README.md

In rpcdsearch: Tools for the Construction of Clinical Code Lists for Primary Care Database Studies

rpcdsearch

Identifies relevant clinical codes and automates the construction of clinical code lists

David A. Springate, Evangelos Kontopantelis, Ivan Olier.

Development of this package has been frozen. This package has been merged with the rEHR package. See the rEHR codelists vignette for details and see the Introduction to rEHR vignette for more details on this package. Further updates will be made there.

Clinical code search and build methodology will be published in a forthcoming paper.

rpcdsearch is not on CRAN but you can install from github using devtools:

install.packages("devtools")
require(devtools)
install_github("rEHR", "rpcdsearch")
require(rpcdsearch)

Building draft definition lists

Definition lists can be defined for:

• clinical terms (by either text search or searching for matching clinical codes)
• test terms (by text search)
• medications (by either text search or product code)

Building definition lists is a two stage process:

1. The search is defined by instantiating an object of class MedicalDefinition, containing the terms to be searched for in the lookup tables
2. A definition_search is performed on the MedicalDefinition object and the relevant lookup tables to return a list of matching dataframes

A MedicalDefinition object can be either made using terms defined within R or with terms imported from an external csv file

Defining searches within R

Use the MedicalDefinition constructor function to generate search definitions. This takes the following arguments:

• terms a list of character vectors representing clinical search terms or NULL
• codes list of character vectors representing clinical code terms or NULL
• tests list of character vectors representing test search terms or NULL
• drugs list of character vectors representing drug search terms or NULL
• drugcodes list of character vectors representing drug product code terms or NULL

vectors of length > 1 are searched for together (AND), in any order. Different vectors in the same list are searched for seperately (OR). Placing a "-" character at the start of a character vector element excludes that terms from the search.

# vectors of length > 1 are combined as a single AND expression
# "-" excludes that term from the search
def <- MedicalDefinition(terms = list("peripheral vascular disease", "peripheral gangrene", "-wrong answer",
                      "intermittent claudication", "thromboangiitis obliterans",
                      "thromboangiitis obliterans", "diabetic peripheral angiopathy",
                      c("diabetes", "peripheral angiopathy"),  # combined as a single AND expression
                      c("diabetes", "peripheral angiopathy"),
                      c("buerger",  "disease presenile_gangrene"),
                      "thromboangiitis obliterans",
                      "-rubbish", # exclusion
                      c("percutaneous_transluminal_angioplasty", "artery"),
                      c("bypass", "iliac_artery"),
                      c("bypass", "femoral_artery"),
                      c("femoral_artery" , "occlusion"),
                      c("popliteal_artery", "occlusion"),
                      "dissecting_aortic_aneurysm", "peripheral_angiopathic_disease",
                      "acrocyanosis", "acroparaesthesia", "erythrocyanosis",
                      "erythromelalgia", "ABPI",
                      c("ankle", "brachial"),
                      c("ankle", "pressure"),
                      c("left", "brachial"),
                      c("left", "pressure"),
                      c("right", "brachial"),
                      c("right", "pressure")),
         codes = list("G73"),
         tests = NULL,
         drugs = list("insulin", "diabet", "aspirin"))

When searching for codes, a range of clinical codes can be searched for by providing two codes seperated by a hyphen. e.g. E114-E117z.

importing searches via a csv file

Searches can be imported from a csv file in this format

The first column in every row determines the list that the term applies to and the second column determines whether the term should be included or excluded. Note that the csv does not have to be a valid format for conversion to a dataframe. Extra columns can be used to include terms to be combined as an AND expression with the other terms on that row. The title row can also be ommitted. You can use standard regex escape patterns in the term definitions.

The data is called into R in the following way:

## Using the example search definition provided with the package
def2 <- import_definitions(system.file("extdata", "example_search.csv", package = "rpcdsearch"))

Running searches

Once a search has been defined, the relevant lookup tables should be called in. Note that these lookup tables are not provided with the package and will be specific to the users EHR database. These examples are using CPRD lookups and EHR definitions (See the ehr_system code for details of how the interface with CPRD is implemented).

## Use fileEncoding="latin1" to avoid any issues with non-ascii characters
medical_table <- read.delim("Lookups//medical.txt", fileEncoding="latin1", stringsAsFactors = FALSE)
drug_table <- read.delim("Lookups/product.txt", fileEncoding="latin1", stringsAsFactors = FALSE)

And the search can be run:

draft_lists <- build_definition_lists(def, medical_table = medical_table,drug_table = drug_table)

This returns a list of dataframes for each of the provided search lists. If terms and codes are provided in the definition, it also contains a combined_terms_codes data frame which is a combination of terms and codes with duplicate rows removed.

Exporting code lists

The code lists produced by build_definition_lists will often want to be reviewed by clinicians or non-technical researchers. To facilitate this, there is an export_definition_search function to export the code lists as an Excel file, with each list occupying a tab in the file. To export a code list:

out_file <- "def_searches.xlsx"
export_definition_search(draft_lists, out_file)

rpcdsearch documentation built on May 2, 2019, 7 a.m.