Daniel R. Feikin, Dennis C. Lezotte, Richard F. Hamman, Daniel A. Salmon, Robert T. Chen, Richard E. Hoffman. Individual and Community Risks of Measles and Pertussis Associated With Personal Exemptions to Immunization. JAMA. 2000;284(24):3145–3150. doi:10.1001/jama.284.24.3145
Author Affiliations: Preventive Medicine Residency, Epidemiology Program Office (Dr Feikin), and the National Immunization Program (Dr Chen), Centers for Disease Control and Prevention, Atlanta, Ga; Department of Preventive Medicine and Biometrics, University of Colorado Health Science Center (Drs Lezotte and Hamman) and the Colorado Department of Public Health and Environment (Dr Hoffman), Denver; and Johns Hopkins University School of Hygiene and Public Health, Baltimore, Md (Mr Salmon).
Context The risk of vaccine-preventable diseases among children who have philosophical
and religious exemptions from immunization has been understudied.
Objectives To evaluate whether personal exemption from immunization is associated
with risk of measles and pertussis at individual and community levels.
Design, Setting, and Participants Population-based, retrospective cohort study using data collected on
standardized forms regarding all reported measles and pertussis cases among
children aged 3 to 18 years in Colorado during 1987-1998.
Main Outcome Measures Relative risk of measles and pertussis among exemptors and vaccinated
children; association between incidence rates among vaccinated children and
frequency of exemptors in Colorado counties; association between school outbreaks
and frequency of exemptors in schools; and risk associated with exposure to
an exemptor in measles outbreaks.
Results Exemptors were 22.2 times (95% confidence interval [CI], 15.9-31.1)
more likely to acquire measles and 5.9 times (95% CI, 4.2-8.2) more likely
to acquire pertussis than vaccinated children. After adjusting for confounders,
the frequency of exemptors in a county was associated with the incidence rate
of measles (relative risk [RR], 1.6; 95% CI, 1.0-2.4) and pertussis (RR, 1.9;
95% CI, 1.7-2.1) in vaccinated children. Schools with pertussis outbreaks
had more exemptors (mean, 4.3% of students) than schools without outbreaks
(1.5% of students; P = .001). At least 11% of vaccinated
children in measles outbreaks acquired infection through contact with an exemptor.
Conclusions The risk of measles and pertussis is elevated in personal exemptors.
Public health personnel should recognize the potential effect of exemptors
in outbreaks in their communities, and parents should be made aware of the
risks involved in not vaccinating their children.
Control of vaccine-preventable diseases in the United States has been
achieved in part by implementation and enforcement of mandatory vaccination
for school entrance in every state.1,2
Despite challenges to immunization laws, the courts have upheld such laws,
citing the substantial public health benefit afforded to society by vaccination.3 Yet, to address concerns regarding mandatory vaccination,
many states allow personal exemptions from vaccination. Religious exemptions
are permitted in 48 states and philosophical exemptions in 15 states.4 Exemptions based on medical grounds are allowed in
Multiple reasons exist for which parents choose not to allow their children
to be vaccinated.5 One reason that has become
prevalent is fear of adverse reactions to vaccination. As the frequency of
vaccine-preventable diseases has decreased, fewer persons have ever witnessed
a child who is severely ill with a vaccine-preventable diseases. As a result,
parents today might perceive a greater risk in having their children vaccinated
than in not having them vaccinated.6 Despite
this belief, some data demonstrate otherwise. A recent study showed that risk
of measles infection during 1985-1992 in the United States was, on average,
35 times greater in children with personal exemptions compared with vaccinated
children.7 Another review demonstrated that
countries that have more active antivaccine movements have higher rates of
pertussis than countries where the majority of children are vaccinated.8
Colorado statute allows both religious and philosophical exemptions,
and parents can take these exemptions simply by signing a statement that declares
they are "an adherent to a religious or personal belief opposed to immunization."
In 1994, the percentage of school-aged children who were unvaccinated as a
result of personal exemptions in Colorado was 1.4%, more than twice the national
average of 0.6% for the same year.4
The goal of this study was to assess whether individuals and communities
experience adverse consequences associated with personal exemptions. In particular,
we evaluated risk of measles and pertussis infection associated with personal
exemptions among school-aged children in Colorado in 1987-1998.
Children aged 3 to 18 years whose parents chose not to allow them to
be vaccinated for religious or philosophical reasons are referred to as exemptors. We did not consider medical exemptions to be
personal exemptions. Children were considered to be vaccinated if they had
received at least 1 measles vaccination or at least 4 pertussis vaccinations.
Cases of disease in unvaccinated or incompletely vaccinated children not declared
as exemptors were excluded.
The proportions of children who were exemptors, fully vaccinated, and
incompletely vaccinated were available through the results of an annual immunization
summary report sent to all Colorado schools, including preschools, that inquired
about the number of students who were vaccinated or exemptors. We assessed
trends in annual proportions of exemptors by using the χ2 test
for trend. A random probability sample of public and private schools was audited
by epidemiologists from the Colorado Department of Public Health and Environment
(CDPHE) to assess accuracy of the school reports.
Children aged 3 to 18 years in Colorado represented the cohort from
which we retrospectively calculated incidence rates for measles and pertussis.
Physicians and clinical laboratories reported all cases to the CDPHE as mandated
by state statute. For each case, a case report form was completed by public
health nurses from the local health department and reviewed by epidemiologists
at the CDPHE. Because of differences in the diagnostic sensitivity of the
2 diseases, the Centers for Disease Control and Prevention (CDC) required
states to report confirmed and probable cases of pertussis but to report only
confirmed cases of measles; therefore, these same categories of cases were
included in this analysis. Cases were defined according to accepted criteria
at the time of diagnosis.9,10
Case definitions did not include information on the case-patient's vaccination
status. For both diseases in 1987-1998, case report forms documented patients'
age, county of residence, and vaccination status. Information regarding the
reason for lack of vaccination, including exemptions, was collected for measles
in 1987-1998 and for pertussis in 1996-1998; therefore, the analysis of incidence
rates among individuals included different periods for the 2 diseases.
We divided children into 4 age groups that correspond to school levels:
preschool and kindergarten (3-5 years), elementary school (6-10 years), middle
school (11-14 years), and high school (15-18 years). We estimated the annual
state population of vaccinated and exemptor children by applying the annual
proportion of vaccinated and exemptor children from the school immunization
summary report to the entire annual state population of the 4 age groups.11
We calculated incidence rates by dividing the total number of cases
among exemptors and vaccinated children by the cumulative number of children
during the indicated period within each group. We calculated relative risks
(RRs) by dividing the average annual incidence rate among exemptors by the
rate in vaccinated children.
County Level. We analyzed whether the frequency of exemptors in a county was predictive
of the average annual incidence rate for measles and pertussis among vaccinated
children in 1987-1998. We performed Poisson regression analysis using average
annual county incidence rates among vaccinated children as the outcome variable
and percentage of exemptor children in the county as the primary independent
variable. The RR was interpreted as the change in incidence rate for each
1% change in exemptors. For example, an RR of 2.0 would indicate a doubling
of the incidence rate for each 1% increase in exemptors.
We adjusted for several potential confounders, including income, educational
status, immigrant population, and population density of individual counties,
available at the county level from the 1990 US Census. In addition, because
of the changing incidence of disease and rates of exemptors during the study
period (measles decreased and pertussis increased), we adjusted for the year
group of diagnosis; the 1987-1998 period was divided into four 3-year intervals.
The average annual county incidence rates during these 4 intervals were used
as 4 separate observations per county. Information regarding exemptions at
the county level was available only for 1992, 1993, 1996, and 1998. We used
data from these 4 years in a linear regression model of the logarithmic transformation
of the percentage of exemptors per county to estimate the percentage of exemptors
in each county for the midyear of the 4 intervals (ie, 1988, 1991, 1994, and
1997). One sparsely populated county was excluded because of lack of information
on the frequency of exemptors. Significant confounders were defined as variables
that caused more than a 10% change in the RR.
School Level. Case report forms for measles (1987-1998) and pertussis (1996-1998)
collected information on the transmission setting of the disease. We considered
a school-based outbreak as one in which at least 1 vaccinated child acquired
infection in a school.
We analyzed whether the probability of having a school-based outbreak
could be predicted by the percentage of exemptors in that school during the
1993 school year for measles or the 1996 school year for pertussis. To prevent
biases of testing and reporting caused by county-level differences, only schools
in counties in which a school-based outbreak occurred were included. Because
the percentage of exemptors in schools was a nonparametrically distributed
continuous variable, univariate analysis was performed by using the Wilcoxon
rank sum test. Multivariate analysis was performed by using logistic regression.
We adjusted for several potential confounders: school size, school level,
and county in which a school was located. Significant confounders were defined
as variables that caused more than a 10% change in the odds ratio.
Initiation Among Exemptors. Only outbreaks that involved 5 or more individuals (of any age) and
that had 2 or more generations were included. Whether the case-patient was
part of an outbreak was indicated on the case report forms. We evaluated whether
outbreaks began among exemptors and spread to vaccinated persons by comparing
the percentage of exemptors in the index and first generation of the outbreak
with that of later generations of the outbreak. The reason we combined the
index and first generations was that often the index generation involved only
1 person. Later generations were combined because many outbreaks had only
1 generation after the first. The χ2 test with continuity correction
was used to compare proportions.
Risk of Exemptors in Outbreaks of Measles. We quantified the contribution of exemptors in measles outbreaks with
5 or more individuals (of any age). The source of exposure was often recorded
on the case report forms. This analysis quantified the percentage of children
in an outbreak who contracted measles through contact with an exemptor. This
analysis was not possible for pertussis outbreaks because the source of exposure
of most pertussis cases was unknown.
During 1987-1998, the percentage of philosophical exemptions among school-aged
children in Colorado increased from 1.02% to 1.87% (Table 1; P<.001). Religious exemptions
decreased slightly, from 0.23% to 0.19% (P<.001).
Overall, philosophical exemptions accounted for 87% of all exemptions.
On average, 11% of schools in the state were audited annually. The annual
percentages of religious and philosophical exemptions were similar in audited
schools and in the annual immunization summary, with the greatest absolute
difference in a single year between the 2 being 0.14% for religious exemptions
and 0.47% for philosophical exemptions.
During 1987-1998, 505 confirmed cases of measles were reported in Colorado,
with 202 (40%) occurring in children aged 3 to 18 years. Among exemptors,
the measles incidence rate was highest among children aged 3 to 10 years,
while among vaccinated children, the rate increased with increasing age (Table 2). On average, exemptors were 22
times more likely to acquire measles than were vaccinated individuals. The
excess risk of measles among exemptors was greatest among children aged 3
to 10 years (RR, 62.0; 95% confidence interval [CI], 39.0-98.6).
During 1996-1998, 1140 confirmed and probable cases of pertussis were
reported in Colorado, with 541 (47%) occurring in children aged 3 to 18 years.
Among exemptors, the pertussis incidence rate was highest among children aged
3 to 10 years (Table 2). On average,
exemptors were 5.9 times more likely to acquire pertussis than were vaccinated
children. Excess risk of pertussis among exemptors was greatest among children
aged 3 to 10 years (RR, 15.9; 95% CI, 10.8-23.4).
Our calculated risk of pertussis among exemptors was most likely an
underestimate of the true amount. Among children with 0 pertussis vaccinations,
43% had an unknown reason for lack of vaccination. Many of these children
were likely exemptors. If all these children are considered to be exemptors,
the risk of acquiring pertussis among exemptors compared with vaccinated children
increased for children aged 3 to 18 years (RR, 10.6; 95% CI, 8.1-13.8) and
for children aged 3 to 10 years (RR, 26.1; 95% CI, 18.9-36.0).
County Level. Measles cases were reported in 19 (30%) of 63 Colorado counties. Of
these cases, 78% occurred in the 10 counties with a metropolitan area of more
than 100 000 persons (approximately 80% of the state population). Adjusting
for the year of diagnosis, the overall annual incidence rate of measles among
vaccinated children aged 3 to 18 years in a county was significantly associated
with the frequency of exemptors in that county (RR, 1.6; 95% CI, 1.0-2.4).
This association was significant for children aged 3 to 10 years (RR, 1.7;
95% CI, 1.2-2.5) but not for children aged 11 to 18 years (RR, 1.4; 95% CI,
0.9-2.1). None of the variables other than year of diagnosis were significant
Pertussis cases were reported in 21 Colorado counties (33%). Of these
cases, 93% occurred in the 10 counties with a metropolitan area of more than
100 000 persons. The annual incidence rate of pertussis among vaccinated
children aged 3 to 18 years in a county was also significantly associated
with the frequency of exemptors in that county after adjustment for year of
diagnosis, the only significant confounder (RR, 1.9; 95% CI, 1.7-2.1). This
association was significant for children aged 3 to 10 years (RR, 1.8; 95%
CI, 1.6-2.0) and 11 to 18 years (RR, 1.8; 95% CI, 1.6-1.9).
School Level. Twenty-two percent of children with measles acquired infection at school;
33% had an unknown transmission setting. Eight school-based outbreaks occurred
in 4 counties (5 elementary, 1 middle, and 2 high schools). A total of 672
schools were included in the analysis. Schools with measles outbreaks had
more exemptors (median, 1.5%; mean, 2.8%) than schools without measles outbreaks
(median, 1.0%; mean, 1.8%), although this difference was not statistically
significant (P = .26).
Twenty-one percent of children with pertussis acquired infection at
school; 46% had an unknown transmission setting. Seventeen school-based outbreaks
occurred in 5 counties (10 elementary, 2 middle, and 5 high schools). A total
of 954 schools were included in the analysis. Schools with pertussis outbreaks
had more exemptors (median, 1.9%; mean, 4.3%) than schools without pertussis
outbreaks (median, 0.7%; mean, 1.5%; P = .001). In
logistic regression analysis, for each 1% increase in exemptors in a school,
the risk of having a pertussis outbreak increased by 12% (odds ratio, 1.12;
95% CI, 1.05-1.20). School level, size, and county were not significant confounders
of this association.
Initiation Among Exemptors. In the 14 measles outbreaks included in this analysis, 24 (14.5%) of
166 individuals were exemptors in the index and first generations compared
with 18 (7.1%) of 254 in later generations (P = .03).
In the 12 pertussis outbreaks included in this analysis, 8 (10.3%) of 77 cases
occurred among exemptors in the index and first generations compared with
9 (8.4%) of 107 in later generations (P = .84).
Risk of Exemptors in Outbreaks. In the 18 measles outbreaks included in this analysis, of 179 cases
among children, 45 (25%) were exemptors. Forty-two percent of exemptors contracted
measles through contact with another exemptor. Among vaccinated children,
11% contracted measles through contact with an exemptor; this amount is probably
an underestimate of the true percentage because 67% of vaccinated children
had an unknown exposure source, some of whom probably were exposed to exemptors.
The epidemic curve of a 1994 outbreak demonstrates how a cluster of exemptors
early in an outbreak can expose other persons, precipitating cases in later
generations (Figure 1).12
Our study showed that during a recent decade in Colorado, exemptors
were, on average, 22 times more likely to acquire measles and 6 times more
likely to acquire pertussis than vaccinated children. In children of day care
and primary school age, in whom contact rates and susceptibility are higher,
these risks were approximately 62-fold and 16-fold greater among exemptors
for measles and pertussis, respectively. Our findings for measles were consistent
with a recent report that estimated that exemptors were at a 35-fold increased
risk of acquiring measles compared with vaccinated children in the United
States overall.7 That study, like ours, documented
that the risk is most elevated for younger children. This could be a result
of the natural epidemiology of measles, in which the majority of unvaccinated
children are infected before age 10 years.13
In addition, the rate of measles among vaccinated children increased with
age in our study, probably because of primary vaccine failures among older
children who were vaccinated before age 15 months in the 1970s.13,14
This is the first study, to our knowledge, to document at the population
level that an excess risk of acquiring pertussis exists among exemptors. Similar
to measles, risk of pertussis among exemptors was highest in younger children.
This could be because the rate of pertussis is highest among young children15 and immunity due to vaccination wanes approximately
10 years after the last vaccination.15,16
Our study also provides specific evidence suggesting that personal exemptors
put vaccinated children at risk of acquiring measles and pertussis. Multiple
outbreaks have occurred in isolated religious communities where most children
claimed religious exemptions to vaccination.17- 19
However, our study suggests that when mixing of exemptor and vaccinated populations
occurs in a county, in a school, or during an outbreak, exemptors can transmit
disease to vaccinated individuals.
Our study had several limitations. For the analysis of individual risk,
the vaccination status was unknown for 3% of measles and 9% of pertussis cases.
Although this decreases the precision of the study, it would not bias the
results unless children with unknown vaccination status were significantly
more or less likely to be exemptors. Second, on the school immunization summary
report, children were labeled as exemptors if they took an exemption from
only 1 vaccine but still received others. Therefore, the population of exemptors
at risk might have been overestimated for either the measles or the pertussis
analysis of individual risk. This bias, however, would tend to underestimate
excess risk of disease among exemptors. Third, the county- and school-level
analyses were ecological. Nonetheless, we were able to control for several
important potential county-level confounders in these analyses. Fourth, for
the analysis of school outbreaks, in many cases the transmission setting of
disease was unknown. Most likely, some of these children acquired their disease
at school; therefore, some school-based outbreaks were missed. This would
decrease the power of the study. However, it should not introduce a bias since
the transmission setting of exemptors should have been no more or less likely
to have been identified than that of vaccinated children. Last, the possibility
exists that children's vaccination status might have influenced their probability
of being diagnosed as a case. There may have been a tendency for clinicians
to perform more laboratory tests on unvaccinated children because of a higher
suspicion of disease. However, our data did not show significantly more laboratory-confirmed
cases among exemptors than among vaccinated children (data not shown). Moreover,
this tendency on the clinicians' part may have been offset by the possibility
that parents of exemptors might have been less likely than those of vaccinated
children to seek traditional medical care that involved laboratory testing.
Public health personnel investigated all reported cases prior to knowing the
case-patient's vaccination status, which was ascertained only after completion
of the case report form. When defining a case as confirmed or probable, personnel
used the standard definitions established by the CDC, which did not include
an individual's vaccination status.
Although their perspectives might differ, both parents and public health
policymakers must balance competing perceptions of risks and benefits in making
immunization decisions.20 The perception of
the prevalence and severity of the disease is often weighed against the perceived
likelihood of an adverse reaction to vaccination. The success of immunization
against vaccine-preventable diseases has shifted the focus of the public's
perception of risk from the diseases to the vaccines themselves. Concerns
regarding the safety of the vaccines against measles and pertussis have been
raised by some reports from the medical community,21,22
although the data often have been misinterpreted and sensationalized when
communicated to the public.8,23
Despite these reports, studies have proven that risk of serious adverse events
to the measles12,24 and the whole-cell15,23 and acellular pertussis15,25
vaccines remains rare.
The decision to forgo vaccination must balance individual rights with
social responsibility. If all vaccine-preventable diseases were confined to
the individual (eg, tetanus), the consequences of forgoing vaccination would
fall only on the child whose parents made the decision. Most vaccine-preventable
diseases, however, are spread from person to person. Therefore, the health
of any individual in the community is intricately dependent on the health
of the rest of that community. A single unvaccinated child in a community
of vaccinated children holds a strategically opportunistic high ground, protected
from risk of disease by herd immunity while avoiding risk of exceedingly rare
adverse events associated with vaccination. Yet, when too many parents want
their child to be that child, the entire community is affected. Such a situation
occurred in both Japan and the United Kingdom in the 1970s, when pertussis
vaccine coverage declined from more than 80% to less than 30% after reports
of severe adverse reactions to the vaccine, resulting in major nationwide
Until vaccines become available that are 100% effective or a disease is eradicated,
an increase in exemptors has the potential to precipitate communitywide outbreaks
of vaccine-preventable diseases.
Parents making decisions regarding vaccination should understand not
only the risks of vaccination but also the risks of not being vaccinated.
Enhanced surveillance for vaccine-preventable diseases, particularly pertussis,
which is underdiagnosed,26 and rapid control
measures (eg, prophylactic antibiotics for pertussis) might be warranted in
areas that have larger populations of personal exemptors. Policymakers might
consider requiring some evidence of parental strength of conviction when claiming
personal exemptions for their children. In Colorado, the ease of claiming
personal exemptions could encourage exemptions of convenience by parents who
have not really considered the issue or who want to avoid the financial and
time commitment of having their children vaccinated. Health care practitioners
and public health officials could do a better job of exchanging useful information
with parents, enabling them to understand the benefit of vaccines and to recognize
the risks for the diseases they prevent, in the context of what is known regarding
risks associated with vaccination. Further studies and interventions should
be pursued that will ensure that individual and societal decisions regarding
childhood immunizations are made on the basis of information and understanding.