DrugUtilisation vs CapR

Marti Catala, Mike Du, Yuchen Guo, Kim Lopez-Guell, Edward Burn, Xintong Li

Creating different numbers of cohorts using DrugUtilisation

generateDrugUtilisationCohortSet function from DrugUtilisation is used here. Example with duckdb shown in this vignette, but the actual computing time comparison from postgres database is provided in the end.

First connect to database

library(DrugUtilisation)
library(CodelistGenerator)
library(Capr)
library(CDMConnector)
library(dplyr)
library(tictoc)

db_name <- Sys.getenv("...")
host <- Sys.getenv("...")
user <- Sys.getenv("...")
password <- Sys.getenv("...")
port <- Sys.getenv("...")


db <- dbConnect(RPostgres::Postgres(),
  dbname = db_name,
  port = port,
  host = host,
  user = user,
  password = password
)

# The name of the schema that contains the OMOP CDM with patient-level data
cdm_database_schema <- "..."

# The name of the schema where results tables will be created
results_database_schema <- "..."

stem_table <- "..."


# cretae cdm object
cdm <- CDMConnector::cdm_from_con(
  con = db,
  cdm_schema = cdm_database_schema,
  write_schema = results_database_schema
)

A function to benchmark this is provided.

benchmarkGenerateDrugUtilisationCohortSet <- function(
    cdm,
    name = "test",
    conceptSet,
    durationRange = c(1, Inf),
    imputeDuration = "none",
    gapEra = 0,
    priorUseWashout = 0,
    priorObservation = 0,
    cohortDateRange = as.Date(c(NA, NA)),
    limit = "all",
    numberOfCohort = c(1:20)) {
  
  time_record <- list()
  
  for (j in numberOfCohort)
  {

    conceptSetList <- conceptSet[c(1:j)]

    name <- paste0("atc_dus_", j)

    tic()

    cdm <- generateDrugUtilisationCohortSet(
      cdm = cdm,
      name = name,
      conceptSet = conceptSet,
      durationRange = durationRange,
      imputeDuration = imputeDuration,
      gapEra = gapEra,
      priorUseWashout = priorUseWashout,
      priorObservation = priorObservation,
      cohortDateRange = cohortDateRange,
      limit = limit
    )

    cohort_count(cdm[[name]])
    
    t <- toc(quiet = TRUE)
  
    time_record[[paste0("DUs number of cohorts", j)]] <-c("timeTaken" = as.numeric(t$toc - t$tic))
  

  }
  

  return(time_record)
}

Now we create ingredient code using CodelistGenerator to get the code for cohort generation

atcCodes <- getATCCodes(cdm, level = "ATC 1st")

ingredientCodes <- getDrugIngredientCodes(cdm)

time_record_cprdgold_dus <- benchmarkGenerateDrugUtilisationCohortSet(cdm,
  numberOfCohort = c(1:12,
  conceptSet = ingredientCodes
)

Now create benachmarking function using Capr

benchmarkCapr <- function(cdm,
                          conceptSetList = atcCodes,
                          numberOfCohort = c(1:20),
                          observationWindow = continuousObservation(priorDays = 365),
                          qualifiedLimit = "All",
                          studyStartDate = as.Date("2015-01-01"),
                          studyEndDate = as.Date("2022-12-31"),
                          persistenceWindow = 0) {
  time_record <- list()

  for (j in numberOfCohort)
  {
    conceptSetList <- conceptSetList[c(1:j)]
    
    atc_test <- lapply(seq_along(conceptSetList), function(i) {
      cs(conceptSetList[[i]], name = names(conceptSetList)[[i]])
    })

    ch <- lapply(atc_test, function(x) {
      cohort(
        entry = entry(
          drugExposure(x),
          observationWindow = observationWindow,
          qualifiedLimit = qualifiedLimit
        ),
        exit = exit(
          endStrategy = drugExit(
            conceptSet = x,
            persistenceWindow = persistenceWindow
          )
        ),
        era = era(studyStartDate = studyStartDate, studyEndDate = studyEndDate)
      )
    })
    
    names(ch) <- paste0("cohort_", seq_along(ch))
    
    tic()

    
    cdm <- generateCohortSet(cdm = cdm,
                             cohortSet = ch, 
                             name = paste0("capr_cohorts_", j),
                             overwrite = TRUE)
  
    
    t <- toc(quiet = TRUE)

    time_record[[paste0("CapR number of cohorts", j)]] <-c("timeTaken" = as.numeric(t$toc - t$tic))
  
  }
  
  
  return(time_record)
  
}

time_record_cprdgold_capr <- benchmarkCapr(
  cdm = cdm,
  numberOfCohort = c(1:20),
  conceptSetList = ingredientCodes)