza. The prosocial and control messages did not differ in their effect on vaccination intention, however, and this study did not show an effect on actual vaccination behavior.

In one of the few studies to examine the effect of prosocial motives on actual vaccination behavior (Rothan-Tondeur et al., 2010), health-care workers were randomized by ward into two conditions; in one, they received a vaccination-message intervention, and in the other, a control condition, they received no message. The message stated that vaccination of health-care workers against influenza benefits patients. The intervention did not affect vaccination coverage, which was equivalent in the two conditions. The messages were delivered via a 2-hour optional information session that health-care workers attended, which may not have been a sufficiently engaging mechanism for delivering the intervention.

A campaign to increase influenza vaccination coverage among health-care workers by emphasizing the prosocial benefits of vaccination to family members and patients had no impact on vaccine uptake (Llupià et al., 2013). Note that in this study, unlike the badge studies described earlier, the prosocial messaging did not have a public-accountability component, which may help to explain the null effect.

Vaccination coverage among health-care workers is often too low, and they are not sufficiently motivated to get vaccinated out of concern for patient safety (Lam, Chambers, MacDougall, & McCarthy, 2010). This has prompted some health-care facilities—particularly academic hospitals in the United States—to institute annual influenza vaccination requirements for health-care workers. Such requirements, sometimes called mandates, communicate a clear injunctive norm that the hospital management views staff vaccination as a priority. The requirements may also circumvent individual motives for vaccination, a topic we return to in Section 4.

The social processes explored in this section provide one class of explanation for why provider recommendations are so strongly associated with patient vaccination behavior. In network terminology, providers are likely to be centrally located in their networks, positioned to influence a large number of people through contagion. Homophily effects on social networks may result in providers being connected to patients who are the most likely to get vaccinated. For example, providers who strongly recommend vaccination attract patients with positive vaccination attitudes (Mergler et al., 2013), even as vaccine-hesitant and vaccine-refusing patients often seek out different providers and complementary and alternative medicines (Frass et al., 2012). Consequently, the strong association between provider recommendation and patient vaccination may be due in part to patient self-selection rather than to a direct causal effect of the recommendation on behavior.

Another reason for the influential role of provider recommendations is that providers can create and communicate injunctive social norms by indicating to their patients, either through explicit recommendations or implicit cues, that they expect patients to get vaccinated. Their recommendations can also communicate descriptive social norms by implying that most other patients get vaccinated or that vaccination is what most previous patients have chosen.

Social media has become an important force in organizing people in social space. Surprisingly little is known about how social media affects vaccination-related attitudes and behavior (Betsch et al., 2012). Theories reviewed in this section on social processes make interesting predictions—for example, that when people opposed to vaccination connect on social media (homophily), it should further polarize their opinions (contagion; Witteman & Zikmund-Fisher, 2012). Illustrating the value of mining social media for vaccination research, Salathé and Khandelwal (2011) coded 300,000 Twitter messages for sentiments about the H1N1 vaccine and found that sentiments expressed on Twitter in each geographical region corresponded to vaccination coverage. They also found homophily in the social network. Broniatowski, Hilyard, and Dredze (2016) examined news articles about the Disneyland measles epidemic and predicted which ones would be shared on Facebook. The articles most likely to be shared and most widely shared contained a take-home message (or easily identified gist), shedding light on a psychological construct that is key to information transmission across a network.

Social media likely magnifies the effects of homophily found in other manifestations of social networks, because social media increases the ease of finding like-minded people. To the extent that homophily amplifies contagion effects in social networks, social media would thus also magnify contagion, consequently increasing the polarization of vaccination views. As a result of these enhanced social network dynamics and the speed with which information spreads through social media, social media could serve as the catalyst for a rapid meltdown of the vaccination system and the loss of confidence built up over many years.

Personal stories are some of the most penetrating content spread on social media. The benefits of narrative storytelling favored by antivaccination activists are widely discussed yet poorly quantified (Betsch, Renkewitz, & Haase, 2013; Bruine de Bruin, Wallin, Parker, Strough, & Hanmer, 2017). As mentioned in Section 2, a large body of research in psychological science has demonstrated that anecdotes and personal stories are more influential than base rates and statistics (e.g., Fagerlin, Wang, & Ubel, 2005). Little research has examined what exactly makes antivaccine efforts specifically effective; instead, research has focused on provaccine messages. Recently, provaccination efforts have tried narrative approaches. For example, the U.S. National HPV Vaccination roundtable recently created a series of videos with survivors of HPV-related cancers. The Someone You Love documentary (Lumiere, Hefti, & Staurulakis, 2014) follows the stories of five women with cervical cancer, including two who die from the disease. No data are available on how these personal stories affect vaccination behavior, which is an important opportunity for future research.

Priorities for future research include examining the spread of rumors compared with the spread of credible information on social media and how social media affects vaccination attitudes and behavior (Reiter et al., in press). Opportunities for intervention include social-media studies that expose people to social networks with provaccination views. Other avenues for further exploration include examining how citizen activists can use social media to build support for vaccination. Finally, social media has helped antivaccination activists to find one another and to spread their message. Understanding the nature of these messages, and helping provaccination activists to use similar methods, is an important opportunity. Research on the effects of exposing people to social-media information that is at odds with their existing beliefs would be an interesting avenue of investigation. In the near future, the social-network and social-norm effects reviewed in this section are likely to play out on social media at least as much as in interactions that occur in-person.

Life’s many obligations and distractions may crowd out intentions to be vaccinated. As a result, planning may not receive the attention it needs. People who are not vaccinated as a result of passivity or inaction may find themselves in this state (Robbins et al., 2010). We consider three intervention strategies designed to increase the likelihood that people will keep their favorable intentions in mind: reminder/recall systems, presumptive vaccine recommendations, and primes such as implementation intentions.

Even when people’s intentions to get vaccinated are at the top of their minds and not forgotten, they may not lead to behavior because barriers—real or perceived—may emerge that make taking action difficult (Gerend, Shepherd, & Shepherd, 2013; Kimmel, Timko Burns, Wolfe, & Zimmerman, 2007). Situational barriers include lack of transport, inadequate maternity leave or childcare, multiple competing priorities, inconvenient clinic opening times, service or vaccine cost, or lack of vaccine availability (Falagas & Zarkadoulia, 2008; Hollmeyer, Hayden, Poland, & Buchholz, 2009; Holman et al., 2014). Their health-care provider may recommend a delay in vaccination because of a sick child or may not take opportunities to vaccinate at sick visits (Smith, Marcuse, Seward, Zhao, & Orenstein, 2015). Language barriers as well as disabilities (e.g., deafness) can make it difficult for people to interact with care systems or obtain information, potentially leaving them unaware that a vaccination is due (Fiscella, Franks, Doescher, & Saver, 2002; McKee, Barnett, Block, & Pearson, 2011). Finally, some people may experience judgment or discrimination from providers (Trivedi & Ayanian, 2006; but see Hausmann, Jeong, Bost & Ibrahim, 2008) or fear that vaccination will be painful (Falagas & Zarkadoulia, 2008; E. Mills, Jadad, Ross, & Wilson, 2005). Taken together, these factors can impede action, creating a gap between people’s inclination to be vaccinated and their behavior. Logistical strategies that can blunt the effect of these barriers should enable people to act on their favorable intentions, thereby increasing vaccination coverage.

Incentive and sanction programs have emerged as a potential strategy for shifting people’s analysis of the costs and benefits of engaging in a desired pattern of behavior. Incentive programs provide people with monetary or nonmonetary rewards contingent on the performance of a specified behavior (e.g., being vaccinated), whereas sanction programs enact monetary or nonmonetary penalties contingent on the failure to perform the behavior (for overviews, see Giles, Robalino, McColl, Sniehotta, & Adams, 2014; Kane, Johnson, Town, & Butler, 2004).

Given this framework, incentives and sanctions offer the most practical value in contexts in which people are not engaging in the desired behavior at a sufficient rate. A key feature of these programs is that they target a set of outcomes that are distinct from the thoughts and feelings about vaccination that are traditionally targeted by intervention programs. Thus, this intervention strategy can be effective without having to change people’s underlying thoughts and feelings about the vaccine.

To date, programs for vaccinations have either provided an incentive (e.g., a cash payment) for engaging in the behavior or have implemented a sanction by limiting an existing benefit for failing to engage in a behavior (e.g., reducing a government payment). However, the overwhelming majority of the programs provide an incentive for obtaining a vaccine; thus, this will be the primary focus of this section.

Incentive programs have focused primarily on promoting childhood vaccination (e.g., Banerjee, Duflo, Latif, Glennerster, & Kothari, 2010; Bond, Davie, Carlin, Lester, & Nolan, 2002; Browngoehl, Kennedy, Krotki, & Mainzer, 1997; Kerpelman, Connell, & Gunn, 2000; Mantzari, Vogt, & Marteau, 2015; Minkovitz et al., 1999) and influenza vaccination for adults (e.g., Bronchetti, Huffman, & Magenheim, 2015; Moran, Nelson, Wofford, Velez, & Case, 1996). However, incentives have also been used to promote hepatitis B vaccine in adult subpopulations (e.g., injection drug users: Seal, 2003; Topp et al., 2013; Weaver et al., 2014). The reviews of incentives diverge in their recommendations: Some assert that this approach is an effective strategy for promoting vaccination (Community Preventive Services Task Force, 2015, 2016a, 2016b); others contend that the evidence for effectiveness is inconclusive (Adams et al., 2015). A consistent theme across these reviews is the heterogeneity in the nature and quality of the underlying studies, which makes it difficult to draw any strong conclusions regarding this intervention strategy. For instance, many tests of incentive programs have been conducted in combination with other intervention strategies, making it difficult to isolate the effect of the incentive. Furthermore, the incentives used in these programs vary considerably in size, by whether they are monetary or nonmonetary, and by whether they provide a fixed reward or variable reward (e.g., a lottery ticket). Currently, investigators have little guidance regarding how to set the magnitude of the incentive, whether it matters if the reward is fixed or variable, or whether the reward should involve monetary or nonmonetary outcomes. A more thorough understanding of how features of the program affect people’s behavioral decisions could help policymakers optimize incentives to make them more effective and efficient (Burns et al., 2012).

Only a small number of programs have tested the effect of introducing a sanction for not receiving a vaccination. In these cases, the focus has been on childhood vaccination programs, and the sanction involved a reduction in the financial support that families would otherwise receive from the government. Thus, the potential adverse outcome is the absence of a benefit rather than the imposition of an additional cost. Although Kerpelman et al. (2000) found that the potential loss of benefits led to higher rates of compliance with childhood vaccination schedules, Minkovitz et al. (1999) found that a similarly structured program had no effect on vaccination coverage. The introduction of a system to link vaccination requirements and certain exemptions to receipt of government payments was associated with increased vaccination coverage in Australia (Lawrence, MacIntyre, Hull, & McIntyre, 2004). The questions raised regarding incentive-based reward programs are relevant here as well. For instance, whether the proposed sanction is sufficiently averse to motivate the behavior requires careful consideration. To date, no study has tested the effectiveness of a program in which the sanction involved imposing an additional cost for failing to be vaccinated, which might be experienced as a more aversive outcome.

Moving forward, analyses of incentive-program effectiveness should also consider the broader consequences of linking an incentive to vaccination behavior. A prevailing concern regarding incentive programs is that they have the potential to undermine people’s intrinsic motivation for the incentivized behavior. To date, this has not been shown to be an issue when using incentives to motivate healthy behavior (Promberger & Marteau, 2013). Given that incentive programs are not designed to change how people think and feel about vaccines, there is little reason to believe that providing an incentive for one vaccine will increase the likelihood that people will obtain other vaccines. In fact, if anything, it may alter what people expect to receive in exchange for being vaccinated, making them less responsive to calls for vaccinations that do not provide incentives.

Consideration should also be given to what the implementation of an incentive program for vaccination might inadvertently communicate about vaccines (Connelly, Certo, Ireland, & Reutzel, 2011). On the one hand, an incentive program may affirm the importance of vaccination. People may think that paying people to take action or penalizing them for inaction signals the value of this behavior. On the other hand, the program may signal that presently people are not choosing to perform the behavior, creating or reinforcing the impression that obtaining the vaccine is not normative (Section 3). Moreover, the observation that people need to be paid (or potentially penalized) in order to agree to be vaccinated could be taken to imply that the case for vaccination, on its own, is not sufficient to motivate action. These latter inferences may serve to undermine people’s interest in the vaccine (Gneezy, Meier, & Rey-Biel, 2011).

An interesting question is for whom an incentive program might be most effective. Given that an incentive is designed to improve people’s cost-benefit analysis of performing the behavior, it may provide limited added value when offered to people who already hold favorable beliefs about the vaccine. However, it seems well suited for people with ambivalent or unfavorable beliefs about the vaccine because the incentive serves as a counterweight to these beliefs. For instance, Bronchetti et al. (2015) observed that providing an incentive for getting the seasonal-influenza vaccine had a more pronounced effect on the behavior of students who had not gotten the vaccine in the prior year.

Finally, the acceptability and feasibility of implementing an incentive program to promote vaccination are important considerations. Several studies have observed that people express discomfort with the notion of paying people to take care of their health (Adams et al., 2015). Regarding feasibility, incentive programs are likely to be viable only within care systems that have the resources to cover the costs of the incentive. Health-care systems could also build incentives into quality-metrics programs. Finally, checklists are available for thinking through how to make incentive programs acceptable and effective (Glasziou et al., 2012).

Another strategy to shape behavior is creating requirements that share some qualities with defaults and sanctions. Requirements clearly set vaccination as a default and impose a sanction or penalty on those who want to opt out. Some jurisdictions have policies that require vaccination as a precondition for access to resources or activities tied to school or work or, in some cases, even residency. In this manner, vaccination requirements are similar in structure to sanction programs that impose penalties for failing to take action. Most requirements are instantiated through legal action by national or regional governments, although some come from employers and providers. All have exemptions for people medically contraindicated for vaccination, and some provide other exemptions relating to personal beliefs or religion. The level of difficulty in attaining these exemptions varies.

Requirements also vary considerably between jurisdictions. Australia requires parents to have their children fully vaccinated or to have obtained an approved exemption (medical or otherwise). The personal-belief or so-called conscientious-objection option was removed from the eligible exemptions in 2016 (Leask & Danchin, 2017). Some Australian states have vaccination requirements for childcare entry, not for school attendance, but legislation enables exclusion of unvaccinated children during an outbreak in both school and childcare. Other countries, such as Slovenia, have highly comprehensive programs, including requiring vaccination against nine diseases; people can submit medical-exemption requests to a committee. Philosophical exemptions are not allowed, and failure to comply results in a fine (Walkinshaw, 2011). Other countries have more modest or disease-specific immunization requirements. Belgium, for example, requires polio vaccination but has not had to enforce the policy because of extremely high rates of vaccination (Walkinshaw, 2011). In 2017, Italy began requiring 12 vaccines. The World Health Organization currently has no official recommendation on vaccination requirements.

The United States requires immigrants to have 14 vaccines but requires no vaccines of travelers and temporary visitors. People infected while outside the United States are responsible for most domestic measles cases (CDC, 2011). All U.S. states require evidence of vaccination against at least some diseases as a condition of school entry or access to childcare (Barraza, Schmit, & Hoss, 2017), but states vary considerably in the vaccines that they require and ages or grade levels to which the policies apply. All U.S. states grant medical exemptions, which are for individuals who have a medical condition that is a contraindication to a vaccine. Forty-eight U.S. states and the District of Columbia offer religious exemptions from school vaccine requirements (Diekema, 2014; Yang & Silverman, 2015), and 20 U.S. states allow philosophical or religious exemptions, which often require only that an individual declare a personal belief opposing vaccination. Only three U.S. states, West Virginia, Mississippi, and California, have laws specifically excluding religious or philosophical exemptions from vaccination requirements.

Vaccination requirements can also be policies that private institutions enact. Most U.S. universities require vaccination in some form and limit exemptions (Noesekabel & Fenick, 2017). Hospitals may require workers to receive certain vaccines as a condition of employment. Physicians may also enforce a form of requirement by refusing to treat children who do not get recommended vaccines (Flanagan-Klygis, Sharp, & Frader, 2005).

A key element of initiatives designed to improve people’s health is that they rely on people initiating changes in their behavior (e.g., diet, physical activity, substance use) and maintaining those changes. Yet, the development of strategies that can elicit sustained changes in behavior has proven challenging (Rothman, Baldwin, Hertel, & Fuglestad, 2011; Rothman, Sheeran, & Wood, 2009). However, some people can maintain the behavior over time, and for them, the behavior becomes a routinized response to a cue or condition in the environment (Wood & Neal, 2016; Wood & Rünger, 2016). In some cases, routinized, repeated behaviors seem to occur with little conscious thought (e.g., putting on your seatbelt every time you drive a car), whereas in other cases, completing the behavior may take conscious thought, but the initial sequence of action is triggered by cues in the environment (e.g., seeing athletic shoes next to the bed in the morning can prompt people to exercise).

To be up to date with the recommended vaccine schedule, people must receive a large number of vaccinations, which may require multiple doses or annual repetition. Successfully navigating this complex set of behaviors can be challenging, yet some people regularly obtain all of the vaccines available to them. People who successfully obtain all, or nearly all, available vaccines may have turned vaccination into a routinized, repeated behavior (Gierisch et al., 2010) that takes limited deliberation.

Consistent with the view that regular vaccination can become routinized, people’s vaccination behavior shows consistency. In particular, prior vaccination behavior is one of the strongest predictors of future uptake of the same vaccine or another vaccine for oneself and one’s children (Schmid et al., 2017). Although the predictive value of prior vaccination behavior would suggest that people have the potential to develop a vaccination routine, several aspects of vaccination behavior are likely make it difficult for a routine to emerge. First, vaccination is not fully under one’s own control; one must receive the vaccine from a provider. Second, the time period between vaccination events is highly variable and in many cases can be quite long, which may make it difficult to form a consistent routine. Third, a routine is more likely to develop when a behavior is repeated under similar conditions (e.g., running first thing in the morning on every weekday). However, the conditions under which vaccinations are obtained are often heterogeneous and vary in their timing, setting, and provider; this is particularly likely across different types of vaccines because one might get a seasonal-influenza vaccine at work but a shingles vaccine in a doctor’s office.

The observation that prior vaccination behavior is a strong predictor of subsequent vaccination may offer important insights into the factors that regulate this behavior and have the potential to support the development of a vaccination routine. A history of prior vaccination behavior may be a marker for individual-level factors that support vaccination, such as the beliefs that underlie vaccine confidence, trust in the health care system, and a perception of vaccination as the normative behavioral response. A prior history may also be a marker for the presence of structural factors that support vaccination, such as the availability of resources or services provided through one’s workplace or health care system. Thus, consistent vaccination behavior may represent the persistent presence of factors that support vaccination behavior each time the opportunity arises.

To the extent that prior vaccination history serves as a marker for the presence or absence of these supportive conditions, it may provide an opportunity for investigators to deploy intervention resources more effectively and efficiently. People who do not have a history of vaccinations may need to be persuaded of the value of a given vaccine and thus are more responsive to intervention strategies that provide information about, or explanations for, vaccination. On the other hand, people with a history of prior vaccinations may not need to be convinced about the value of vaccination, and thus it would be sufficient to rely on intervention strategies that provide reminders or put in place structural changes that make it easier to access vaccination services.

Research on the psychological determinants of vaccination behavior has focused almost exclusively on the explicit (conscious) thoughts and feelings that people report regarding vaccination (e.g., appraisals of risk, confidence in vaccination). This focus on conscious thoughts and feelings is consistent with prevailing models of health behavior (Conner & Norman, 2015; Rothman & Salovey, 2007), and compelling evidence shows that experimentally elicited changes in conscious thoughts and feelings can elicit changes in behavior (e.g., Sheeran et al., 2016, 2014). Yet because the observed magnitude of these effects have proven modest, investigators have begun to explore the hypothesis that people also hold implicit (or nonconscious) thoughts and feelings about behavior that may affect their behavioral decisions (Marteau et al., 2012; Sheeran et al., 2016; Sheeran et al., 2013). As noted earlier, these implicit thoughts and feelings are associated in memory with the behavior, but they are not readily available to conscious awareness.

Research on implicit processes has focused on the determinants of addictive or hedonic behaviors such as smoking, drinking, drug use, and unhealthy dietary choices (see Sheeran & Bosch, 2016). This work has been guided by the premise that people may explicitly report that they hold unfavorable thoughts or feelings toward cigarettes, beer, or doughnuts, but at the same time hold implicit thoughts or feelings that are more favorable. Moreover, under certain conditions (e.g., times of stress, reductions in executive function or self-control), these implicit beliefs can have a more pronounced influence on behavior (Strack & Deutsch, 2004).

A similar discrepancy may exist between people’s explicit and implicit thoughts and feelings about vaccination. People’s explicit thoughts and feelings may indicate a favorable evaluation of vaccination, whereas their implicit thoughts and feelings may reveal a more unfavorable or ambivalent evaluation of the behavior. For people whose thoughts and feelings are marked by this discrepancy, their favorable (explicit) thoughts and feelings should guide their behavior when they have sufficient time and mental energy to make decisions. However, their more unfavorable (implicit) thoughts and feelings should guide their behavior when they are under stress or other conditions that constrain the time and energy they can direct toward their decisions. This pattern of predictions might help elucidate the factors that underlie inconsistencies in people’s vaccination behavior over time. Formative research is needed to examine people’s implicit and explicit thoughts and feelings regarding vaccination, their relation to each other, and the conditions under which they are predictive of vaccination behavior.

If evidence emerges in support of the predicted, differential influence of implicit and explicit thoughts and feelings on vaccination behavior, investigators may want to pursue the development of strategies that can mitigate the adverse effect of people’s implicit thoughts and feelings. These strategies could take one of two approaches. One would be to develop strategies that support the activation and use of people’s (more favorable) explicit thoughts and feelings about vaccination, especially when they are under stress or are grappling with multiple demands on their time. Another would be to develop strategies that inhibit the influence of implicit thoughts and feelings on vaccination behavior. For example, the formation of if-then plans to promote the use of deliberation in situations in which people might typically react impulsively has been shown to reduce the predictive effect of implicit attitudes on behavior (Webb, Sheeran, & Pepper, 2012). In either case, these approaches should increase the likelihood that people’s vaccination behavior is more likely to be consistent with their (favorable) explicit beliefs regarding the vaccine.

Consistent with the discussions in previous sections, health-care provider recommendations are strongly associated with vaccination behavior. Physician recommendations may be effective because they serve as a cue to action, but several additional factors may contribute to their effectiveness. First, to the extent that the recommendation can be acted on immediately, it sidesteps several of the logistical challenges associated with obtaining the vaccine (e.g., scheduling an appointment; making arrangements to get to the clinic). Second, a recommendation from a health-care provider may affirm or augment people’s perceptions of the value and safety of the vaccine. Third, with the recommendation, the health-care provider may lead the patient to construe vaccination as the default.

Provider recommendations are especially effective when they take the form of presumptive announcements. Such recommendations may be successful, at least in part, because of their direct effect on patient behavior. Specifically, they have the potential to make vaccination the default response or to serve as a strong prime or reminder that activates the favorable intentions patients already have. Whether recommendations are beneficial for all patients may depend on why physician recommendations are effective. For example, if recommendations increase vaccination coverage because they serve as a reminder or cue to action, they should be effective for people who already hold favorable attitudes or intentions regarding the vaccine but may show limited benefit for people with unsure or ambivalent attitudes or intentions. If recommendations instead increase vaccination coverage because they increase confidence or communicate a social norm, they may be particularly effective when directed toward people who are unsure or ambivalent regarding the vaccine.

The most promising avenues for policy implementation come from the literature reviewed in this section. Improving vaccination coverage is likely to necessitate some combination of reminders, defaults, incentives, and requirements, and other similar programs.

Most intervention approaches that we suggest are suitable for people inclined toward or perhaps ambivalent about getting vaccinated. One of the most promising tools for addressing low overage is the use of health-care-provider recommendations because they are so strongly associated with vaccination, but the specific approach to recommendations matters. Physicians should use presumptive announcements about vaccination at the outset, reserving open-ended conversations for addressing questions if parents raise them. Other tools that may be effective in raising coverage are those reviewed in Section 4 for closing the intention-behavior gap, such as the use of reminders and recalls or removing barriers to vaccination by providing standing orders. Harnessing social processes is an intriguing potential tool that merits exploration but remains poorly understood. Building provaccination descriptive and injunctive social norms among vaccine recipients and health-care professionals may increase coverage.

For people disinclined to get vaccinated, approaches include respectful engagement, such as offering decision aids and using motivational interviewing. Neither approach has evidence showing that it can increase vaccination coverage, but both build on the principle of patient empowerment, which may be meaningful for people whose opposition to vaccination is likely to generate disapproval from health-care providers. Likewise, self-affirmation can make people more receptive to risk communication (Harris & Epton, 2009; McQueen & Klein, 2006). Vaccination requirements are increasingly popular for childhood and adolescent vaccines, but they are not yet in place for adults other than health-care workers. Requirements may also be most practical if used only after most other approaches have been exhausted.

Interventions that address thoughts and feelings may offer a small impact, such as communication campaigns focused on vaccine effectiveness, vaccine safety, and the harms of not getting vaccinated, with the goal of increasing motivation. We suspect that their impact is most meaningful in early phases of a program or for rollout of a new vaccine, when awareness of a vaccine is low, and when the campaign can reach people who are already most inclined to get vaccinated. Some data suggest that educational campaigns may be effective in low- and middle-income countries but not in high-income countries (Harvey, Reissland, & Mason, 2015).

Very few evidence-based methods are available to aid clinicians in convincing hesitant parents and patients to get vaccinated. Surveys of primary-care providers demonstrate their frustration in perceiving that what they are currently doing is not very effective (Kempe, O’Leary, et al., 2015). Clearly, to better address the concerns of parent and patients about vaccines, providers need more evidence-based tools and vaccination programs need more evidence-based interventions that are practical in primary care, given time and resource constraints. Some of the methods that show promise in other medical contexts (e.g., motivational interviewing) may be difficult to incorporate into clinical practice given the amount of time required for training and use (Lundahl et al., 2013; Söderlund, Madson, Rubak, & Nilsen, 2011).

Even when parents have their children fully vaccinated, they may delay or spread out the vaccines such that the child does not complete the entire vaccination schedule by the recommended age. School requirements increase timeliness. Because parents who delay but do not refuse vaccination presumably have reasonably favorable vaccination intentions, reminders and recalls may be quite effective with this group. Default automatic appointments are also effective at bringing patients in at the desired time.

Establishing a social norm for timely vaccination may be promising, although little evidence is yet available to show how such an approach might work. Social-marketing efforts may be useful, especially if led by vaccine advocates who can counter the promotion of delay by health-care providers (e.g., the author of a popular parenting book; see Sears, 2011). As with encouraging vaccine uptake, communication is likely to be a weak tool. However, national campaigns could reinforce social-promotion efforts. If so, they should emphasize the risks of waiting to get vaccinated. Of course, one-on-one counseling by health providers remains an option, as does use of decision aids, which reduced delay in one trial.

In the past century, vaccination has eradicated one human disease (smallpox), and two others are close to eradication (polio and guinea worm). At the same time, three other disease-eradication programs experienced setbacks, in part because of the erosion or collapse of support from policymakers and challenges to vaccine confidence (Taylor, 2009). Understanding why policymakers adopt effective policies and maintain their support is an area with little evidence. It is likely that policymakers and everyday citizens are subject to similar psychological processes.

Vaccination takes place within a larger system of healthy policy, industry production, government oversight, and program delivery. Vaccination of individuals is important, as is support for national vaccine policies. Such public support can lead to stable vaccination coverage and make the national vaccination system more resilient when facing vaccine scares or spikes in antivaccination sentiment. Although risk appraisals and confidence are weak motivators of behavior, they may be important for generating and maintaining support of vaccination policies and programs (Petrescu, Hollands, Couturier, Ng, & Marteau, 2016). So while public education and communication efforts are weak interventions when targeted toward behavior, they may be useful when all else fails, and they may generate support for vaccination policies and programs and may be especially important during a safety scare. Other potentially viable approaches rely on social processes by strengthening vaccination social norms and developing provaccination activist networks. Other key areas for future research include developing methods for intervening through social media to address vaccination concerns, thereby increasing the robustness of vaccine coverage.