R/change_cleanr.koch6.R

In JayAchar/tbcleanr: tbcleanr: tools for cleaning TB datade

#' Clean drug change data
#'
#' Take drug change data from routine TB programmes and clean for 
#' further analysis.
#' 
#' @inheritParams change_cleanr
#' @author Jay Achar 
#' @seealso \code{\link{tbcleanr}}
#' @importFrom assertthat assert_that
#' @importFrom lubridate dmy
#' @importFrom purrr as_mapper modify_at
#' @export

change_cleanr.koch6 <- function(x,
                                add = NULL) {
  
  # save class
  start_class <- class(x)
  
  # check and adjust add arg
  add <- add[add %in% names(x)]
  
  ## Adjust variable names
  # variable names to change
  original_names <- c("changedate",
                        "CBdq", 
                        "CDld", 
                        "CMfx", 
                        "CCfz", 
                        "CLfx",
                        "CImpCln",
                        "CLzd",
                        "CH", 
                        "CR", 
                        "CE", 
                        "CZ",  
                        "CS",  
                        "CKm", 
                        "CCm", 
                        "COfx", 
                        "CEto", 
                        "CCs", 
                        "CPAS", 
                        "CAmx-Clv", 
                        "CPto")
  
  # define new names for drug variables
  new_names <- c("change_dt", 
                 "bdq_change",
                 "dlm_change",
                 "mfx_change",
                 "cfz_change",
                 "lfx_change",
                 "imp_change",
                 "lzd_change",
                 "inh_change",
                 "rif_change",
                 "eth_change",
                 "pza_change",
                 "str_change",
                 "kan_change",
                 "cap_change ",
                 "ofx_change",
                 "eto_change",
                 "cyc_change",
                 "pas_change",
                 "amx_change",
                 "pto_change")
  
  # define drug variables only
  drug_names <- new_names[!new_names == "change_dt"]
  
  # confirm all variables present in data set
  assert_that(all(original_names %in% names(x)))
  
  # check length of old vs new names
  assert_that(length(original_names) == length(new_names))
  
  # Change selected variable names
  name_position <- match(original_names, names(x))
  
  names(x)[name_position] <- new_names
  
  ## Subset and clean data
    ## subset to only keep specified variables
    x <- x[, c("RegistrationNb", new_names, add)]
  
  ## convert drug dose changes to NA
  x[x == 3] <- NA_integer_
  
  ## convert 0 (ND) to NA
  x[x == 0] <- NA_integer_
  
  ## convert 4 (stop permanently) to 2 (stop)
  x[x == 4] <- 2L
  
  # generate count variable to look for missing values in drug changes
  x$row_na <- as.numeric(rowSums(is.na(x[, drug_names])))
  
  # remove rows with all NA in drug variables
  x <- x[x$row_na < length(drug_names), ]
  
  x$row_na <- NULL
  
  # define mapper to convert drugs to factors
  fct_drugs <- purrr::as_mapper(~ factor(.x, 
                                         levels = 1:2, 
                                         labels = c("Start", "Stop")))
  # convert drug variables to factor
  x <- x %>% 
    purrr::modify_at(.f = fct_drugs,
                     .at = drug_names)
  
  # convert to dates
  x$change_dt <- lubridate::dmy(x$change_dt)
  
  # convert all variable to lower case
  names(x) <- tolower(names(x))
  
  # re-apply starting class
  class(x) <- start_class
  
  x
}