Mapping between clinical coding systems

Source: vignettes/map_codes.Rmd 
library(codemapper)
library(dplyr)
#> 
#> Attaching package: 'dplyr'
#> The following objects are masked from 'package:stats':
#> 
#>     filter, lag
#> The following objects are masked from 'package:base':
#> 
#>     intersect, setdiff, setequal, union

all_lkps_maps_dummy <- build_all_lkps_maps_dummy()

Basic usage

Use map_codes() to map a vector of codes from one clinical coding system to another:

map_codes(codes = "G20..",
          from = "read2",
          to = "icd10",
          all_lkps_maps = all_lkps_maps_dummy)
#> # A tibble: 1 × 3
#>   code  description                      code_type
#>   <chr> <chr>                            <chr>    
#> 1 I10X  Essential (primary) hypertension icd10

Alternatively, use get_mapping_df() to return a mapping data frame:

get_mapping_df(from = "read3",
               to = "icd10",
               all_lkps_maps = all_lkps_maps_dummy)
#> # A tibble: 5 × 2
#>   read3 icd10
#>   <chr> <chr>
#> 1 X40J4 E109 
#> 2 C10.. E149 
#> 3 XaIP9 L721 
#> 4 XE0e0 N390 
#> 5 XE0Uc I10X

Filters

Mapping between different clinical coding systems is not perfect and should be manually reviewed. Some mapping tables list multiple possible code mappings, with a separate column indicating the type of mapping.1 For example, the Read 2 to Read 3 mapping table includes an ‘IS_ASSURED’ column, where ‘1’ and ‘0’ indicate that a code mapping is or is not assured respectively:

all_lkps_maps_dummy$read_v2_read_ctv3 %>% 
  dplyr::select(READV2_CODE,
                READV3_CODE,
                IS_ASSURED)
#> # A tibble: 26 × 3
#>    READV2_CODE READV3_CODE IS_ASSURED
#>    <chr>       <chr>       <chr>     
#>  1 C106.       XE10H       1         
#>  2 C106.       X00Ag       1         
#>  3 C106.       XE15k       1         
#>  4 C106.       Xa0lK       0         
#>  5 C106.       XaPmX       1         
#>  6 C108.       X40J4       1         
#>  7 C108.       X40J4       0         
#>  8 C108.       X40J4       1         
#>  9 C108.       X40J4       0         
#> 10 C108.       X40J4       1         
#> # … with 16 more rows

Similarly, the Read 3 to ICD10 mapping table includes ‘mapping_status’ and ‘refine_flag’ columns:

all_lkps_maps_dummy$read_ctv3_icd10 %>% 
  dplyr::select(read_code:refine_flag)
#> # A tibble: 36 × 4
#>    read_code icd10_code mapping_status refine_flag
#>    <chr>     <chr>      <chr>          <chr>      
#>  1 X40J4     E109       D              C          
#>  2 X40J4     E10        A              M          
#>  3 X40J4     O240       R              C          
#>  4 C10..     E149       D              C          
#>  5 C10..     E14        A              M          
#>  6 C10..     E109       R              C          
#>  7 C10..     E119       R              C          
#>  8 C10..     E129       R              C          
#>  9 C10..     E139       R              C          
#> 10 C10..     O249       R              C          
#> # … with 26 more rows

It is important to decide which mappings to include. For example, if no filters are applied then the Read 3 code for sebaceous cyst ‘XaIP9’ will map to a number of ICD10 codes, some of which are sex-specific (‘N508’ and ‘N948’):

map_codes(codes = "XaIP9",
          from = "read3",
          to = "icd10",
          all_lkps_maps = all_lkps_maps_dummy,
          col_filters = NULL)
#> # A tibble: 5 × 3
#>   code  description                                                    code_type
#>   <chr> <chr>                                                          <chr>    
#> 1 H028  Other specified disorders of eyelid                            icd10    
#> 2 L721  Trichilemmal cyst                                              icd10    
#> 3 N508  Other specified disorders of male genital organs               icd10    
#> 4 N608  Other benign mammary dysplasias                                icd10    
#> 5 N948  Other specified conditions associated with female genital org… icd10

Default filters are applied by default_col_filters():

map_codes(codes = "XaIP9",
          from = "read3",
          to = "icd10",
          all_lkps_maps = all_lkps_maps_dummy,
          col_filters = default_col_filters())
#> # A tibble: 1 × 3
#>   code  description       code_type
#>   <chr> <chr>             <chr>    
#> 1 L721  Trichilemmal cyst icd10

See additional notes at https://rmgpanw.gitlab.io/codemapper_notes/.

  1. Refer to the accompanying documentation for UKB resource 592 for further details.↩︎