Religious Vaccine Exemptions in Kindergartners: 2011–2018

We performed a cross-sectional, retrospective, state-level analysis of publicly available aggregated data from the Centers for Disease Control and Prevention yearly vaccine reports for kindergartners (2011–2012 through 2017–2018 school years).^12–18  We chose this time frame to extend and provide meaningful comparisons to a previous study of exemption data for kindergartners from 2011–2012 through 2015–2016.⁶  For each report, states provide estimates of the proportion of children with medical and nonmedical (ie, religious and/or personal beliefs) exemptions as well as the proportion of kindergartners who are up to date on several vaccines of interest (eg, measles, mumps, and rubella; diphtheria, tetanus, and acellular pertussis; and varicella). States use different approaches to calculate these proportions, including censuses, 1- or 2-stage cluster surveys, simple random sampling, or mixed designs. However, most states use censuses and sample >80% of their kindergartners, so sampling variability in their estimates should be minimal. Therefore, we assumed these measurements were taken without error in our analyses.

We calculated the incidence proportion of kindergartners with religious vaccine exemptions for each state-year from these reported proportions. For states with religious exemptions only, the reported nonmedical exemption incidence proportion equaled the incidence proportion of religious exemptions. For states with religious and personal belief exemptions, we calculated the religious exemption proportion by dividing the state’s number of children with religious exemptions by the overall number of children with nonmedical exemptions and multiplied by the nonmedical exemption proportion.

Because the exemption process is heterogeneous, we accounted for the strength of state exemption policies (ie, whether it was “easy,” “medium,” or “difficult” for parents to claim an exemption) on the basis of standards previously reported by Omer et al.⁶  We chose this adjustment because previous studies suggest that nonmedical exemption rates decrease when exemption policies are more difficult to obtain.^5,19  Because many states revised their exemption laws during the study period, we recoded exemption law strengths for each state every year with published exemption law summaries.^1,6  We did not include Mississippi and West Virginia because they did not allow nonmedical exemptions. We also excluded Arizona, which allowed personal belief exemptions only in kindergarten. California eliminated nonmedical exemptions in 2015; we excluded state-years after this change. As noted, Vermont eliminated philosophical exemptions in 2016. In total, we considered 357 state-years for inclusion, excluding 62 state-years for a lack of separate reporting of nonmedical exemptions, a lack of religious exemptions, or a lack of data. There were 295 state-years in the final analysis.

Our dependent variable of interest was the religious exemption incidence proportion for a single state in a given year. We performed univariable analyses by year, calculating mean religious exemption proportions by state policy availability, school year, and exemption policy strength; we calculated robust SEs to account for correlation within states. Because our outcome variable was a proportion between 0 and 1, standard linear regression techniques were not appropriate. Therefore, we performed a quasi-binomial regression analysis with SEs adjusted for correlation within states.²⁰  We chose a log link function to provide estimates of risk ratios rather than odds ratios. Because we used aggregated (state-level rather than individual-level) data, our risk ratios were ratios of mean proportions, or mean population risks, rather than individual risks. We included exemption policy availability (religious only versus religious and personal beliefs), exemption policy strength (easy versus medium and difficult), and school year (treated as a categorical variable) as covariates. All analyses were conducted with R version 3.4.4 (The R Foundation, Vienna, Austria).

In unadjusted analyses, states with religious and personal belief exemptions had a significantly lower mean proportion of kindergartners with religious exemptions (0.41%; 95% confidence interval [CI] 0.28%–0.53%) compared with states with religious exemptions only (1.63%; 95% CI 1.30%–1.97%). This phenomenon was stable for each school year during the study period (Fig 1). In adjusted analyses, states with religious and personal belief exemptions were one-fourth as likely to have kindergartners with religious exemptions compared with states with religious exemptions only (risk ratio 0.25; 95% CI 0.16–0.38). After Vermont eliminated personal belief exemptions in 2016, the mean proportion of kindergartners with religious exemptions increased from 0.5% (2011–2012 through 2015–2016) to 3.7% (2016–2017 through 2017–2018). Overall, the mean proportion of Vermont kindergartners with any nonmedical exemption decreased from 5.7% to 3.7%. Using 2011–2012 as a reference year, states were more likely to have kindergartners with religious exemptions during the 2017–2018 school year (P = .04). Compared with states with easy exemption policies, states with medium or difficult policies were equally likely to have kindergartners with religious exemptions (Table 1).

In this retrospective analysis of data on religious exemptions for American kindergartners, we found that states with religious and personal belief exemptions were significantly less likely to have kindergartners with religious exemptions than were states with religious exemptions alone. Next, we observed that Vermont’s mean proportion of kindergartners with religious exemptions increased sevenfold after its elimination of personal belief exemptions, suggesting a replacement effect. Finally, relative to the 2011–2012 school year, we found that states were more likely to have kindergartners with religious exemptions in the 2017–2018 school year.

First, this study enhances our understanding of how vaccine exemption policies influence state-reported vaccine exemption data. Two recent longitudinal analyses of nonmedical exemption data for kindergartners found that states offering both religious and personal belief exemptions had higher overall rates of nonmedical exemptions than did states with religious exemptions alone.^6,19  However, authors in the first study did not report religious exemption rates separately by state,⁶  and authors in the second study restricted their analyses of religious exemptions to states with religious exemptions only.¹⁹  Thus, both studies did not examine how religious exemption rates differ according to the availability of personal belief exemptions. We augment these findings by demonstrating that states with both nonmedical exemption policies are significantly less likely to have kindergartners with religious exemptions than were states with religious exemptions only. Put differently, state-level religious exemption rates appear to be a function of personal belief exemption availability, decreasing significantly when states offer a personal belief exemption alternative.

The suggestion of a replacement effect in Vermont after its 2016 legislative change provides further evidence of the impact of policy availability on religious exemption rates. Previous studies document a vaccine exemption replacement effect in other settings after analogous policy changes. In 2016, California eliminated all nonmedical exemptions.¹  Delamater et al²¹  studied the rate of medical exemptions in California before and after the legislative change, finding that medical exemptions decreased in the first year after the change. However, by the second year, medical exemption rates rose, and geographic analyses showed that the areas with rising medical exemption rates coincided with areas that previously had high rates of nonmedical exemptions. The authors did not study individual-level behavior, but they suggested that their results were evidence of a replacement effect, whereby parents obtained medical exemptions for their children in the absence of an alternative. We hypothesize a similar effect with Vermont’s 2016 change. Although our study was also not at the individual level, at least 2 small individual-level studies suggest that parents claim religious exemptions in the absence of a personal belief alternative.^22,23  Furthermore, it is interesting to note that Vermont ranked 48th out of 50 in state religiosity in the year before its policy change.²⁴  Ultimately, additional work at the individual level must confirm this hypothesis. Regardless, the overall proportion of Vermont kindergartners with nonmedical exemptions decreased after its 2016 legislative change. This finding aligns with aforementioned studies that showed lower overall nonmedical exemption rates in states with religious exemptions only, and vaccine advocates will likely interpret this as a public health victory.^6,19  In the last year, 10 other states have enacted or proposed legislation to eliminate nonmedical exemptions, and policy makers in other states could consider Vermont’s experience as an instructive example when considering policy changes to decrease exemption rates.²⁵  However, further studies are needed to determine if sudden apparent increases in religious vaccine exemption rates may stigmatize religious groups or alienate religious leaders who support vaccination.

Third, our data suggest that religious exemptions may be rising again. Whereas previous longitudinal studies of religious exemptions suggested a plateau as of the 2015–2016 school year,⁶  our up-to-date results show an increase for the 2017–2018 school year, most prominently in states with religious exemptions only. Although the increase is not surprising in light of recent outbreaks, it is curious when considering the simultaneous decrease in Americans’ religious affiliations. A 2014 survey found that 22.8% of Americans reported no religious affiliation, up from 16.1% in 2007.²⁴  It may be that insular religious groups are growing more rapidly than secular communities or that they are increasingly falling victim to vaccine misinformation. Yet, the rise of religious exemptions in an increasingly secular society also questions whether the religious exemption category can still serve its intended purpose. Religious vaccine exemptions originated >80 years ago and were formalized in the 1960s to exempt children from specific religious groups whose official teachings prohibited vaccination.^26,27  Today, all major religions support vaccination, and fewer Americans are religious.^2,24  In contrast, Americans are increasingly spiritual, with 27% of Americans self-identifying as such in 2014, up from 19% in 2007.²⁴  Whereas definitions of religion emphasize adherence to official teachings and traditions, definitions of spirituality emphasize one’s individual experience of the divine and individual interpretations of sacred texts or traditions.²⁸  Thus, laws formalized in the 1960s may be increasingly outdated in an increasingly secular and spiritual society that prioritizes individual experience over dogmatic adherence to teachings. Future work is needed that explores the rise in religious exemptions at the individual level, differentiating religious from spiritual concerns.

This study has limitations. First, in 16 state-years, states did not report the number of children with exemptions but reported the number of exemptions for each vaccine, which could count some children more than once. In these instances, reported proportions were calculated by using lower-bound estimates of individual vaccine exemption rates. Second, individual states collected data using a variety of methods. Although this could have resulted in varying precision among reported exemption proportions, the impact of this on our results should be minimal because ∼75% of states performed censuses with at least 80% of children counted. Next, our adjusted analyses included state policy type availability, exemption policy strength, and school year only. We did not adjust for state religiosity because of a lack of data on this for all years except 2014. We did not adjust for race, income, education, and population density (though other longitudinal analyses of kindergartner exemption data have done so^4–6 ) because state-reported proportions represent aggregated data, and the interpretation of effects estimated from models including such covariates could be problematic. Finally, because we used state-level data and our risk ratios are estimates of population-level risks, we cannot make individual-level conclusions about parents’ vaccine-related behaviors regarding religious exemptions.

State-reported kindergarten religious exemption rates differ by policy availability, may be subject to a replacement effect on personal belief exemption elimination, and were increasing as of the 2017–2018 school year. Furthermore, religious exemptions may be an increasingly problematic or outdated exemption category, and researchers and policy makers must work together to determine how best to balance a respect for religious liberty with the need to protect public health. Researchers should study individual-level behaviors to determine why religious exemption rates might change with personal belief exemption availability, document whether parents choose religious exemptions in the absence of a personal belief alternative, and differentiate religious from spiritual concerns. Policy makers must engage key stakeholders to discuss exemption laws and determine the viability of religious exemptions as an independent exemption category. This important work will help school vaccine exemption laws evolve in our rapidly changing society, ensuring they remain a cornerstone of public health for decades to come.

CI

confidence interval