R/clean_process_clinical_data.R

In pwatrick/DrugRepurposingToolKit: A High-Throughput Approach Integrating Gene Expression Signatures and Clinical Data to Repurpose Drugs

#' Clean and process EHR data
#'
#'
#' @description
#' \code{clean_process_clinical_data} cleans and processes clinical EHR data.
#'
#' @details
#' This function cleans and processes clinical data from electronic health
#' record (EHR) databases that have adopted the Observational Medical Outcomes
#' Partnership (OMOP) Common Data Model (CDM) format.
#'
#' Updated: 2021-03-27
#'
#' @param drug                  A string, e.g., "simvastatin"
#' @param phenotype             A string, e.g., "Hyperlipidemia"
#' @param biomarker             A string, e.g., "LDL Cholesterol"
#' @param indication_drugs      A tibble with columns
#'                              c('rxcui_in', 'drug_desc', 'indication'),
#'                              eg, DrugRepurposingToolKit::drugsHyperlipidemia
#' @param concept_id_exclusions A vector, e.g., c(2212451). Used for excluding
#'                              concept_ids which may not contain the correct
#'                              values
#' @param r_covariates          A tibble with columns
#'                              c('person_id','dob','gender_concept_id',
#'                              'race_concept_id', drug_concept_id',
#'                              'first_drug_exposure','last_drug_exposure')
#' @param r_drugs               A tibble with columns
#'                              c('person_id','start_date','end_date',
#'                              'drug_exposure_start_date','drug_concept_id')
#' @param r_biomarkers          A tibble with columns
#'                              c('person_id','start_date','measurement_date',
#'                              'end_date', 'measurement_concept_id',
#'                              'value_as_number')
#' @export

clean_process_clinical_data <- function(drug, phenotype, biomarker, indication_drugs, concept_id_exclusions,
                                  r_covariates, r_drugs, r_biomarkers) {

  #Make sure that patients are in all three tables
  r_biomarker_temp <- r_biomarkers %>%
    select(person_id) %>% distinct() %>%
    rename(biomarker_person_id = person_id) %>%
    mutate(biomarker_tbl = 1)
  r_covariates_temp <- r_covariates %>%
    select(person_id) %>% distinct() %>%
    rename(covariates_person_id = person_id) %>%
    mutate(covariates_tbl = 1)
  r_drugs_temp <- r_drugs %>%
    select(person_id) %>%
    distinct() %>%
    rename(drugs_person_id = person_id) %>% mutate(drugs_tbl = 1)
  r_merged <- inner_join(r_biomarker_temp, r_drugs_temp, by = c("biomarker_person_id" = "drugs_person_id"))
  r_merged <- inner_join(r_merged, r_covariates_temp, by = c("biomarker_person_id" = "covariates_person_id"))
  overlapping_patients <- r_merged$biomarker_person_id
  rm(r_biomarker_temp, r_covariates_temp, r_drugs_temp, r_merged); gc()

  #Remove patients that are not in all three tables
  r_covariates <- r_covariates %>% filter(person_id %in% overlapping_patients)
  r_drugs <- r_drugs %>% filter(person_id %in% overlapping_patients)
  r_biomarkers <- r_biomarkers %>% filter(person_id %in% overlapping_patients)

  #Define treatment and baseline periods
  p_obsperiod1 <- r_drugs %>%
    select(person_id, start_date, end_date) %>%
    distinct()
  p_obsperiod2 <- r_covariates %>%
    select(person_id, drug_concept_id, first_drug_exposure, last_drug_exposure) %>%
    distinct()
  p_obsperiod3 <- inner_join(p_obsperiod1, p_obsperiod2, by = "person_id")
  f_obsperiod <- p_obsperiod3 %>%
    DrugRepurposingToolKit::define_observation_periods()
  f_obsperiod <- f_obsperiod %>%
    filter(person_id %in% overlapping_patients)
  rm(p_obsperiod1, p_obsperiod2, p_obsperiod3); gc()

  #Map drugs to ingredients
  p_drugs <- r_drugs %>%
    select(person_id, drug_exposure_start_date, drug_concept_id) %>%
    distinct()
  f_drugs <- p_drugs %>%
    DrugRepurposingToolKit::map_drugs_to_ingredients()
  rm(r_drugs, p_drugs); gc()

  #Process covariates
  p_covariates <- r_covariates %>%
    select(person_id, dob, gender_concept_id, race_concept_id)
  f_covariates <- DrugRepurposingToolKit::process_covariates(p_covariates, f_obsperiod, european = 0)
  rm(p_covariates, r_covariates); gc()

  #Get drug exposures in baseline and treatment periods
  f_drugs_periods <-
    DrugRepurposingToolKit::observation_period_drug_exposures(f_obsperiod, f_drugs)

  #Flag indication drug exposed patients
  ##Remove `drug` from indication drugs list. This checks and removes drug of interest from the list.
  indication_drugs <- indication_drugs %>%
    filter(!str_detect(drug_desc, drug))
  ##Get indication drug exposed patients
  f_indication_drug_exposed <- f_drugs_periods %>%
    filter(rxcui_ingr %in% indication_drugs$rxcui_in)
  ##Flag patients for subanalyses, stratified on exposure to approved drugs for target disease
  f_covariates_flagged <-
    DrugRepurposingToolKit::add_subanalysis_flags(f_indication_drug_exposed, f_covariates)


  #Process biomarker data
  p_biomarkers <-
    DrugRepurposingToolKit::process_biomarker(biomarker_tbl = r_biomarkers,
                                             biomarker_name = biomarker,
                                             covariates_flagged = f_covariates_flagged,
                                             baseline_cols = c('person_id', 'start_date',
                                                               'first_drug_exposure', 'final_end_date'),
                                             inclusion_cols = c('person_id', 'measurement_date',
                                                                'measurement_concept_id', 'value_as_number'),
                                             concept_id_exclusions = concept_id_exclusions)

  #Only keep non-approved drug exposed patients
  f_covariates_flagged <- f_covariates_flagged %>%
    filter(drug_exposed == 0)
  processed_biomarker_values <- inner_join(f_covariates_flagged, p_biomarkers, by = "person_id") %>%
    select(-c("dob", "drug_exposed", "treatment_new_exposed"))

  #Return tibble with processed data of patients not exposed to known approved drugs
  return(processed_biomarker_values)
}