key: cord-0431141-8kk6ysbx
authors: Anderson, K.-A. M.; Creanza, N.
title: The Cultural Evolution of Vaccine Hesitancy: Modeling the Interaction between Beliefs and Behaviors
date: 2022-05-27
journal: nan
DOI: 10.1101/2022.05.26.22275604
sha: 8790c006b53f52692f48d8e29f59e37772c5579a
doc_id: 431141
cord_uid: 8kk6ysbx

In the last decade, despite the proven efficacy of vaccines, the developed world has seen a resurgence of vaccine-preventable diseases (VPDs) such as measles, pertussis, and polio. Vaccine hesitancy, an individual behavior influenced by historical, political, and socio-cultural factors, is believed to be a primary factor responsible for decreasing vaccine coverage, thereby increasing the risk and occurrence of VPD outbreaks. Society, culture, and individual motivations affect human decisions regarding health behavior and preventative care, and health perceptions and health-related behaviors can change at the population level as cultures evolve. In recent years, mathematical models of disease dynamics have begun to incorporate aspects of human behavior, however they do not address how evolving cultures influence these health behaviors. Here, using a mathematical modeling framework, we explore the effects of cultural evolution on vaccine hesitancy and vaccination behavior. With this model, we shed light on the facets of cultural evolution (vertical and oblique transmission, homophily, etc.) that promote the spread of vaccine hesitancy, ultimately affecting levels of vaccination coverage and VPD outbreak risk in a population. In addition, we present our model as a generalizable framework for exploring cultural evolution when beliefs influence, but do not strictly dictate, human behaviors. We show vaccine confidence and vaccine-conferred benefits can be driving forces of vaccine coverage, and we demonstrate that an assortative preference among vaccine-hesitant individuals can lead to increased vaccine hesitancy and lower vaccine coverage. Further, we show that vaccine mandates can lead to a phenomenon in which high vaccine hesitancy co-occurs with high vaccination coverage, and that high vaccine confidence can be maintained even in areas where access to vaccines is limited.

6 can better address how to mitigate VPD outbreaks by understanding the cultural dynamics of 158 vaccine hesitancy. In this manuscript, we aim to take a cultural approach to understanding the 159 evolution of vaccine hesitancy and its interactions with vaccination coverage and vaccine-160 preventable disease using a generalizable modeling framework for belief-behavior interactions. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101/2022.05.26.22275604 doi: medRxiv preprint Bm,n Probability that parental pairs vaccinate their children, which depends upon the parents' vaccination states (bm) and vaccine attitude (cn) (given in Table S2 )

Probability that parental pairs transmit vaccine confidence to their children from random mating. We define a 'choosing parent,' arbitrarily assigned as the first member of 194 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101/2022.05.26.22275604 doi: medRxiv preprint each mating pair. The choosing parent's A state dictates the level of assortative mating, that is, 195 the degree to which an individual of a given A state will preferentially mate with another 196 individual of the same state, expressed by parameters αk where k = {1, 2} and 0≤αk≤1 (Table S1) . 197

If the choosing parent is A + , this individual mates preferentially with other A + individuals with 198 probability α1, and mates randomly with probability 1−α1, whereas if the choosing parent is A − , 199 this individual mates preferentially with other A − individuals with probability α2, and mates 200 randomly with probability 1−α2. There are sixteen possible mating pairs from the four 201 phenotypes described, and we use the notation mi,j to indicate the frequency of a mating 202 between a choosing parent of type i and the second parent of type j where i, j = {1, 2, 3, 4} (Table  203 S1); in other words, m1,3 represents the mating frequency of V + A + (x1) and V − A + (x3). 204

Since the two traits in question (A and V) are transmitted vertically, for each phenotype we 205 must specify the probability that the mating produces an offspring of that phenotype. The 206 vaccine confidence trait (A + ) is transmitted with probability Cn, and the vaccine hesitancy trait (A − ) 207

is transmitted with probability 1−Cn (for n = {0, 1, 2, 3} as shown in Tables 2 and S2). If C0 = 0, two 208 A − parents will always produce A − offspring, and if C3 = 1, two A + parents will always produce A + 209 offspring. However, if C0 > 0, two A − parents can produce A + offspring at some probability, and 210 similarly if C3 < 1, two A + parents can produce A − offspring with some probability. 211

Transmission of vaccination (V + with probability Bm,n for m, n = {0, 1, 2, 3}; Table 1 ) is more 212 complex, since parents' vaccine attitudes (A), in addition to their own vaccination states (V), can 213 influence their behavior in vaccinating their offspring via a set of "influence parameters" that 214 inform vaccination probabilities. The probability that each mating pair produces an offspring with 215 the V + trait (i.e. vaccinates their offspring) is a scaled product of the influence of parental 216 attitudes (cn for n = {0, 1, 2, 3}) and the influence of parental vaccination states (bm for m = {0, 1, 2, 217 3} ) (Tables 2 and S2) . For example, for mating pair V + A + × V + A − , their combined vaccination states 218 (V + × V + ) will influence vaccination behavior by b3, and their combined attitude states, (A + × A − ), 219 will influence vaccination behavior by c2. Therefore, a V + A + × V + A − mating will produce a V + 220 offspring with probability %,' = ' (

); this pair will also produce an A + offspring with 221 probability C2 based on their combined attitude states. Thus, according to the model, this pairing 222 will produce a V + A + offspring with probability B3,2C2 and a V + A − offspring with probability 223 B3,2(1−C2). We note that assortative mating (αk>0) will increase the frequency of matings between 224 individuals that share an attitude trait, with these non-random interactions in turn skewing 225 vaccination outcomes in line with those of same-state couples (via c0 and c3). 226 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) The cultural selection pressure on vaccination is given by the parameter σ, such that the 240 frequency of the V + A + and V + A − phenotypes are multiplied by 1+σ after vertical cultural 241 transmission has occurred. At the end of each timestep, the frequency of each phenotype is 242 divided by the sum of all four frequencies, ensuring that the frequencies sum to 1. This cultural 243 selection coefficient is implemented in the same way as a selection coefficient in a population-244 genetic model, but unlike the latter, it is structured to encompass both biological fitness and 245 cultural selection pressures, including perceived risks or benefits of the vaccine itself, personal 246 cost-benefit analyses of preventative health behaviors, and the structural or societal-level factors 247 influencing vaccination rates [56, 57] . Since the frequencies of V + phenotypes are multiplied by 248 1+σ, this parameter modulates whether there are more or fewer vaccinated individuals than 249 expected: in other words, when σ>0, vaccinated individuals are more common in a set of 250 offspring than would be expected strictly by parental beliefs and vaccination statuses. We 251 calculate σ in each timestep as a function of the current vaccination coverage (frequency of V + , 252

i.e. x1 + x2), and in each simulation we specify σmax as the maximum cultural selection pressure of 253 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101/2022.05.26.22275604 doi: medRxiv preprint getting vaccinated (−1≤σmax≤1) (see the cultural selection coefficient function in Figure 2 ). This 254 function was constructed by fitting a curve to pre-specified conditions: we assume that when 255 vaccination coverage is low, the real and perceived benefits of vaccination are highest and thus 256 the cultural selection pressure is near σmax, however, as vaccination coverage increases, the 257 perceived benefits of vaccination decrease and the cultural selection pressure is reduced ( Figure  258 2). As vaccination coverage (V + ) increases in the population, and thus disease occurrence is low, the 287 benefits to being vaccinated are less obvious, while low-probability costs such as adverse 288 reactions become more apparent and could be perceived as being riskier than the disease itself. 

To compute the frequency of a given phenotype in the next iteration, we sum the 304 probability that each mating pair produces offspring of that phenotype over each of the sixteen 305 possible mating pairs. Cultural selection (σ), described above, then operates on offspring with the 306 V + trait. The full recursions, giving xiʹ phenotype frequencies in the next iteration in terms of xi in 307 the current iteration, are given in Text S1. If xiʹ is equal to xi, the system is at equilibrium. Unless 308 otherwise stated, the model is initialized with phenotypic frequencies structured to represent 309 those of the United States: x1 (frequency of V This Cn threshold value is more sensitive to σmax than to bm, cn, or Bm,n: the threshold value is 340 lowered as σmax increases (diagonal line in Figure 4A -B). Although vaccination probability (Bm,n) is 341 dependent on both cn, the influence of parental vaccine attitude, and bm, the influence of 342 parental vaccination state (Table S2) , modulating either type of influence of mixed-state parents 343 has little effect on the level of vaccination coverage and negligible effects on confidence levels at 344 each non-threshold Cn (Figure 4C-F) . 345

Interestingly, direct modulation of the mixed-trait couple vaccination probability (B1,1 = 346 B1,2 = B2,1 = B2,2) also has little power in affecting coverage and confidence levels at equilibrium 347 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (Table  359 S3). In these tests, we observe increasing equilibrium vaccination coverage as Bm,n probabilities 360 increase, with higher coverage in high-confidence transmission environments ( Figure 5 ). If we 361 vary both confidence transmission parameters and vaccination probability parameters by 362 implementing range shifts in both Cn and Bm,n, we observe an interaction between confidence 363 transmission and vaccination probability that determines vaccination coverage ( Figure S1 ). In 364 both, we confirm vaccination coverage levels are determined by an interaction between 365 confidence transmission and vaccination probability, whereas confidence levels are dictated 366 primarily by levels of confidence transmission. In sum, the degree to which parents with mixed 367 vaccine-hesitant and vaccine-confident attitudes transmit vaccine confidence instead of vaccine 368 hesitancy to their offspring is a key factor in determining population trait majorities which can 369 drastically shift population dynamics. 370 371 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

Next, we hold vaccine confidence transmission (Cn) at default probabilities, reminiscent of 397

Mendelian transmission, such that two vaccine confident or two vaccine hesitant parents 398 predictably transmit their vaccine attitude, and parents with differing vaccine attitudes each have 399 a ~50% chance of transmitting their phenotype, e.g. C0 near 0, C1 and C2 at 0.5, C3 near 1 (Table  400 1). We then varied cultural selection in combination with vaccination-associated probabilities (bm, 401 cn, Bm,n). With Cn held constant, cultural selection (σmax) is the primary factor determining 402 vaccination coverage and confidence levels ( increased Bm,n for σmax ⪅ 0.3 ( Figure 6D, F) , as well as for both increased cn and increased bm 409 ( Figure S2 . This dynamic is interesting as these parameters influence vaccination behavior, 410 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

To test whether the equilibrium phenotype frequencies were sensitive to starting 429 frequencies, we plotted the dynamics of each phenotype over time at default parameters (given 430 in Table 1 ). For each set of initial phenotype proportions tested, each phenotypic frequency 431 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101/2022.05.26.22275604 doi: medRxiv preprint dominates at lower values of these parameters, and the V + A + phenotype, which dominates at 447 higher values (Figures 8-9) . Interestingly, the "conflicting" phenotypes (when an individual's 448 attitude toward vaccinating their children does not match their own vaccination state: V -A + and 449 V + A -) are present at their highest frequencies at neutral cultural selection (σmax = 0, Figure 8B ) 450 and/or neutral confidence transmission (C1 = C2 = 0.5, Figure 9B ). Vaccinated individuals have the 451 same fitness regardless of their attitude (V + A + bears the same selection pressure as V + A -), so it is 452 worth noting that at higher levels of confidence transmission and cultural selection, V + A + 453 increases in frequency but V + Adecreases in frequency (compare Figure 8B-C, Figure 9B However, if confidence is highly transmitted (C1 = C2 = 0.8), the V + Afrequency will be reduced, as 476 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101/2022.05.26.22275604 doi: medRxiv preprint this phenotype is more likely to produce A + offspring than A -, thus increasing V + A + phenotype 477 frequencies in the population (Figure 9 and Figure S6 ). If we turn to the other conflicting 478 phenotype, unvaccinated but vaccine-confident (V -A + ) individuals become more common when 479 A + increases in frequency in the population as C1 = C2 increases from 0.1 to 0.5 (Figure 9 and 480 Figure S6 ). In contrast, higher vaccine confidence transmission (C1 = C2 = 0.8) can lead to a 481 vaccination-promoting environment in which Vfrequencies are reduced over time; thus the V -A + 482 phenotype becomes rare and V + A + predominates (Figure 9 and Figure S6) . 

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

With neutral confidence transmission (C1 = C2 = 0.5), we also observe an expansion of the 536 polymorphic space when we modulate cultural selection (σmax) alongside the influence and 537 transmission parameters (Figure 11) . Interestingly, in the cultural environment defined by this 538 parameter space, we observe a pattern that deviates from the expected association between 539 high vaccine confidence and high vaccination coverage: as the influence of vaccine attitudes (cn) 540 and vaccination probabilities (Bm,n) increase (Figure 11, horizontal axes) , the population's 541 equilibrium vaccination coverage increases while its vaccine confidence decreases. This pattern 542 persisted across all tested levels of maximum cultural selection (σmax) (Figure 11, vertical axes) . In 543 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101/2022.05.26.22275604 doi: medRxiv preprint other words, we observe higher levels of confidence at low influence and transmission 544 probabilities than we do at higher probabilities (Figure 11D, F) . Table 1 .

We explored the interaction between the influence parameters, bm and cn, at various 559 maximum cultural selection coefficients (σmax) (Figure 12) . Vaccination coverage and vaccine 560 confidence equilibrate at mid-range frequencies (between 0.3 and 0.8) across the range of bm and 561 cn, indicating that these trait frequencies are not particularly sensitive to either parameter. 562

Cultural selection favoring vaccination increases the equilibrium level of vaccination coverage and 563 vaccine confidence (Figure 12 and Figure S7) . The most notable deviation between equilibrium 564 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. preferences ( 1 ≈ 2; diagonal lines in Figure 13A -B); above this boundary (when 1 > 2) 593 vaccination coverage and confidence are much higher than below this boundary (when 1 < 2) . 594

When cultural selection favors being vaccinated (σmax > 0, Figure 13C Table 1) to c0 = 637 0.5, c1 = c2 = 0.9, c3 = 0.99, then examined the effect of cultural selection coefficient and 638 confidence transmission probability (Figure 14A-B) ; we then compare the results of this strict 639 mandate both to our baseline parameters and to a more lenient mandate represented by c0 = 640 0.3, c1 = c2 = 0.7, c3 = 0.99 ( Figure S10) . 641

Our previous simulations consistently showed a positive correlation between population-642 level frequencies of vaccination coverage and vaccine confidence (Figures 8-10, 13) . However, 643 modeling a cultural environment under strict mandates reveals a decoupling of vaccination 644 coverage and vaccine confidence at low confidence transmission probabilities (Figure 14A-B) . 645

Even when vaccine confidence is very low (specifically at mixed-trait couple confidence 646 transmission probabilities below 0.5; red region in Figure 14B ), vaccination coverage is much 647 higher than without the mandate (compare Figure 14A -B to Figure 10A -B, see also Figure S10 ). 648

This suggests that an external pressure to vaccinate is able to overcome the opposing cultural 649 pressure imposed by hesitancy in the population. If we lower the barrier to acquiring an 650 exemption (using more lenient mandate parameters), vaccination coverage and vaccine 651 confidence dynamics begin to mirror one another, as they do in simulations without vaccine 652 mandates ( Figure S10) . 653 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101/2022.05.26.22275604 doi: medRxiv preprint We next modeled the dynamics of vaccination behaviors and attitudes when a population 656 has limited access to vaccines. To represent this scenario, we reduced the influence of parental 657 vaccine attitudes on vaccination behaviors for couples with at least one confident individual (i.e. 658 reducing c1, c2, c3 from default values). In this simple representation of a vaccine-scarce 659 environment, we assume that parents' confidence in vaccines would have reduced influence on 660 their ability to vaccinate their offspring, that is, their vaccine confidence does not ensure their 661 ability to overcome vaccine inaccessibility. Hesitant couples are least likely to vaccinate their 662 offspring regardless of vaccine availability, but couples who would likely vaccinate their offspring 663

given the chance would have difficulty doing so due to the lack of resources. This scenario can be 664 contrasted with the strict vaccination mandate scenario described above, in which a couple 665 would often vaccinate their children despite their hesitancy. Attitude influence parameters were 666 set to c0 = 0.01 ; c1 = c2 = 0.1 , and c3 = 0.5, and as before, we modulated the maximum cultural 667 selection coefficient of vaccination (σmax) and confidence transmission (C1 and C2) (Figure 14C-D) . 668

Vaccination coverage was noticeably reduced overall, while vaccine confidence increased slightly 669 across the parameter space. Juxtaposed with the strict mandate ( Figure 14A-B) , our vaccine 670 scarcity models produce an opposite deviation of vaccination coverage from vaccine confidence 671 levels: when vaccines are mandated, we observe increased vaccination coverage in low-672 confidence environments, and when vaccines are inaccessible, we observe lower than expected 673 vaccination coverage (<50%) in a predominantly vaccine-confident environment (>90%) (Figure  674   14) . 675 676 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. is not being robustly transmitted (or maintained in adulthood). Further, we show that a culture in 719 which vaccine-hesitant individuals preferentially assort with one another more so than vaccine-720 confident individuals ( 2 > 1) can allow vaccine hesitancy to more easily gain a foothold, 721 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. In a vaccine-scarce environment, an individual's attitude regarding vaccines has less influence on 752 vaccination behavior due to the barrier imposed by resource availability. As a result, a population 753 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. and to a lesser degree, the themes of vaccination. However, they do not report parent 782 vaccination states or whether the child was actually vaccinated (on schedule). With data 783 presenting parent vaccination states alongside their vaccine attitudes and vaccination decisions, 784 we would be able to more accurately inform phenotype frequencies, possibly extending the 785 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity.

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101/2022.05.26.22275604 doi: medRxiv preprint model to incorporate various types of hesitancy. We note, however, that our results did not 786 depend on the initial proportions of vaccination status or vaccine hesitancy, so these data would 787 primarily be for comparison to our equilibrium outcomes. We were also constrained by limited 788 data to inform the cultural transmission and transition probabilities. In our model, baseline 789 confidence transmission and influence probabilities are structured according to a simple pattern 790 of inheritance, such that each parent is equally likely to influence an offspring's phenotype. 791

However, cultural traits and vaccination attitudes may not strictly follow this pattern: one parent 792 might have more influence, or one variant of a trait might be more likely to be transmitted. In 793 addition, transmission probabilities are constant in our model, remaining unaffected by changing 794 cultural conditions throughout each simulation, but in reality, these probabilities may fluctuate in 795 response to a variety of factors including vaccine type or family structure. Future developments 796 of the model could include modulating the probability of transmission of vaccine confidence 797 dependent on aspects of the cultural environment, such as the attitude frequencies in the 798 population. Our cultural selection coefficient and attitude transition probabilities did vary with 799 the frequency of vaccination coverage, but the exact relationships could not be informed by 800 existing data. As a result, we generated frequency-dependent functions to fit a set of 801 assumptions: for example, that the transition to vaccine hesitancy might be most common when 802 the vaccination coverage is low and the benefits of vaccination are more apparent. 803

Though vaccination coverage and vaccine confidence stabilized (reached equilibria) in our 804 simulations, in reality vaccination rates fluctuate over time in response to changing population 805 dynamics, perhaps never arriving at a stable equilibrium. For example, the increasingly rapid 806 spread of information [23] may cause attitudes and behaviors to change frequently over short 807 periods of time. In our model, most of the phenotype frequency fluctuations occur in the first few 808 iterations before quickly adjusting to an equilibrium. Unlike many models of population 809 dynamics, this model has a discrete-time format and does not consider a birth-death cycle or 810 asynchrony in population turnover. Thus, the timescale of our model might not translate directly 811 to years or generations, and we avoid interpreting the number of iterations in literal terms. It is 812 possible that if more realistic birth and death processes were incorporated, the cultural dynamics 813 would occur at different timescales and would continue to fluctuate instead of approaching a 814 stable equilibrium. In addition, parents' vaccination decisions are also influenced by the 815 grandparents of the children to be vaccinated [72] . A restructuring of the timescale or the 816 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101/2022.05.26.22275604 doi: medRxiv preprint incorporation of population asynchrony in our model could allow for consideration of these 817 influences. 818

Finally, we constructed the offspring vaccination probability to be informed primarily by 819 parents' vaccine attitudes and secondarily by their own vaccination status. Though it is 820 understood that there is an interaction between parents' beliefs and their own experiences with 821 vaccines regarding their decision to vaccinate their children, accurately modeling the relative 822 contribution of these two factors could benefit from empirical studies on parental willingness to 823 vaccinate based on their beliefs and vaccination status. With our current formula (Bm,n, Table S2 ), 824 vaccine-confident parents who did not themselves receive childhood vaccines have a reduced 825 likelihood of vaccinating their offspring than vaccinated parents. In reality, parental vaccine 826 attitudes might even further outweigh their own vaccination status in their decision making than 827 we model here. 828

Our findings, which are based on cultural evolutionary modeling in a public-health 829 context, suggest several avenues for policy and outreach recommendations. First, we note that 830 vaccine mandates with limited exemptions will increase the vaccination coverage but may mask 831 the spread of vaccine hesitancy. In addition to vaccine policy, our results suggest that a broad 832 effort to encourage and inform the public about vaccine safety and efficacy will foster higher 833 vaccine coverage. In this vein, we note that individuals who are skeptical about vaccines might 834 invest more time in seeking out information about them [73-75], thus we recommend that 835 accurate information about vaccines should be readily accessible through a variety of means, be 836 easily understood, and be supported by personal anecdotes. If misinformation is easier to 837 encounter and digest than accurate information, it could have an outsized impact on individuals 838 who are "on the fence" [61]. Relatedly, we note that research suggests transparency and trust go 839 hand in hand: if individuals perceive the healthcare system is concealing information to make 840 vaccines appear more safe, trust in that system will decrease and people will be more susceptible 841 to vaccine hesitancy [73] [74] [75] [76] . Finally, dialogue between people with different beliefs and 842 attitudes can help to break the "echo chambers" of homophily, encouraging individuals to 843 communicate and empathize with one another. Therefore, to address vaccine hesitancy, our 844 results underscore the importance of considering the cultural beliefs that underpin health 845 behaviors. 846 847 848 . CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review) 

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. 

. CC-BY-NC-ND 4.0 International license It is made available under a is the author/funder, who has granted medRxiv a license to display the preprint in perpetuity. (which was not certified by peer review)

The copyright holder for this preprint this version posted May 27, 2022. ; https://doi.org/10.1101/2022.05.26.22275604 doi: medRxiv preprint

References: Public Health. The Future of Public Health

A successful eradication campaign. Global eradication of smallpox

THE 854 USA: VIRTUAL ELIMINATION OF DISEASE CAUSED BY WILD VIRUS. The Lancet

Eradication of Poliomyelitis in the United States. I. Live Virus Vaccine-Associated and Wild 857

Clinical Infectious Diseases

Enabling implementation of the Global 859

Vaccine Action Plan: developing investment cases to achieve targets for measles and rubella 860 prevention

Pertussis resurgence and vaccine uptake: implications for reducing vaccine 862 hesitancy. Pediatrics

Is There a Resurgence of Vaccine Preventable Diseases in the U.S.? Journal of Pediatric 864 Nursing

Factors associated with suboptimal compliance to vaccinations in children 866 in developed countries: a systematic review

Vaccine hesitancy: an overview

Patterns of vaccination acceptance

Group on Vaccine Hesitancy. Vaccine hesitancy: Definition, scope and 872 determinants

The Ten Threats to Global Health in 2018 and 2019. A welcome and 874 informative communication of WHO to everybody. Zdrowie Publiczne i Zarządzanie

New media, old messages: themes in the history of vaccine hesitancy and refusal. Virtual 877 Mentor

Durbach N. Bodily Matters: The Anti-Vaccination Movement in England

The Leicester anti-vaccination movement. The Lancet. 1992

Anti-vaccinationists past and present

State laws on compulsory immunization in the United States

Individual and Community Risks of Measles and Pertussis Associated With Personal 892 Exemptions to Immunization

Health Consequences of Religious 894 and Philosophical Exemptions From Immunization Laws

Information dissemination via electronic word-of-mouth: 897 Good news travels fast, bad news travels faster! Computers in Human Behavior

Perception of risk of vaccine adverse events: a historical perspective

Lancet retracts 12-year-old article linking autism to MMR vaccines

The MMR vaccine and autism: Sensation, refutation, retraction, and fraud. Indian 904 J Psychiatry

Attitudes to vaccination: A critical review

The meaning of assortative mating

Complexity in models of cultural niche construction with selection and 909 homophily

Models of cultural niche construction with selection and 911 assortative mating

Vaccination usage among an old-order Amish community in Illinois. Pediatr 919

A Measles Outbreak in an 921

Underimmunized Amish Community in Ohio

Addressing MMR Vaccine Resistance in Minnesota's Somali Community

Childhood vaccine beliefs reported by Somali and non-Somali parents

An introduction to niche construction theory

Niche Construction: The Neglected Process in 929 Evolution (MPB-37)

The niche construction of cultural complexity: interactions between 933 innovations, population size and the environment

Genes, culture, and 936 agriculture: An example of human niche construction

Cooperation, conflict, and niche construction in the genus homo. War, peace, and human 938 nature

Towards a characterization of behavior-disease models

Coupling infectious diseases, human preventive behavior, and networks--a conceptual 942 framework for epidemic modeling

Stability analysis of SIR model with vaccination

Parents' health beliefs and HPV vaccination of 953 their adolescent daughters

The Health Belief Model: a decade later

The Health Belief Model and Preventive Health Behavior

The Health Belief Model as an 958 explanatory framework in communication research: exploring parallel, serial, and moderated 959 mediation

Nonmedical exemptions from school immunization 961 requirements: a systematic review

Association Between Vaccine Refusal and Vaccine-963

Preventable Diseases in the United States: A Review of Measles and Pertussis

Perceived Barriers to Childhood Immunization Among Rural 966

Journal of Community Health Nursing

Cultural transmission and evolution: a quantitative approach

Impact of anti-vaccine 971 movements on pertussis control: the untold story

Cost-effectiveness and economic benefits of 973 vaccines in low-and middle-income countries: a systematic review

Vaccine beliefs of parents who oppose compulsory vaccination

Target the fence-sitters

Age 24 Months Among Children Born in 2015 and 2016 -National Immunization Survey-Child

United States

Parents who refuse or delay HPV vaccine: 988 Differences in vaccination behavior, beliefs, and clinical communication preferences

Papillomavirus Vaccination Intentions and Reasons for Hesitancy

vaccine based on the health belief model: A population-based survey in Hong Kong

Vaccine production, distribution, access, and uptake

A global database of COVID-19 999 vaccinations

Global COVID-19 vaccine inequity

COVID-19 vaccine 1002 acceptance and hesitancy in low-and middle-income countries

Do grandparents influence parents' decision to vaccinate their 1004 children? A systematic review

The rise (and fall?) of parental vaccine hesitancy

Parents' Behaviors and Experiences Associated with Four 1008

Vaccination Behavior Groups for Childhood Vaccine Hesitancy

Qualitative Analysis of Mothers' 1011

Decision-Making About Vaccines for Infants: The Importance of Trust

Transparency as a Trust-Building Practice in Physician 1014 Relationships With Patients