R/create_ofi.R

In martingerdin/rofi: Utility Functions for the Opportunities For Improvement Project

#' Create Opportunities For Improvement Variable
#'
#' Create opportunities for improvement variable, which is either yes
#' or no
#'
#' Yes: The case was flagged and reviewed in a meeting and the
#' consensus was that there were opportunities for improvement
#' (variable Problemomrade_.FMP), or that the patient died and the
#' death was determined as preventable or potentially preventable
#' (variable Fr1.14 2 or 3).
#' 
#' No: The consensus was that there were no opportunities for
#' improvement, or the nurses in the initial review did not send the
#' case for review because everything was okay (variable VK_avslutad).
#'
#' NA: The case was never selected for review or the patient died but
#' whether the death was preventable is registered as unknown (Fr1.14
#' 999)
#' @param data A data.frame. The data needed to create the
#'     opportunities for improvements variable. Has to include columns
#'     with tne names specified in the arguments quality.review.done,
#'     problem.area, mortality.review.done, and preventable.death.  No
#'     default.
#' @param quality.review.done Character. The name of the quality
#'     review done variable, which indicates if the quality review has
#'     been completed. Defaults to "VK_avslutad".
#' @param problem.area Character. The name of the problem area
#'     variable, which indicates what, if any, problem area that were
#'     identified. Defaults to "Problemomrade_.FMP".
#' @param mortality.review.done Character. The name of the mortatlity
#'     review done variable, which indicates if the mortality review
#'     has been completed. Defaults to "tra_DodsfallsanalysGenomford".
#' @param preventable.death Character. The name of the preventable
#'     death variable, which indicates if a death was prevantable,
#'     potentially proventable, and not preventable. Defaults to
#'     "Fr1.14".
#' @param include.preventable.death Logical. Should preventable death
#'     count as an opportunity for improvement? Defaults to TRUE.
#' @export
create_ofi <- function(data,
                       quality.review.done = "VK_avslutad",
                       problem.area = "Problemomrade_.FMP",
                       mortality.review.done = "tra_DodsfallsanalysGenomford",
                       preventable.death = "Fr1.14",
                       include.preventable.death = TRUE) {
    ## Check arguments
    assertthat::assert_that(is.data.frame(data))
    variable.names <- c(quality.review.done = quality.review.done,
                        problem.area = problem.area,
                        mortality.review.done = mortality.review.done,
                        preventable.death = preventable.death)
    for (variable.name in variable.names) assertthat::assert_that(is.character(variable.name) & length(variable.name) == 1)
    assertthat::assert_that(all(variable.names %in% names(data)))
    assertthat::assert_that(is.logical(include.preventable.death) & length(include.preventable.death) == 1)

    ## Check for data shift, i.e. that input data has changed since
    ## this code was written
    ofi.data <- data[, variable.names]
    names(ofi.data) <- names(variable.names)
    ofi.data <- check_data_shift(ofi.data)

    ## Create ofi variable

    ## The starting point is that either the quality review or
    ## mortality review process is done
    ofi <- !with(ofi.data, quality.review.done == "ja" | mortality.review.done == 1)
    ## If neither of them are done, then ofi should be NA
    ofi[ofi] <- NA
    ## If the problem area is not labelled as okay then there is an
    ## opportunity for improvement
    
    ############
    # New code # Why & (and) for problem.area and | (or) for preventable death?
    ############
    ofi[with(ofi.data, problem.area != "ok" & problem.area != "föredömligt handlagd" 
             & problem.area != "inget problemområde" & problem.area != "nej")] <- TRUE

    if (include.preventable.death) {
        ## If the death is preventable or potentially prevantable then
        ## there is an opportunity for improvement
        ofi[with(ofi.data, preventable.death == "2" | preventable.death == "3")] <- TRUE
        ## If the preventability is unknown then it is unknown if there is
        ## an opportunity for improvement, unless there is an opportunity
        ## for improvement according to the quality review
        ofi[ofi.data$preventable.death == 999 & ofi == FALSE] <- NA
    }
    
    ## Make ofi character
    ofi <- ifelse(ofi, "Yes", "No")
    
    ## Return ofi vector
    return (ofi)
}

check_data_shift <- function(ofi.data) {
    ## Check quality review done variable
    ofi.data$quality.review.done <- tolower(as.character(ofi.data$quality.review.done))
    levels.quality.review.done <- unique(ofi.data$quality.review.done)
    original.levels.quality.review.done <- c("ja", NA, "nej")
    if (!all(levels.quality.review.done %in% original.levels.quality.review.done))
        stop ("Levels in the quality review done variable have changed.")

    ## Check problem area variable
    ofi.data$problem.area <- tolower(as.character(ofi.data$ problem.area))
    levels.problem.area <- unique(ofi.data$problem.area)
    original.levels.problem.area  <- c(NA,"ok","triage på akutmottagningen","vårdnivå","handläggning",
                                       "logistik/teknik","resurs","missad skada","lång tid till op",
                                       "kompetens brist","inget problemområde","föredömligt handlagd",
                                       "kommunikation","handläggning/logistik","traumakriterier/styrning",
                                       "lång tid till dt","triage på akm","tertiär survey","ok","nej",
                                       "dokumentation","bristande rutin","ok","neurokirurg","dokumetation",
                                       "handläggning\r\ndokumentation","annat", "vårdnivå+\r\nmissade skador",
                                       "kommunikation+missad skada", "handläggning prehosp")
    if (!all(levels.problem.area %in% original.levels.problem.area))
        stop ("Levels in the problem area variable have changed.")

    ## Check mortality review done variable
    ofi.data$mortality.review.done <- tolower(as.character(ofi.data$mortality.review.done))
    levels.mortality.review.done <- unique(ofi.data$mortality.review.done)
    original.levels.mortality.review.done <- c(NA, 2, 1)
    if (!all(levels.quality.review.done %in% original.levels.quality.review.done))
        stop ("Levels in the mortality review done variable have changed.")

    ## Check preventable death variable
    levels.preventable.death <- unique(ofi.data$preventable.death)
    original.levels.preventable.death <- c(NA, "1", "3", "2", "999")
    if (!all(levels.preventable.death %in% original.levels.preventable.death))
        stop ("Levels in the preventable death variable have changed.")

    ## Return data
    return (ofi.data)
}