PEDIATRICS Volume 137 , number 4 , April 2016
providers inside and outside the
Kaiser network or verified through
school or provider records for
vaccinations given before enrollment.
Kaiser Permanente also was an active
surveillance reporting site in AV. The
Philadelphia Department of Public
Health immunization registry has
used health department birth records
and vaccine administration reporting
from health care providers to
establish and maintain immunization
records for all children aged ≤6
years in Philadelphia since 1995. The
registry expanded to include children
aged ≤18 years in 2007.
For each varicella case identified
that met study inclusion criteria,
we selected potential controls using
incidence density sampling by
extracting age-matched (±2 years)
records for all children from the pool
of eligible controls aged 1 to 18 years
who did not have a previous varicella
history documented in historic
surveillance data or immunization
registry.
25
A 2-year age range for
controls was chosen, because the
routine 2-dose varicella vaccination
recommendation spans ages 4 to 6
years.
13
Moreover, because routine
1-dose coverage reached higher
levels (>80%), protection from
1-dose varicella vaccination appears
to remain consistent during the
first few years after vaccination.
13
Between 5 and 60 potential control
subjects were randomly selected
from the corresponding control pool
for each incident case. To be able to
analyze VE among the age groups for
which the first and second doses of
varicella vaccine are recommended,
controls selected for cases aged 1
to 3 years had to be <4 years of age,
and controls selected for cases ≥4
years of age had to be ≥4 years of
age. Study staff sent an invitation
letter and contacted parents or
guardians of eligible control subjects
via telephone. The first 3 eligible
respondents who consented to
participate were the controls for each
incident case. Recruited controls
were eligible to be controls for
subsequent incident cases, and if she
or he developed varicella at a later
time point, the subject was eligible
for the study as a case subject.
Data Collection
Study staff obtained verbal consent
and collected data from a parent or
guardian of each subject by telephone
by using a standard questionnaire.
Given limited study resources, we
did not recruit cases and controls
with non-English-speaking parents
or guardians who could not provide
consent because of the language
barrier. The questionnaires captured
information on demographics,
varicella vaccination history, recent
VZV exposures, underlying medical
conditions, and use of immune-
suppressing medications. The case
questionnaire, which has been
described previously, included
additional disease-specific questions
and standardized prompts to obtain
the number of lesions.
23
VASP staff
scheduled home visits to collect
lesion specimens from eligible cases
reported before their rashes had
resolved. The CDC National VZV
Laboratory performed PCR testing
26,
27
on lesion specimens collected
from suspected varicella cases. For
a few cases, VZV-specific testing was
conducted by hospital or commercial
reference laboratories. Participating
families received a $10–$20 gift
card after completion of study-
related activities, and AV health care
providers were offered a $20 gift
card for every case reported with
lesion specimens collected.
For case and control subjects,
varicella vaccination administration
dates were collected from the
registries used for control selection,
parental records, and the subject’s
health care provider. Study staff
made efforts to validate vaccination
information for all subjects with the
immunization registry or health care
provider records. If a discrepancy
existed between these 2 sources,
the source with the most complete
information (ie, highest number of
doses) was used. We considered
1-dose varicella vaccination to be
valid when given 4 days before a
child’s first birthday or later. Second-
dose varicella vaccination was
considered valid when administered
≥4 weeks after the first dose. The
last dose also needed to be given
>42 days before rash onset for cases
(breakthrough varicella) or the
matched incident case’s onset for
controls. Those given doses within 42
days were excluded.
Data Analysis
For our main analysis, we combined
data from both sites, because
varicella vaccine coverage and risk
for breakthrough varicella have not
differed between sociodemographic
subgroups,
13, 28
and the directions
of estimates from each site were
similar. We used 2 case definitions
for varicella: clinically diagnosed
and laboratory confirmed. Varicella
severity was categorized based on
the number of lesions reported as
mild (<50 lesions), moderate (50–500
lesions), or severe (>500 lesions).
The following VE estimates were
calculated to examine protection
against any varicella or moderate or
severe disease alone (≥50 lesions):
incremental 2-dose VE (2 doses
versus 1 dose), overall 2-dose VE
(2 doses versus unvaccinated), and
overall 1-dose VE (1 dose versus
unvaccinated). All VE estimates were
calculated with Greenwood and Yule’s
formula: VE = 1 − relative risk (RR).
19
In our study, RR refers to the risk
of developing varicella among the
subgroup with the higher number
of varicella vaccine doses compared
with the subgroup with fewer or no
doses and was estimated with an
odds ratio (OR) from conditional
logistic regression to account for the
matching variable (age). We were
able to adjust for other potential
confounders when examining VE
against clinically diagnosed disease
3