ard_incidence_rate: ARD Incidence Rate

In cardx: Extra Analysis Results Data Utilities

ARD Incidence Rate

Description

Function takes a time at risk variable (time) and event count variable (count) and calculates the incidence rate in person-years.

Incidence rate is calculated as: Total number of events that occurred / Total person-time at risk

Usage

ard_incidence_rate(
  data,
  time,
  count = NULL,
  id = NULL,
  by = NULL,
  strata = NULL,
  n_person_time = 100,
  unit_label = "time",
  conf.level = 0.95,
  conf.type = c("normal", "normal-log", "exact", "byar")
)

Arguments

data	(data.frame) a data frame.
time	(tidy-select) column name of time at risk variable.
count	(tidy-select) column name of variable indicating count of events that occurred. If NULL, each row in data is assumed to correspond to a single event occurrence.
id	(tidy-select) column name used to identify unique subjects in data. If NULL, each row in data is assumed to correspond to a unique subject.
by, strata	(tidy-select) columns to tabulate by/stratify by for summary statistic calculation. Arguments are similar, but with an important distinction: by: results are calculated for all combinations of the columns specified, including unobserved combinations and unobserved factor levels. strata: results are calculated for all observed combinations of the columns specified. Arguments may be used in conjunction with one another.
n_person_time	(numeric) amount of person-time to estimate incidence rate for. Defaults to 100.
unit_label	(string) label for the unit of values in time and estimated person-time output (e.g. "years" for person-years, "days" for person-days, etc.). If the desired person-time estimate unit does not match the current time unit, values of time should be converted to the correct unit during pre-processing. Defaults to "time" (person-time).
conf.level	(numeric) confidence level for the estimated incidence rate.
conf.type	(string) confidence interval type for the estimated incidence rate. One of: normal (default), normal-log, exact, or byar.

Details

The formulas used to calculate the confidence interval for each CI type are as follows, where x_i and t_i represent the number of events and follow-up time for subject i, respectively.

• byar: Byar's approximation of a Poisson CI. A continuity correction of 0.5 is included in the calculation.

  CI = (\sum{x_i} + 0.5) (1 - 1 / (9 \times (\sum{x_i} + 0.5)) \pm Z_{1 - \alpha / 2} / (3 \sqrt{\sum{x_i} + 0.5}))^3 / \sum{t_i}

• normal: Normal CI.

  CI = \sum{x_i} / \sum{t_i} \pm Z_{1 - \alpha / 2} \sqrt{\sum{x_i}} / \sum{t_i}

• normal-log: Normal-Log CI.

  CI = \exp(\log(\sum{x_i} / \sum{t_i}) \pm Z_{1 - \alpha / 2} / \sqrt{\sum{x_i}})

• exact: Exact CI for a Poisson mean.

  CI_{lower} = \chi^2_{\alpha / 2, 2\sum{x_i} + 2} / {2 \sum{t_i}}

  CI_{upper} = \chi^2_{1 - \alpha / 2, 2\sum{x_i} + 2} / {2 \sum{t_i}}

Value

an ARD data frame of class 'card'

Examples

set.seed(1)
data <- data.frame(
  USUBJID = 1:100,
  TRTA = sample(LETTERS[1:3], 100, replace = TRUE),
  AETTE1 = abs(rnorm(100, mean = 0.5)),
  AETOT1 = sample(0:20, 100, replace = TRUE)
)

data |>
  ard_incidence_rate(time = AETTE1, count = AETOT1, id = USUBJID, by = TRTA, unit_label = "years")

cardx documentation built on July 3, 2025, 5:08 p.m.