Merge branch 'feat/test-data-for-lab-and-health-insurance' of https:/…

…/github.com/steno-aarhus/osdc into test/create-small-fake-data

# Conflicts:
#	DESCRIPTION
#	dev/create_test_data.R

.Rbuildignore

@@ -1,6 +1,7 @@
 ^.*\.Rproj$
 ^\.Rproj\.user$
 ^LICENSE\.md$
+^data-raw$
 ^\.github$
 ^data-raw$
 ^dev$

DESCRIPTION

@@ -28,6 +28,7 @@ Imports:
     rlang
 Suggests:
     arrow,
+    codeCollection,
     data.table,
     dbplyr,
     duckdb,

data-raw/testdata.R

@@ -0,0 +1,298 @@
+# Script to generate synthetic data for tests
+
+# Load required libraries
+library(tidyverse)
+library(data.table)
+library(here)
+library(lubridate)
+
+# Set seed for reproducibility
+set.seed(123)
+
+# Set number of rows to create for fake data
+number_rows <- 1000
+
+
+# Functions ---------------------------------------------------------------
+
+#' Create a vector of fake PNR (person ID number).
+#'
+#' These numbers range from 001 to 100.
+#'
+#' @param n The number (length) of items to output.
+#' @param number_subjects Number of patients to create
+#'
+#' @return A vector.
+#'
+#' @examples
+#' create_fake_pnr(10)
+create_fake_pnr <- function(n, number_subjects = 200, offset = 0) {
+  (1 + offset):(number_subjects + offset) |>
+    # Pad the string to match the number values, so that if 1000 `number_subjects`
+    # is given, the string length of the output is 4 characters wide (e.g. "0001").
+    stringr::str_pad(width = stringr::str_length(number_subjects), pad = "0") |>
+    sample(number_rows, replace = TRUE)
+}
+
+#' Create a vector of fake ATC Codes.
+#'
+#' @inheritParams create_fake_pnr
+#'
+#' @return A vector.
+#'
+#' @examples
+#' create_fake_atc(10)
+create_fake_atc <- function(n) {
+  codeCollection::ATCKoodit |>
+    tibble::as_tibble() |>
+    dplyr::filter(stringr::str_length(Koodi) == 7) |>
+    dplyr::pull(Koodi) |>
+    sample(n, replace = TRUE)
+}
+
+create_fake_indication <- function(number_rows) {
+  sample(1:9e8, number_rows, replace = TRUE) |>
+    stringr::str_trunc(width = 7, ellipsis = "") |>
+    stringr::str_pad(width = 7, pad = "0")
+}
+
+assign_drugname_from_atc <- function(data) {
+  codeCollection::ATCKoodit |>
+    tibble::as_tibble() |>
+    dplyr::select(ATC = Koodi, drugname = en) |>
+    dplyr::right_join(data, by = "ATC", relationship = "many-to-many")
+}
+
+# Medication data ---------------------------------------------------------
+
+# Pseudo-lmdb:
+
+## Non-diabetes data (ID 1-200):
+
+# Create a tibble with 1000 rows from 200 individuals
+med_df <- tibble(
+  # ID variable
+  pnr = create_fake_pnr(number_rows),
+  # Date of purchase
+  eksd = as_date(sample(
+    seq(as_date("1995-01-01"), as_date("2014-12-31"), by = "day"), 1000,
+    replace = TRUE
+  )),
+  # Number of packages
+  apk = sample(1:3, 1000, replace = TRUE),
+  # Indication for treatment
+  indo = create_fake_indication(number_rows),
+  # ATC code
+  ATC = create_fake_atc(number_rows),
+  # Volume
+  volume = sample(20:100, 1000, replace = TRUE)
+) |>
+  assign_drugname_from_atc()
+
+## Diabetes data (ID 201-250):
+
+# Create a dataframe with 1000 rows from 50 individuals
+med_a10_df <- data.frame(
+  pnr = create_fake_pnr(number_rows, number_subjects = 50, offset = 200),
+  # ID variable
+  eksd = as.Date(sample(
+    seq(as.Date("1995-01-01"), as.Date("2014-12-31"), by = "day"), 1000,
+    replace = TRUE
+  )),
+  # Date of purchase
+  apk = sample(1:3, 1000, replace = TRUE),
+  # Number of packages
+  indo = create_fake_indication(number_rows),
+  # Indication for treatment
+  ATC = paste(
+    rep(
+      c(
+        "A10AB",
+        "A10AC",
+        "A10AD",
+        "A10AE",
+        "A10BA",
+        "A10BB",
+        "A10BD",
+        "A10BG",
+        "A10BH",
+        "A10BJ",
+        "A10BK",
+        "A10BX"
+      ),
+      80
+    ),
+    sample(0:9, 1000, replace = TRUE),
+    sample(0:9, 1000, replace = TRUE),
+    sep = ""
+  ),
+  # ATC code
+  volume = sample(20:100, 1000, replace = TRUE) # Volume
+)
+
+# Hardcode half of purchases to be metformin, liraglutid or semaglutid:
+
+med_a10_df[sample(nrow(med_a10_df), nrow(med_a10_df) / 2), ]$ATC <-
+  sample(c("A10BA02", "A10BJ02", "A10BJ06"),
+    nrow(med_a10_df) / 2,
+    replace = TRUE
+  )
+
+# Assign drugnames:
+med_a10_df <- med_a10_df |>
+  assign_drugname_from_atc()
+
+replaceDrugNames <- function(data) {
+  # Check if the data frame contains the necessary columns
+  if (!all(c("ATC", "drugname") %in% colnames(data))) {
+    cat("Required columns not found in the data frame.\n")
+    return(NULL)
+  }
+
+  # Define replacement mappings for ATC codes
+  replacement_mappings <- list(
+    "A10BJ02" = "Saxenda",
+    "A10BJ06" = "Wegovy Flextouch"
+  )
+
+  # Iterate through rows and make replacements
+  for (i in 1:nrow(data)) {
+    atc_code <- data$ATC[i]
+    if (atc_code %in% names(replacement_mappings)) {
+      # Check if the ATC code is in the mappings
+      if (runif(1) < 0.5) {
+        # Replace with the corresponding drug name with 50% probability
+        data$drugname[i] <- replacement_mappings[[atc_code]]
+      }
+    }
+  }
+
+  return(data)
+}
+
+# Apply the function to create drug names
+med_a10_df <- replaceDrugNames(med_a10_df)
+
+setDT(med_a10_df)
+setDT(med_df)
+
+# Handcode a few false-positive cases with purchases of metformin:
+med_df[pnr %in% c(sprintf("%03d", 180:190)), `:=`(
+  indo = sample(c("0000092", "0000276", "0000781"), nrow(med_df[pnr %in% c(sprintf("%03d", 180:190))]), replace = TRUE),
+  ATC = "A10BA02",
+  drugname = "Metformin"
+)]
+
+# Handcode a few false-positive cases with purchases of Saxenda:
+med_df[pnr %in% c(sprintf("%03d", 190:195)), `:=`(
+  ATC = "A10BJ02",
+  drugname = "Saxenda"
+)]
+
+# Handcode a few false-positive cases with purchases of Wegovy:
+med_df[pnr %in% c(sprintf("%03d", 195:200)), `:=`(
+  ATC = "A10BJ06",
+  drugname = "Wegovy Flextouch"
+)]
+
+# Combine the two:
+med_df <- rbind(med_df, med_a10_df)
+
+setDT(med_df)
+
+# Hospital diagnoses ------------------------------------------------------
+
+# lpr_adm -----------------------------------------------------------------
+
+
+
+
+# lpr_diag ----------------------------------------------------------------
+
+# Health Service data -----------------------------------------------------
+
+#' Create synthetic health insurance data
+#'
+#' @param num_samples Number of samples to create
+#'
+#' @return A [tibble::tibble()] with columns `pnr`, `BARNMAK`, `SPECIALE`, and `HONUGE`.
+#'
+#' @examples
+#' create_test_hi_df(num_samples = 100)
+create_test_hi_df <- function(num_samples) {
+  tibble::tibble(
+    pnr = create_fake_pnr(num_samples),
+
+    # BARNMAK: service performed on patient' child or not (binary)
+    # 1 = child, 0 = not, 5% are 1's
+    BARNMAK = sample(c(0, 1), num_samples,
+      replace = TRUE,
+      prob = c(0.95, 0.05)
+    ),
+
+    # SPECIALE: service code (6-digit int)
+    # 50% random samples between 100000 and 600000
+    # 50% random samples from 540000 to 549999
+    SPECIALE = ifelse(
+      # repeat 0 and 1 num_samples times and randomise them
+      sample(rep(c(0, 1), length.out = num_samples)),
+      # sample 100000:600000 for all 1's
+      sample(100000:600000, num_samples, replace = TRUE),
+      # sample 540000:549999 for all 0's
+      sample(540000:549999, num_samples, replace = TRUE)
+    ),
+
+    # HONUGE: year/week of the service being billed (4-digit chr)
+    # first and second digits are random numbers between 01-52
+    # third and fourth digits are random numbers corresponding to 1990 onward
+    HONUGE = sprintf(
+      "%02d%02d",
+      sample(1:52, num_samples, replace = TRUE),
+      sample(c(90:99, 01:22), num_samples, replace = TRUE)
+    )
+  )
+}
+
+# create test health insurance df with 100 rows
+health_insurance_df <- create_test_hi_df(num_samples = 1000)
+
+# Laboratory data ---------------------------------------------------------
+
+#' Create synthetic lab data
+#'
+#' @param num_samples Number of samples to create
+#'
+#' @return A [tibble::tibble()] with columns `pnr`, `SAMPLINGDATE`, `ANALYSISCODE`, and `VALUE`.
+#'
+#' @examples
+#' create_test_lab_df(num_samples = 100)
+create_test_lab_df <- function(num_samples) {
+  tibble::tibble(
+    pnr = create_fake_pnr(num_samples),
+
+    # SAMPLINGDATE: date of sample (date)
+    # random dates between 1995-01-01 and 2015-12-31
+    SAMPLINGDATE = sample(
+      seq(lubridate::as_date("1995-01-01"), lubridate::as_date("2015-12-31"), by = "day"),
+      num_samples,
+      replace = TRUE
+    ),
+
+    # ANALYSISCODE: npu code of analysis type (chr)
+    # 50% is either NPU27300 or NPU03835
+    # other 50% is 'NPU'+random sample from 10000:99999
+    ANALYSISCODE = sample(
+      c(sample(c("NPU27300", "NPU03835"), num_samples / 2, replace = TRUE),
+      paste0("NPU", sample(10000:99999, num_samples / 2, replace = TRUE)))
+    ),
+
+    # VALUE: numerical result of test (num)
+    # random decimal number between 0.1-99.9
+    VALUE = runif(num_samples, 0.1, 99.9)
+  )
+}
+
+# create test lab df with 100 rows
+test_lab_df <- create_test_lab_df(num_samples = 1000)
+
+# usethis::use_data(test_lab_df, overwrite = TRUE, internal = TRUE)