Handle analysis of categorical concepts

dev/omop_analyses/analyse_omop_cdm.R

@@ -114,10 +114,43 @@ analyse_summary_stats <- function(cdm) {
     # Combine mean and standard deviation
     bind_rows(df_mean, df_sd)
   }
+
+  # Function to analyse a categorical column - present in observation and measurement
+  # by joining value_as_concept_id to cdm$concept by concept_id
+  analyse_categorical_column <- function(cdm, name_of_table) {
+
+    name_of_id_col <- paste0(name_of_table, "_concept_id")
+
+    # Rename columns and remove empty values
+    table <- cdm[[name_of_table]] |>
+      rename_with(~"concept_id", all_of(name_of_id_col)) |>
+      # beware CDM docs: NULL=no categorical result, 0=categorical result but no mapping
+      filter(value_as_concept_id != 0 & !is.null(value_as_concept_id))
+
+    # count freq and join to concept table to get name
+    df_freq_val_as_concept_named <- table |>
+      count(concept_id, value_as_concept_id) |>
+      left_join(select(cdm$concept, concept_id, concept_name),
+                by = c('value_as_concept_id' = 'concept_id')) |>
+      mutate(concept_id = concept_id,
+             #TODO as agreed 2024-08-16 enable concept_name here and in analyse_numeric_column
+             #OR could join concept_name at end of analyse_summary_stats()
+             #concept_name = concept_name,
+             summary_attribute = "frequency",
+             value_as_string = concept_name,
+             value = n,
+             .keep="none") |>
+      collect()
+  }
+
   # Combine results for all columns
   bind_rows(
+    #numeric results
     cdm$measurement |> analyse_numeric_column(measurement_concept_id, value_as_number),
-    cdm$observation |> analyse_numeric_column(observation_concept_id, value_as_number)
+    cdm$observation |> analyse_numeric_column(observation_concept_id, value_as_number),
+    #categorical results
+    cdm |> analyse_categorical_column("measurement"),
+    cdm |> analyse_categorical_column("observation")
   )
 }