Skip to contents

Prepare Person-Time and Event Data for SIR/IRR Calculations

Source: R/poisson_analysis.R
pirr_data.Rd

This function merges exposure, response, and population-level data to produce an aggregated dataset for person-time and event rate analysis (e.g., Standardized Incidence Ratios [SIR] or Incidence Rate Ratios [IRR]). It compares disease occurrence across exposure time windows using registry-style longitudinal data.

Usage

pirr_data(
  exposure_diagnoses,
  response_diagnoses,
  pop_dates,
  all_cases = FALSE,
  censoring_age = c(50, 90),
  censoring_date = c(as.Date("1953-01-01"), as.Date("2024-12-31")),
  custom_responses = list()
)

Arguments

exposure_diagnoses

A `data.frame` containing exposure diagnoses. Must include columns `ID`, `DATE`, and `DG`. Typically created with `search_diagnoses()`.

response_diagnoses

A `data.frame` containing response diagnoses. Must include columns `ID`, `DATE`, and `DG`. Typically created with `search_diagnoses()`.

pop_dates

A `data.frame` with population registry information, including `ID`, `DATE_BIRTH`, `DATE_DEATH`, and `DATE_MIGRATION`. Usually from `classify_population()`.

all_cases

Logical; if `TRUE`, follow-up continues after the first response case. If `FALSE`, follow-up stops at the first response diagnosis.

censoring_age

Numeric vector (length 2) specifying the lower and upper ages for follow-up inclusion (e.g., `c(50, 90)`).

censoring_date

A `Date` vector (length 2) defining the administrative start and end of follow-up (e.g., `c(as.Date("1960-01-01"), as.Date("2022-12-31"))`).

custom_responses

Optional named list defining custom response diagnosis groupings. For example: `list(Any_fracture = "ankle+forearm+hip+humerus+vertebral", Osteoporotic = "forearm+hip+humerus+vertebral", Hip = "hip")`.

Value

A `data.frame` summarizing:

pyrs

Person-years within each exposure and age stratum.

Death

Count of deaths within each stratum.

Diagnosis counts

Optional columns for each response diagnosis or custom grouping defined in `custom_responses`.

caika

Exposure time category (e.g., `<1y`, `1–4y`, `5–9y`, etc.).

Age

Age group at risk.

Details

The function requires the **heaven** package for Lexis splitting utilities. See: [https://github.com/tagteam/heaven](https://github.com/tagteam/heaven)

The function performs the following steps:

  • Extracts and merges first exposure and response diagnoses per individual.

  • Computes time differences between exposure and response dates.

  • Splits follow-up time using `heaven::lexisSeq()` and `heaven::lexisTwo()`.

  • Categorizes person-time into exposure windows (`<1y`, `1–4y`, `5–9y`, `10–14y`, `15+y`).

  • Optionally aggregates diagnoses using custom groupings from `custom_responses`.

  • Restricts follow-up to the age range specified in `censoring_age`.

Examples

if (FALSE) { # \dontrun{
result <- pirr_data(
  exposure_diagnoses = exposure_data,
  response_diagnoses = response_data,
  pop_dates = population_data,
  censoring_age = c(50, 90),
  censoring_date = c(as.Date("1960-01-01"), as.Date("2023-12-31")),
  custom_responses = list(
    Any_fracture = "ankle+forearm+hip+humerus+vertebral",
    Osteoporotic = "forearm+hip+humerus+vertebral",
    Hip = "hip"
  )
)
} # }