Row-wise flatten_chr() or unlist() to convert string to vector-CodePudding

I'm using a dataset that included diagnoses coding, and I'm trying to see whether follow_up_code is in discharge_codes. Sadly, the discharge codes have been provided as a string, rather than a vector.

mre <- tribble(
 ~patient_id, ~discharge_codes, ~follow_up_code,
 1234       , "A_B_C"         , "A",
 4567       , "D_E_F"         , "C",
 7890.      , "A_C_E".        , "E"
)

I've tried using flatten_chr() with str_split() to allow me to search follow_up_code %in% discharge_codes, but this flattens discharge_codes entirely (rather than by patient), and using rowwise() %>% mutate(... flatten_chr()) errors as ".x must be a list, not a character vector".

I feel I must be missing something, either in the approach I'm taking, or there being a much more straightforward way to achieve this?

CodePudding user response：

You can use grepl to check if the string exists in your other column. In the column match you can see if the string was there (TRUE/FALSE) like this:

mre <- data.frame(patien_id = c(1234, 4567, 7890),
                  discharge_codes = c("A_B_C", "D_E_F", "A_C_E"),
                  follow_up_code = c("A", "C", "E"))

mre$match <- grepl(mre$follow_up_code, mre$discharge_codes)
mre
#>   patien_id discharge_codes follow_up_code match
#> 1      1234           A_B_C              A  TRUE
#> 2      4567           D_E_F              C FALSE
#> 3      7890           A_C_E              E  TRUE

^{Created on 2022-07-08 by the reprex package (v2.0.1)}

CodePudding user response：

You can convert the text column into a character vector and then see if the code is within that vector. The benefit of this is that the discharge_codes are now available for other uses, if needed.

library(dplyr)
library(purrr)
library(stringr)

mre %>% 
  mutate(discharge_codes = str_split(discharge_codes, "_"),
         match = map2_lgl(discharge_codes, follow_up_code, ~ .y %in% .x))

You can see that discharge_codes is now a list column with character vectors.

# A tibble: 3 x 4
  patient_id discharge_codes follow_up_code match
       <dbl> <list>          <chr>          <lgl>
1       1234 <chr [3]>       A              TRUE 
2       4567 <chr [3]>       C              FALSE
3       7890 <chr [3]>       E              TRUE

CodePudding user response：

You can simply do this with the help of str_detect() from stringr package.

library(dplyr)
library(stringr)
library(tibble)

mre <- tibble::tribble(
  ~patient_id, ~discharge_codes, ~follow_up_code,
  1234,          "A_B_C",             "A",
  4567,          "D_E_F",             "C",
  7890,          "A_C_E",             "E"
)

mre %>% 
  mutate(
    matched = str_detect(discharge_codes, follow_up_code)
  )

#> # A tibble: 3 × 4
#>   patient_id discharge_codes follow_up_code matched
#>        <dbl> <chr>           <chr>          <lgl>  
#> 1       1234 A_B_C           A              TRUE   
#> 2       4567 D_E_F           C              FALSE  
#> 3       7890 A_C_E           E              TRUE

^{Created on 2022-07-08 by the reprex package (v2.0.1)}

CodePudding user response：

This one will automatically filter all rows where there is a match:

library(tidyverse)

mre %>% 
  mutate(match = str_match_all(discharge_codes, follow_up_code)) %>% 
  unnest(c(match))

  patient_id discharge_codes follow_up_code match[,1]
       <dbl> <chr>           <chr>          <chr>    
1       1234 A_B_C           A              A        
2       7890 A_C_E           E              E