Accept per-microliter units for blood neutrophil counts

DESCRIPTION

@@ -44,6 +44,7 @@ Suggests:
     markdown,
     shiny,
     ggplot2,
+    RSQLite,
     Eunomia (>= 2.0.0),
     R.utils,
     devtools

NEWS.md

@@ -1,6 +1,8 @@
 # DataQualityDashboard (unreleased)
 This release adds support for OMOP CDM v5.5, drops support for OMOP CDM v5.2, consolidates the concept level threshold files, and refreshes the CDM documentation columns in the threshold files.
 
+- Added `8647` (per microliter, `/uL`) to the plausible units for blood neutrophil count concept `3017732` (LOINC `26499-4`), fixing false unit violations (#698).
+
 ### Removed CDM v5.2 support
 
 **`cdmVersion = "5.2"` is no longer accepted, and the `OMOP_CDMv5.2_Table_Level.csv`, `OMOP_CDMv5.2_Field_Level.csv`, and `OMOP_CDMv5.2_Check_Descriptions.csv` files have been removed.** CDM v5.2 is no longer published in the [CommonDataModel](https://github.com/OHDSI/CommonDataModel) repository, so its threshold files could not be kept in step with the specification, and the package was never tested against it. Sites still on v5.2 should either remain on an earlier release of this package or pass their own threshold files via `tableCheckThresholdLoc` and `fieldCheckThresholdLoc`.

inst/csv/OMOP_CDM_Concept_Level.csv

@@ -404,7 +404,7 @@ MEASUREMENT,MEASUREMENT_CONCEPT_ID,3013429,Basophils [#/volume] in Blood by Auto
 MEASUREMENT,MEASUREMENT_CONCEPT_ID,3023599,MCV [Entitic volume] by Automated count,,,,,,,,8583,5,
 MEASUREMENT,MEASUREMENT_CONCEPT_ID,3036588,Neutrophil cytoplasmic Ab.perinuclear [Presence] in Serum,,,,,,,,"8525,-1",5,
 MEASUREMENT,MEASUREMENT_CONCEPT_ID,4298431,White blood cell count,,,,,,,,"8848,8961,9444,8784,8785",5,
-MEASUREMENT,MEASUREMENT_CONCEPT_ID,3017732,Neutrophils [#/volume] in Blood,,,,,,,,"8848,8961,9444",5,
+MEASUREMENT,MEASUREMENT_CONCEPT_ID,3017732,Neutrophils [#/volume] in Blood,,,,,,,,"8848,8961,9444,8647",5,
 MEASUREMENT,MEASUREMENT_CONCEPT_ID,3024561,Albumin [Mass/volume] in Serum or Plasma,,,,,,,,"8636,8713",5,
 MEASUREMENT,MEASUREMENT_CONCEPT_ID,3034639,Hemoglobin A1c [Mass/volume] in Blood,,,,,,,,"8713,8840,9579,8923",5,
 MEASUREMENT,MEASUREMENT_CONCEPT_ID,3013650,Neutrophils [#/volume] in Blood by Automated count,,,,,,,,"8784,8848,8961,9444,8647,8785",5,

tests/testthat/test-plausibleUnitConceptIds.R

@@ -0,0 +1,115 @@
+library(testthat)
+
+plausibleUnitThresholdFiles <- list.files(
+  system.file("csv", package = "DataQualityDashboard"),
+  pattern = "^OMOP_CDM(v5\\.[2-5])?_Concept_Level\\.csv$"
+)
+
+runNeutrophilUnitCheck <- function(thresholdFile, measurementConceptIds, unitConceptIds) {
+  conceptChecks <- .readThresholdFile("default", thresholdFile)
+  check <- conceptChecks[
+    conceptChecks$cdmTableName == "MEASUREMENT" &
+      conceptChecks$cdmFieldName == "MEASUREMENT_CONCEPT_ID" &
+      conceptChecks$conceptId == 3017732,
+  ]
+  expect_equal(nrow(check), 1, info = thresholdFile)
+
+  databaseFile <- tempfile(fileext = ".sqlite")
+  on.exit(unlink(databaseFile))
+  connectionDetails <- DatabaseConnector::createConnectionDetails(
+    dbms = "sqlite",
+    server = databaseFile
+  )
+  connection <- DatabaseConnector::connect(connectionDetails)
+  on.exit({
+    DatabaseConnector::disconnect(connection)
+    unlink(databaseFile)
+  })
+  DatabaseConnector::executeSql(connection, "
+    CREATE TABLE measurement (
+      measurement_concept_id INTEGER,
+      unit_concept_id INTEGER
+    );
+  ")
+  if (length(measurementConceptIds) > 0) {
+    values <- sprintf(
+      "(%i, %s)",
+      measurementConceptIds,
+      ifelse(is.na(unitConceptIds), "NULL", unitConceptIds)
+    )
+    DatabaseConnector::executeSql(
+      connection,
+      paste0("INSERT INTO measurement VALUES ", paste(values, collapse = ", "), ";")
+    )
+  }
+
+  sql <- SqlRender::loadRenderTranslateSql(
+    sqlFilename = "concept_plausible_unit_concept_ids.sql",
+    packageName = "DataQualityDashboard",
+    dbms = "sqlite",
+    warnOnMissingParameters = FALSE,
+    cdmDatabaseSchema = "main",
+    cdmTableName = check$cdmTableName,
+    cdmFieldName = check$cdmFieldName,
+    conceptId = check$conceptId,
+    plausibleUnitConceptIds = check$plausibleUnitConceptIds,
+    cohort = FALSE
+  )
+  DatabaseConnector::querySql(connection, sql, snakeCaseToCamelCase = TRUE)
+}
+
+test_that("blood neutrophil counts accept /uL and reject unrelated units (#698)", {
+  skip_if_not_installed("RSQLite")
+  expect_gt(length(plausibleUnitThresholdFiles), 0)
+  for (thresholdFile in plausibleUnitThresholdFiles) {
+    result <- runNeutrophilUnitCheck(thresholdFile, 3017732L, 8647L)
+    expect_equal(result$numViolatedRows, 0, info = thresholdFile)
+    expect_equal(result$numDenominatorRows, 1, info = thresholdFile)
+    expect_equal(result$pctViolatedRows, 0, info = thresholdFile)
+
+    result <- runNeutrophilUnitCheck(
+      thresholdFile,
+      rep(3017732L, 4),
+      c(8848L, 8961L, 9444L, 8554L)
+    )
+    expect_equal(result$numViolatedRows, 1, info = thresholdFile)
+    expect_equal(result$numDenominatorRows, 4, info = thresholdFile)
+    expect_equal(result$pctViolatedRows, 1 / 4, info = thresholdFile)
+  }
+})
+
+test_that("neutrophil unit checks preserve zero and NULL denominator behavior", {
+  skip_if_not_installed("RSQLite")
+  expect_gt(length(plausibleUnitThresholdFiles), 0)
+  for (thresholdFile in plausibleUnitThresholdFiles) {
+    # A different measurement concept must not affect either count.
+    result <- runNeutrophilUnitCheck(
+      thresholdFile,
+      c(3017732L, 3000000L),
+      c(0L, 8554L)
+    )
+    expect_equal(result$numViolatedRows, 0, info = thresholdFile)
+    expect_equal(result$numDenominatorRows, 0, info = thresholdFile)
+    expect_equal(result$pctViolatedRows, 0, info = thresholdFile)
+
+    result <- runNeutrophilUnitCheck(
+      thresholdFile,
+      c(3017732L, 3000000L),
+      c(NA_integer_, 8554L)
+    )
+    expect_equal(result$numViolatedRows, 0, info = thresholdFile)
+    expect_equal(result$numDenominatorRows, 1, info = thresholdFile)
+    expect_equal(result$pctViolatedRows, 0, info = thresholdFile)
+  }
+})
+
+test_that("neutrophil unit checks return zero for an empty measurement table", {
+  skip_if_not_installed("RSQLite")
+  expect_gt(length(plausibleUnitThresholdFiles), 0)
+  for (thresholdFile in plausibleUnitThresholdFiles) {
+    result <- runNeutrophilUnitCheck(thresholdFile, integer(), integer())
+    expect_equal(result$numViolatedRows, 0, info = thresholdFile)
+    expect_equal(result$numDenominatorRows, 0, info = thresholdFile)
+    expect_equal(result$pctViolatedRows, 0, info = thresholdFile)
+  }
+})