This function builds pairs of vaccinated and unvaccinated
individuals with similar characteristics. The function relies on the matching
algorithm implemented in the package {MatchIt}, setting, by default,
method = "nearest", ratio = 1, and distance = "mahalanobis".
Exact and near characteristics are accepted, passed in the parameters exact
and nearest, respectively. The parameter nearest must be provided
together with the calipers as a named vector (e.g.,
nearest = c(characteristic1 = n1, characteristic2 = n2), where n1
and n2 are the calipers). The default matching method of the function is
static. This means that pairs are matched once, without taking into
account their vaccination, censoring, and outcome dates. After this, the
pairs whose exposition time do not coincide are removed to avoid negative
time-to-events.
The function returns a matched and adjusted by exposition cohort, with
information of the beginning of follow-up period of pairs (t0_follow_up),
corresponding to the vaccination date of the vaccinated individual, the
individual time-to-event (time_to_event) and the outcome status
(outcome_status), both taking into account the right-censoring dates.
Pairs are censored if the vaccinated or unvaccinated partner was previously
censored (i.e., if censoring_date_col is informed) and the censor occurs
before their outcomes. Rolling calendar matching method will be included in
future releases.
Usage
match_cohort(
data_set,
outcome_date_col,
censoring_date_col,
start_cohort,
end_cohort,
method = "static",
nearest = NULL,
exact = NULL,
immunization_date_col = "immunization_date",
vacc_status_col = "vaccine_status",
vaccinated_status = "v",
unvaccinated_status = "u"
)Arguments
- data_set
data.framewith cohort information (see example).- outcome_date_col
Name of the column that contains the outcome dates.
- censoring_date_col
Name of the column that contains the censoring date.
NULLby default.- start_cohort
Start date of the study.
- end_cohort
End date of the study.
- method
Method to match the cohort. Default is
static.- nearest
Named vector with name(s) of column(s) for
nearestmatching and caliper(s) for each variable (e.g.,nearest = c("characteristic1" = n1, "characteristic2" = n2), wheren1andn2are the calipers). Default isNULL.- exact
Name(s) of column(s) for
exactmatching. Default isNULL.- immunization_date_col
Name of the column that contains the immunization date to set the beginning of the follow-up period (
t0_follow_up). Default isimmunization_date.- vacc_status_col
Name of the column containing the vaccination. Default is
vaccine_status.- vaccinated_status
Status assigned to the vaccinated population. Default is
v.- unvaccinated_status
Status assigned to the unvaccinated population. Default is
u.
Value
object of the class match. List with results from static match:
match: data.frame with adjusted cohort,
summary: matching summary,
balance_all: balance of the cohort before matching,
balance_matched: balance of the cohort after matching.
Four columns are added to the structure provided in data_set:
subclass: ID of, matched pair,
t0_follow_up: beginning of follow-up period for pair,
time_to_event: time to event,
and outcome_status: outcome status
(1:positive, 0: negative).
Examples
# Define start and end dates of the study
start_cohort <- as.Date("2044-01-01")
end_cohort <- as.Date("2044-12-31")
# Create `data.frame` with information on immunization
cohortdata <- make_immunization(
data_set = cohortdata,
outcome_date_col = "death_date",
censoring_date_col = "death_other_causes",
immunization_delay = 14,
vacc_date_col = "vaccine_date_2",
end_cohort = end_cohort
)
# Match the data_set
matching <- match_cohort(
data_set = cohortdata,
outcome_date_col = "death_date",
censoring_date_col = "death_other_causes",
start_cohort = start_cohort,
end_cohort = end_cohort,
method = "static",
exact = "sex",
nearest = c(age = 1)
)
# Check match balance and summary
# `warnings_log = TRUE` displays the logs created
# during the iterative process.
summary(matching, warnings_log = TRUE)
#> Balance all:
#> u v smd
#> age 30.9928364 48.1656078 0.8599792
#> sex_F 0.4834661 0.5758635 0.1859192
#> sex_M 0.5165339 0.4241365 -0.1859192
#>
#> Balance matched:
#> u v smd
#> age 43.1952984 45.6560579 0.1330454
#> sex_F 0.5555154 0.5555154 0.0000000
#> sex_M 0.4444846 0.4444846 0.0000000
#>
#> Summary:
#> u v
#> All 62538 37462
#> Matched 33180 33180
#> Unmatched 29358 4282
#>
#> Warnings:
#> Error at iteration 2: Assertion on 'data_set' failed: Must have at least 1 rows, but has 0 rows.- skipping to next
#> Error at iteration 3: Assertion on 'data_set' failed: Must have at least 1 rows, but has 0 rows.- skipping to next
#> Error at iteration 4: Assertion on 'data_set' failed: Must have at least 1 rows, but has 0 rows.- skipping to next
#> Error at iteration 5: Assertion on 'data_set' failed: Must have at least 1 rows, but has 0 rows.- skipping to next
#> Error at iteration 5: Assertion on 'data_set' failed: Must have at least 1 rows, but has 0 rows.- skipping to next
#> Matches before iterating: 66014
#> Removed before iterating 946
#> Matches after iterating: 66360
#> Removed after iterating 600
# Extract matched data
cohortdata_match <- get_dataset(matching)
# View of mached cohort
head(cohortdata_match)
#> id sex age death_date death_other_causes vaccine_date_1 vaccine_date_2
#> 1 afade1b2 F 37 <NA> <NA> <NA> <NA>
#> 2 556c8c76 M 19 <NA> <NA> <NA> <NA>
#> 3 04edf85a M 50 <NA> <NA> <NA> <NA>
#> 4 7e51a18e F 8 <NA> <NA> <NA> <NA>
#> 5 c5a83f56 M 66 <NA> <NA> <NA> <NA>
#> 6 7f675ec3 M 29 <NA> <NA> 2044-04-09 2044-04-30
#> vaccine_1 vaccine_2 immunization_date vaccine_status subclass t0_follow_up
#> 1 <NA> <NA> <NA> u 1330 2044-12-05
#> 2 <NA> <NA> <NA> u 14154 2044-12-15
#> 3 <NA> <NA> <NA> u 22648 2044-08-14
#> 4 <NA> <NA> <NA> u 8438 2044-12-24
#> 5 <NA> <NA> <NA> u 5550 2044-06-26
#> 6 BRAND1 BRAND1 2044-05-14 v 20552 2044-05-14
#> time_to_event outcome_status
#> 1 26 0
#> 2 16 0
#> 3 139 0
#> 4 7 0
#> 5 188 0
#> 6 231 0