NHSDataDictionaRy - a package for accessing NHS Data Dictionary with web scraping and other useful functions

Context

This package has been commissioned by the NHS-R community and is intended to be used to web scrape the NHS Data Dictionary website for useful lookup tables. The NHS-R community have been pivotal in getting this package off the ground.

The package is maintained by Gary Hutson - Head of Advanced Analytics at Arden and GEM Commissioning Support Unit and to contact the maintainer directly you can navigate to this site.

Additionally, the package has been developed with generic web scraping functionality to allow other websites containing data tables and elements to be scraped.

Loading the package

To load the package, you can use the below command:

library(NHSDataDictionaRy)
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
library(magrittr)
library(tibble)

This brings in the functions needed to work with the package. The below sub sections will show how to use the package, as intended.

Text manipulation of the tibble

The NHSDataDictionaRy package provides a couple of Microsoft Excel convenience functions for working with text data. These are:

I will demonstrate how these can be used on the tibble extracted from the previous example in the following sub sections.

left_xl() function

To utilise the left_xl function it expects two parameters - the first is the text to work with and the second is the number of characters to left trim by:

#Grab a sub set of the data frame
df <- nhs_tibble[10,]
result <- NHSDataDictionaRy::left_xl(df$link_name, 22)
print(result)
#> [1] "ACCESSIBLE INFORMATION"
class(result)
#> [1] "character"

right_xl() function

This works the same way as the left function, but trims from the right of the text inward:

#Grab a sub set of the data frame
df <- nhs_tibble[10,]
result <- NHSDataDictionaRy::right_xl(df$link_name, 23)
print(result)
#> [1] "FORMAT CODE (SNOMED CT)"
class(result)
#> [1] "character"

mid_xl() function

This function takes a slightly different approach and expects 3 input parameter, the first being the text to trim, the second being where to start trimming and the third parameter is the termination point i.e. where to stop the trimming of the string:

#Grab a sub set of the data frame
df <- nhs_tibble[10,]
original <- df$link_name
#Original string
result <- NHSDataDictionaRy::mid_xl(df$link_name, 12, 20)
print(original); print(result)
#> [1] "ACCESSIBLE INFORMATION SPECIFIC INFORMATION FORMAT CODE (SNOMED CT)"
#> [1] "INFORMATION SPECIFIC"
class(result)
#> [1] "character"

len_xl() function

This is a simple, but useful function, as it gets the length of the string:

#Grab a sub set of the data frame
df <- nhs_tibble[10,]
#Original string
original <- df$link_name
string_length <- NHSDataDictionaRy::len_xl(original)
print(string_length)
#> [1] 67
class(string_length)
#> [1] "integer"

Working with the NHS R Data Dictionary lookup

This package provides functionality for working with the nhs_data_elements extracted from the NHS Data Dictionary website. The two main useful function to extract elements are the tableR function and the xPathTextR function. These can work with the tibble returned to extract useful lookups.

tableR function (utilising scrapeR function)

The scrapeR function is the workhorse, but the tableR wraps the results of the function in a nice tibble output. This will show you how to utilise the return tibble and to pass the function through the tableR to scrape a tibble to be utilised for lookups:

# Filter by a specific lookup required
reduced_tibble <- 
  dplyr::filter(nhs_tibble, link_name == "ACTIVITY TREATMENT FUNCTION CODE")

#Use the tableR function to query the NHS Data Dictionary website and return the associate tibble

national_codes <- NHSDataDictionaRy::tableR(url=reduced_tibble$full_url,
                          xpath = reduced_tibble$xpath_nat_code, 
                          title = "NHS Hospital Activity Treatment Function National Codes")

default_codes <- NHSDataDictionaRy::tableR(url=reduced_tibble$full_url,
                          xpath = reduced_tibble$xpath_default_code, 
                          title = "NHS Hospital Activity Treatment Function Default Codes")


# Here you could merge the codes - as you will have national and default codes

merged_frame <- national_codes %>% 
  dplyr::bind_rows(default_codes)


# The query has returned results, if the url does not have a lookup table an error will be thrown

print(head(national_codes,10))
#> # A tibble: 10 x 4
#>    Code  Description              Dict_Type                  DttmExtracted      
#>    <chr> <chr>                    <chr>                      <dttm>             
#>  1 100   General Surgery Service  NHS Hospital Activity Tre~ 2021-02-17 14:16:44
#>  2 101   Urology Service          NHS Hospital Activity Tre~ 2021-02-17 14:16:44
#>  3 102   Transplant Surgery Serv~ NHS Hospital Activity Tre~ 2021-02-17 14:16:44
#>  4 103   Breast Surgery Service   NHS Hospital Activity Tre~ 2021-02-17 14:16:44
#>  5 104   Colorectal Surgery Serv~ NHS Hospital Activity Tre~ 2021-02-17 14:16:44
#>  6 105   Hepatobiliary and Pancr~ NHS Hospital Activity Tre~ 2021-02-17 14:16:44
#>  7 106   Upper Gastrointestinal ~ NHS Hospital Activity Tre~ 2021-02-17 14:16:44
#>  8 107   Vascular Surgery Service NHS Hospital Activity Tre~ 2021-02-17 14:16:44
#>  9 108   Spinal Surgery Service   NHS Hospital Activity Tre~ 2021-02-17 14:16:44
#> 10 109   Bariatric Surgery Servi~ NHS Hospital Activity Tre~ 2021-02-17 14:16:44
print(head(default_codes), 10)
#> # A tibble: 2 x 4
#>   Code  Description                   Dict_Type              DttmExtracted      
#>   <chr> <chr>                         <chr>                  <dttm>             
#> 1 199   Non-UK provider; TREATMENT F~ NHS Hospital Activity~ 2021-02-17 14:16:44
#> 2 499   Non-UK provider; TREATMENT F~ NHS Hospital Activity~ 2021-02-17 14:16:44
print(head(merged_frame))
#> # A tibble: 6 x 4
#>   Code  Description              Dict_Type                   DttmExtracted      
#>   <chr> <chr>                    <chr>                       <dttm>             
#> 1 100   General Surgery Service  NHS Hospital Activity Trea~ 2021-02-17 14:16:44
#> 2 101   Urology Service          NHS Hospital Activity Trea~ 2021-02-17 14:16:44
#> 3 102   Transplant Surgery Serv~ NHS Hospital Activity Trea~ 2021-02-17 14:16:44
#> 4 103   Breast Surgery Service   NHS Hospital Activity Trea~ 2021-02-17 14:16:44
#> 5 104   Colorectal Surgery Serv~ NHS Hospital Activity Trea~ 2021-02-17 14:16:44
#> 6 105   Hepatobiliary and Pancr~ NHS Hospital Activity Trea~ 2021-02-17 14:16:44

Not all lookups will have associated national code tables, if they are not returned you will receive a message saying the lookup table is not available for this NHS Data Dictionary type.

Using my lookup with NHS data

There are common lookups that are needed, and this is one such mapping between specialty code, to get the description of the specialty unit description. I will show an example with a made up data frame to illustrate the use case for these lookups and to have up to date lookups:


act_aggregations <- tibble(SpecCode = as.character(c(101,102,103, 104, 105)),
                             ActivityCounts = round(rnorm(5,250,3),0), 
                             Month = rep("May", 5))

# Use dplyr to join the NHS activity by specialty code

act_aggregations %>% 
  left_join(merged_frame, by = c("SpecCode"="Code"))
#> # A tibble: 5 x 6
#>   SpecCode ActivityCounts Month Description    Dict_Type     DttmExtracted      
#>   <chr>             <dbl> <chr> <chr>          <chr>         <dttm>             
#> 1 101                 247 May   Urology Servi~ NHS Hospital~ 2021-02-17 14:16:44
#> 2 102                 252 May   Transplant Su~ NHS Hospital~ 2021-02-17 14:16:44
#> 3 103                 243 May   Breast Surger~ NHS Hospital~ 2021-02-17 14:16:44
#> 4 104                 251 May   Colorectal Su~ NHS Hospital~ 2021-02-17 14:16:44
#> 5 105                 252 May   Hepatobiliary~ NHS Hospital~ 2021-02-17 14:16:44
  
# This easily joins the lookup on to your data

The benefit of having it in an R package is that you can instantaneously have a lookup of the most relevant and up to date NHS lookups, replacing the need to have a massive data warehouse to capture this information.

xpathTextR function

This function has been provided to return elements from a website, other than html tables, as these functions predominately work with tables. The below example shows how this can be implemented, but requires the retrieval of the xpath via the Inspect command in Google Chrome (CTRL + SHIFT + I):


url <- "https://datadictionary.nhs.uk/data_elements/abbreviated_mental_test_score.html"
xpath_element <- '//*[@id="element_abbreviated_mental_test_score.description"]'

# Run the xpathTextR function to retrieve details of the element retrieved

result_list <- NHSDataDictionaRy::xpathTextR(url, xpath_element)
print(result_list)
#> $result
#> [1] "Description\n  \n  \n  \n    \n      ABBREVIATED MENTAL TEST SCORE\n is the \n              PERSON SCORE\n where the \n              ASSESSMENT TOOL TYPE\n is \n              'Abbreviated Mental Test Score'.        \n    The score is in the range 0 to 10.\n  \n\n"
#> 
#> $website_passed
#> [1] "https://datadictionary.nhs.uk/data_elements/abbreviated_mental_test_score.html"
#> 
#> $xpath_passed
#> [1] "//*[@id=\"element_abbreviated_mental_test_score.description\"]"
#> 
#> $html_node_result
#> {html_document}
#> <html xmlns="http://www.w3.org/1999/xhtml" xmlns:whc="http://www.oxygenxml.com/webhelp/components" xml:lang="en" lang="en" whc:version="21.1">
#> [1] <head>\n<link rel="shortcut icon" href="../oxygen-webhelp%5Ctemplate%5Cre ...
#> [2] <body class="wh_topic_page frmBody">\n        <a href="#wh_topic_body" cl ...
#> 
#> $datetime_access
#> [1] "2021-02-17 14:16:44 GMT"
#> 
#> $person_accessed
#> [1] "GARYH - LAPTOP-GE3S96EI"

This provides details of the result, the text retrieved live from the website - this would need some cleaning, the website passed to the function, the xpath included, the result of the node search, the date and time the list was generated and the person and domain accessing this.

Cleaning the text example

The example below shows how the text could be cleaned once it is retrieved:

# Use the returned result and do some text processing
clean_text <- trimws(unlist(result_list$result))
clean_text <- clean_text %>% 
  gsub("[\r\n]", "", .) %>% #Remove new line and breaks
  trimws() %>% #Get rid of any white space
  as.character() #Cast to a character vector

print(clean_text)
#> [1] "Description                ABBREVIATED MENTAL TEST SCORE is the               PERSON SCORE where the               ASSESSMENT TOOL TYPE is               'Abbreviated Mental Test Score'.            The score is in the range 0 to 10."

I have used the trim white space function to extract the result element from the returned list from the previous function and now I use piping to a gsub function to remove newlines and spaces, I use the trimws() command again to make sure the spacing is sorted and then I convert (cast) this into a character string. Finally, the results are printed.

Wrapping up

There are lots of use cases for this, but I would like to keep iterating this tool so please contact me with suggestions of what could be included in future versions.