# ==================================================================== # # TITLE # # Antimicrobial Resistance (AMR) Analysis # # # # SOURCE # # https://gitlab.com/msberends/AMR # # # # LICENCE # # (c) 2019 Berends MS (m.s.berends@umcg.nl), Luz CF (c.f.luz@umcg.nl) # # # # This R package is free software; you can freely use and distribute # # it for both personal and commercial purposes under the terms of the # # GNU General Public License version 2.0 (GNU GPL-2), as published by # # the Free Software Foundation. # # # # This R package was created for academic research and was publicly # # released in the hope that it will be useful, but it comes WITHOUT # # ANY WARRANTY OR LIABILITY. # # Visit our website for more info: https://msberends.gitab.io/AMR. # # ==================================================================== # #' Calculate resistance of isolates #' #' @description These functions can be used to calculate the (co-)resistance of microbial isolates (i.e. percentage S, SI, I, IR or R). All functions support quasiquotation with pipes, can be used in \code{dplyr}s \code{\link[dplyr]{summarise}} and support grouped variables, see \emph{Examples}. #' #' \code{portion_R} and \code{portion_IR} can be used to calculate resistance, \code{portion_S} and \code{portion_SI} can be used to calculate susceptibility.\cr #' @param ... one or more vectors (or columns) with antibiotic interpretations. They will be transformed internally with \code{\link{as.rsi}} if needed. Use multiple columns to calculate (the lack of) co-resistance: the probability where one of two drugs have a resistant or susceptible result. See Examples. #' @param minimum the minimal amount of available isolates. Any number lower than \code{minimum} will return \code{NA} with a warning. The default number of \code{30} isolates is advised by the Clinical and Laboratory Standards Institute (CLSI) as best practice, see Source. #' @param as_percent a logical to indicate whether the output must be returned as a hundred fold with \% sign (a character). A value of \code{0.123456} will then be returned as \code{"12.3\%"}. #' @param also_single_tested a logical to indicate whether (in combination therapies) also observations should be included where not all antibiotics were tested, but at least one of the tested antibiotics contains a target interpretation (e.g. S in case of \code{portion_S} and R in case of \code{portion_R}). \strong{This would lead to selection bias in almost all cases.} #' @param data a \code{data.frame} containing columns with class \code{rsi} (see \code{\link{as.rsi}}) #' @param translate_ab a column name of the \code{\link{antibiotics}} data set to translate the antibiotic abbreviations to, using \code{\link{abname}}. This can be set with \code{\link{getOption}("get_antibiotic_names")}. #' @param combine_IR a logical to indicate whether all values of I and R must be merged into one, so the output only consists of S vs. IR (susceptible vs. non-susceptible) #' @details \strong{Remember that you should filter your table to let it contain only first isolates!} Use \code{\link{first_isolate}} to determine them in your data set. #' #' These functions are not meant to count isolates, but to calculate the portion of resistance/susceptibility. Use the \code{\link[AMR]{count}} functions to count isolates. \emph{Low counts can infuence the outcome - these \code{portion} functions may camouflage this, since they only return the portion albeit being dependent on the \code{minimum} parameter.} #' #' \code{portion_df} takes any variable from \code{data} that has an \code{"rsi"} class (created with \code{\link{as.rsi}}) and calculates the portions R, I and S. The resulting \emph{tidy data} (see Source) \code{data.frame} will have three rows (S/I/R) and a column for each variable with class \code{"rsi"}. #' #' The old \code{\link{rsi}} function is still available for backwards compatibility but is deprecated. #' \if{html}{ #' \cr\cr #' To calculate the probability (\emph{p}) of susceptibility of one antibiotic, we use this formula: #' \out{
}\figure{mono_therapy.png}\out{
} #' To calculate the probability (\emph{p}) of susceptibility of more antibiotics (i.e. combination therapy), we need to check whether one of them has a susceptible result (as numerator) and count all cases where all antibiotics were tested (as denominator). \cr #' \cr #' For two antibiotics: #' \out{
}\figure{combi_therapy_2.png}\out{
} #' \cr #' For three antibiotics: #' \out{
}\figure{combi_therapy_3.png}\out{
} #' \cr #' And so on. #' } #' @source \strong{M39 Analysis and Presentation of Cumulative Antimicrobial Susceptibility Test Data, 4th Edition}, 2014, \emph{Clinical and Laboratory Standards Institute (CLSI)}. \url{https://clsi.org/standards/products/microbiology/documents/m39/}. #' #' Wickham H. \strong{Tidy Data.} The Journal of Statistical Software, vol. 59, 2014. \url{http://vita.had.co.nz/papers/tidy-data.html} #' @seealso \code{\link[AMR]{count}_*} to count resistant and susceptible isolates. #' @keywords resistance susceptibility rsi_df rsi antibiotics isolate isolates #' @return Double or, when \code{as_percent = TRUE}, a character. #' @rdname portion #' @name portion #' @export #' @inheritSection AMR Read more on our website! #' @examples #' # septic_patients is a data set available in the AMR package. It is true, genuine data. #' ?septic_patients #' #' # Calculate resistance #' portion_R(septic_patients$amox) #' portion_IR(septic_patients$amox) #' #' # Or susceptibility #' portion_S(septic_patients$amox) #' portion_SI(septic_patients$amox) #' #' # Do the above with pipes: #' library(dplyr) #' septic_patients %>% portion_R(amox) #' septic_patients %>% portion_IR(amox) #' septic_patients %>% portion_S(amox) #' septic_patients %>% portion_SI(amox) #' #' septic_patients %>% #' group_by(hospital_id) %>% #' summarise(p = portion_S(cipr), #' n = n_rsi(cipr)) # n_rsi works like n_distinct in dplyr #' #' septic_patients %>% #' group_by(hospital_id) %>% #' summarise(R = portion_R(cipr, as_percent = TRUE), #' I = portion_I(cipr, as_percent = TRUE), #' S = portion_S(cipr, as_percent = TRUE), #' n = n_rsi(cipr), # works like n_distinct in dplyr #' total = n()) # NOT the amount of tested isolates! #' #' # Calculate co-resistance between amoxicillin/clav acid and gentamicin, #' # so we can see that combination therapy does a lot more than mono therapy: #' septic_patients %>% portion_S(amcl) # S = 67.1% #' septic_patients %>% count_all(amcl) # n = 1576 #' #' septic_patients %>% portion_S(gent) # S = 74.0% #' septic_patients %>% count_all(gent) # n = 1855 #' #' septic_patients %>% portion_S(amcl, gent) # S = 92.0% #' septic_patients %>% count_all(amcl, gent) # n = 1517 #' #' #' septic_patients %>% #' group_by(hospital_id) %>% #' summarise(cipro_p = portion_S(cipr, as_percent = TRUE), #' cipro_n = count_all(cipr), #' genta_p = portion_S(gent, as_percent = TRUE), #' genta_n = count_all(gent), #' combination_p = portion_S(cipr, gent, as_percent = TRUE), #' combination_n = count_all(cipr, gent)) #' #' # Get portions S/I/R immediately of all rsi columns #' septic_patients %>% #' select(amox, cipr) %>% #' portion_df(translate = FALSE) #' #' # It also supports grouping variables #' septic_patients %>% #' select(hospital_id, amox, cipr) %>% #' group_by(hospital_id) %>% #' portion_df(translate = FALSE) #' #' #' \dontrun{ #' #' # calculate current empiric combination therapy of Helicobacter gastritis: #' my_table %>% #' filter(first_isolate == TRUE, #' genus == "Helicobacter") %>% #' summarise(p = portion_S(amox, metr), # amoxicillin with metronidazole #' n = count_all(amox, metr)) #' } portion_R <- function(..., minimum = 30, as_percent = FALSE, also_single_tested = FALSE) { rsi_calc(..., type = "R", include_I = FALSE, minimum = minimum, as_percent = as_percent, also_single_tested = also_single_tested, only_count = FALSE) } #' @rdname portion #' @export portion_IR <- function(..., minimum = 30, as_percent = FALSE, also_single_tested = FALSE) { rsi_calc(..., type = "R", include_I = TRUE, minimum = minimum, as_percent = as_percent, also_single_tested = also_single_tested, only_count = FALSE) } #' @rdname portion #' @export portion_I <- function(..., minimum = 30, as_percent = FALSE, also_single_tested = FALSE) { rsi_calc(..., type = "I", include_I = FALSE, minimum = minimum, as_percent = as_percent, also_single_tested = also_single_tested, only_count = FALSE) } #' @rdname portion #' @export portion_SI <- function(..., minimum = 30, as_percent = FALSE, also_single_tested = FALSE) { rsi_calc(..., type = "S", include_I = TRUE, minimum = minimum, as_percent = as_percent, also_single_tested = also_single_tested, only_count = FALSE) } #' @rdname portion #' @export portion_S <- function(..., minimum = 30, as_percent = FALSE, also_single_tested = FALSE) { rsi_calc(..., type = "S", include_I = FALSE, minimum = minimum, as_percent = as_percent, also_single_tested = also_single_tested, only_count = FALSE) } #' @rdname portion #' @importFrom dplyr %>% select_if bind_rows summarise_if mutate group_vars select everything #' @export portion_df <- function(data, translate_ab = getOption("get_antibiotic_names", "official"), minimum = 30, as_percent = FALSE, combine_IR = FALSE) { if (!"data.frame" %in% class(data)) { stop("`portion_df` must be called on a data.frame") } if (data %>% select_if(is.rsi) %>% ncol() == 0) { stop("No columns with class 'rsi' found. See ?as.rsi.") } if (as.character(translate_ab) == "TRUE") { translate_ab <- "official" } options(get_antibiotic_names = translate_ab) resS <- summarise_if(.tbl = data, .predicate = is.rsi, .funs = portion_S, minimum = minimum, as_percent = as_percent) %>% mutate(Interpretation = "S") %>% select(Interpretation, everything()) if (combine_IR == FALSE) { resI <- summarise_if(.tbl = data, .predicate = is.rsi, .funs = portion_I, minimum = minimum, as_percent = as_percent) %>% mutate(Interpretation = "I") %>% select(Interpretation, everything()) resR <- summarise_if(.tbl = data, .predicate = is.rsi, .funs = portion_R, minimum = minimum, as_percent = as_percent) %>% mutate(Interpretation = "R") %>% select(Interpretation, everything()) data.groups <- group_vars(data) res <- bind_rows(resS, resI, resR) %>% mutate(Interpretation = factor(Interpretation, levels = c("R", "I", "S"), ordered = TRUE)) %>% tidyr::gather(Antibiotic, Value, -Interpretation, -data.groups) } else { resIR <- summarise_if(.tbl = data, .predicate = is.rsi, .funs = portion_IR, minimum = minimum, as_percent = as_percent) %>% mutate(Interpretation = "IR") %>% select(Interpretation, everything()) data.groups <- group_vars(data) res <- bind_rows(resS, resIR) %>% mutate(Interpretation = factor(Interpretation, levels = c("IR", "S"), ordered = TRUE)) %>% tidyr::gather(Antibiotic, Value, -Interpretation, -data.groups) } if (!translate_ab == FALSE) { if (!tolower(translate_ab) %in% tolower(colnames(AMR::antibiotics))) { stop("Parameter `translate_ab` does not occur in the `antibiotics` data set.", call. = FALSE) } res <- res %>% mutate(Antibiotic = abname(Antibiotic, from = "guess", to = translate_ab)) } res } #' Calculate resistance of isolates #' #' This function is deprecated. Use the \code{\link{portion}} functions instead. #' @inheritParams portion #' @param ab1,ab2 vector (or column) with antibiotic interpretations. It will be transformed internally with \code{\link{as.rsi}} if needed. #' @param interpretation antimicrobial interpretation to check for #' @param ... deprecated parameters to support usage on older versions #' @importFrom dplyr tibble case_when #' @export rsi <- function(ab1, ab2 = NULL, interpretation = "IR", minimum = 30, as_percent = FALSE, ...) { .Deprecated(new = paste0("portion_", interpretation)) if (all(is.null(ab2))) { df <- tibble(ab1 = ab1) } else { df <- tibble(ab1 = ab1, ab2 = ab2) } if (!interpretation %in% c("S", "SI", "IS", "I", "RI", "IR", "R")) { stop("invalid interpretation") } result <- case_when( interpretation == "S" ~ portion_S(df, minimum = minimum, as_percent = FALSE), interpretation %in% c("SI", "IS") ~ portion_SI(df, minimum = minimum, as_percent = FALSE), interpretation == "I" ~ portion_I(df, minimum = minimum, as_percent = FALSE), interpretation %in% c("RI", "IR") ~ portion_IR(df, minimum = minimum, as_percent = FALSE), interpretation == "R" ~ portion_R(df, minimum = minimum, as_percent = FALSE)) if (as_percent == TRUE) { percent(result, force_zero = TRUE) } else { result } }