Generate a candidate codelist
In CodelistGenerator: Identify Relevant Clinical Codes and Evaluate Their Use

knitr::opts_chunk$set(
  collapse = TRUE,
  comment = "#>",
  out.width = "100%"
)

In this example we will create a candidate codelist for osteoarthritis, exploring how different search strategies may impact our final codelist. First, let's load the necessary packages and create a cdm reference using mock data.

library(dplyr)
library(CodelistGenerator)

cdm <- mockVocabRef()

The mock data has the following hypothetical concepts and relationships:

knitr::include_graphics("Figures/1.png")

Search for keyword match

We will start by creating a codelist with keywords match. Let's say that we want to find those codes that contain "Musculoskeletal disorder" in their concept_name:

knitr::include_graphics("Figures/2.png")

getCandidateCodes(
  cdm = cdm,
  keywords = "Musculoskeletal disorder",
  domains = "Condition", 
  standardConcept = "Standard",
  includeDescendants = FALSE,
  searchInSynonyms = FALSE,
  searchNonStandard = FALSE,
  includeAncestor = FALSE
)

Note that we could also identify it based on a partial match or based on all combinations match.

getCandidateCodes(
  cdm = cdm,
  keywords = "Musculoskeletal",
  domains = "Condition",
  standardConcept = "Standard",
  searchInSynonyms = FALSE,
  searchNonStandard = FALSE,
  includeDescendants = FALSE,
  includeAncestor = FALSE
)

getCandidateCodes(
  cdm = cdm,
  keywords = "Disorder musculoskeletal",
  domains = "Condition",
  standardConcept = "Standard",
  searchInSynonyms = FALSE,
  searchNonStandard = FALSE,
  includeDescendants = FALSE,
  includeAncestor = FALSE
)

Notice that currently we are only looking for concepts with domain = "Condition". However, we can expand the search to all domains using domain = NULL.

getCandidateCodes() function will generate a table with class "candidate_codes", which contains an atribute with the details of the search strategy:

candidate_codes <- getCandidateCodes(
  cdm = cdm,
  keywords = "Musculoskeletal",
  domains = "Condition",
  standardConcept = "Standard",
  searchInSynonyms = FALSE,
  searchNonStandard = FALSE,
  includeDescendants = FALSE,
  includeAncestor = FALSE
)

searchStrategy(candidate_codes)

Include non-standard concepts

Now we will include standard and non-standard concepts in our initial search. By setting standardConcept = c("Non-standard", "Standard"), we allow the function to return, in the final candidate codelist, both the non-standard and standard codes that have been found.

knitr::include_graphics("Figures/3.png")

getCandidateCodes(
  cdm = cdm,
  keywords = "Musculoskeletal disorder",
  domains = "Condition",
  standardConcept = c("Non-standard", "Standard"),
  searchInSynonyms = FALSE,
  searchNonStandard = FALSE,
  includeDescendants = FALSE,
  includeAncestor = FALSE
)

Multiple search terms

We can also search for multiple keywords simultaneously, capturing all of them with the following search:

knitr::include_graphics("Figures/4.png")

getCandidateCodes(
  cdm = cdm,
  keywords = c(
    "Musculoskeletal disorder",
    "arthritis"
  ),
  domains = "Condition",
  standardConcept = c("Standard"),
  includeDescendants = FALSE,
  searchInSynonyms = FALSE,
  searchNonStandard = FALSE,
  includeAncestor = FALSE
)

Add descendants

Now we will include the descendants of an identified code using includeDescendants argument

knitr::include_graphics("Figures/5.png")

getCandidateCodes(
  cdm = cdm,
  keywords = "Musculoskeletal disorder",
  domains = "Condition",
  standardConcept = "Standard",
  includeDescendants = TRUE,
  searchInSynonyms = FALSE,
  searchNonStandard = FALSE,
  includeAncestor = FALSE
)

Notice that now, in the column found_from, we can see that we have obtain concept_id=1 from an initial search, and concept_id_=c(2,3,4,5) when searching for descendants of concept_id 1.

With exclusions

We can also exclude specific keywords using the argument exclude

knitr::include_graphics("Figures/6.png")

getCandidateCodes(
  cdm = cdm,
  keywords = "Musculoskeletal disorder",
  domains = "Condition",
  exclude = c("Osteoarthrosis", "knee"),
  standardConcept = "Standard",
  includeDescendants = TRUE,
  searchInSynonyms = FALSE,
  searchNonStandard = FALSE,
  includeAncestor = FALSE
)

When multiple words are added within a term (e.g., "knee osteoarthritis"), each word will be searched independently, so that for example, "osteoarthritis of knee" is excluded:

getCandidateCodes(
  cdm = cdm,
  keywords = "Musculoskeletal disorder",
  domains = "Condition",
  exclude = c("knee osteoarthritis"),
  standardConcept = "Standard",
  includeDescendants = TRUE,
  searchInSynonyms = FALSE,
  searchNonStandard = FALSE,
  includeAncestor = FALSE
)

If we only want to exclude exact matching terms (without accounting for words boundaries) we need to add "/" at the beginning and at the end of the term. Hence, using "knee osteoarthritis", "osteoarthritis of knee" won't be excluded. However, if we had "rightknee osteoarthritis", it would be excluded.

# No exclusion:
getCandidateCodes(
    cdm = cdm,
    keywords = "Knee",
    domains = "Condition",
    exclude = NULL,
    standardConcept = c("Standard", "Non-standard"),
    includeDescendants = TRUE,
    searchInSynonyms = FALSE,
    searchNonStandard = FALSE,
    includeAncestor = FALSE
)

# Exclusion looking for terms:
getCandidateCodes(
  cdm = cdm,
  keywords = "Knee",
  domains = "Condition",
  exclude = c("knee osteoarthritis"),
  standardConcept = c("Standard", "Non-standard"),
  includeDescendants = TRUE,
  searchInSynonyms = FALSE,
  searchNonStandard = FALSE,
  includeAncestor = FALSE
)

# Exclusion looking for partial matching terms (without word boundaries)
getCandidateCodes(
  cdm = cdm,
  keywords = "Knee",
  domains = "Condition",
  exclude = c("/knee osteoarthritis/"),
  standardConcept = c("Standard", "Non-standard"),
  includeDescendants = TRUE,
  searchInSynonyms = FALSE,
  searchNonStandard = FALSE,
  includeAncestor = FALSE
)

# Exclusion looking for partial matching terms (without word boundaries)
getCandidateCodes(
  cdm = cdm,
  keywords = "Knee",
  domains = "Condition",
  exclude = c("/e osteoarthritis/"),
  standardConcept = c("Standard", "Non-standard"),
  includeDescendants = TRUE,
  searchInSynonyms = FALSE,
  searchNonStandard = FALSE,
  includeAncestor = FALSE
)

If we want to do exact matching (that means, to find the exact two words "knee osteoarthritis" in the concept name) we need to use "/\b" at the beginning and at the end of the expression.

getCandidateCodes(
  cdm = cdm,
  keywords = "Knee",
  domains = "Condition",
  exclude = c("/\bKnee osteoarthritis/\b"),
  standardConcept = c("Standard", "Non-standard"),
  includeDescendants = TRUE,
  searchInSynonyms = FALSE,
  searchNonStandard = FALSE,
  includeAncestor = FALSE
)

# We will now only search for "ee osteoarthritis" to show that 
# "knee osteoarthritis" won't be excluded:
getCandidateCodes(
  cdm = cdm,
  keywords = "Knee",
  domains = "Condition",
  exclude = c("/\bee osteoarthritis/\b"),
  standardConcept = c("Standard", "Non-standard"),
  includeDescendants = TRUE,
  searchInSynonyms = FALSE,
  searchNonStandard = FALSE,
  includeAncestor = FALSE
)

For example, if we look for 

Notice that, for example, if we wanted `keywords = "depression"` and `exclude = "ST depression"`, concepts like "poSTpartum depression" would be excluded. To avoid this,
we could use `exclude = "/ST depression/"`. Notice that, "poST depression" would also be excluded with this option.

Hence, there is another option to exclude exact matching terms accounting for words boundaries: adding "/\b" at the beginning and at the end of the term. For example, if we look for "/\bp osteoarthritis/\b", concepts like "hip osteoarthritis **won't** be excluded.



## Add ancestor
To include the ancestors one level above the identified concepts, we can use the argument `includeAncestor`
```r
knitr::include_graphics("Figures/7.png")

codes <- getCandidateCodes(
  cdm = cdm,
  keywords = "Osteoarthritis of knee",
  includeAncestor = TRUE,
  domains = "Condition",
  standardConcept = "Standard",
  includeDescendants = TRUE,
  searchInSynonyms = FALSE,
  searchNonStandard = FALSE,
)

codes

Search using synonyms

We can also pick up codes based on their synonyms. For example, Osteoarthrosis has a synonym of Arthritis.

knitr::include_graphics("Figures/8.png")

getCandidateCodes(
  cdm = cdm,
  keywords = "osteoarthrosis",
  domains = "Condition",
  searchInSynonyms = TRUE,
  standardConcept = "Standard",
  includeDescendants = FALSE,
  searchNonStandard = FALSE,
  includeAncestor = FALSE
)

Notice that if includeDescendants = TRUE, Arthritis descendants will also be included:

knitr::include_graphics("Figures/9.png")

getCandidateCodes(
  cdm = cdm,
  keywords = "osteoarthrosis",
  domains = "Condition",
  searchInSynonyms = TRUE,
  standardConcept = "Standard",
  includeDescendants = TRUE,
  searchNonStandard = FALSE,
  includeAncestor = FALSE
)

Search via non-standard

We can also pick up concepts associated with our keyword via non-standard search.

knitr::include_graphics("Figures/10.png")

codes1 <- getCandidateCodes(
  cdm = cdm,
  keywords = "Degenerative",
  domains = "Condition",
  standardConcept = "Standard",
  searchNonStandard = TRUE,
  includeDescendants = FALSE,
  searchInSynonyms = FALSE,
  includeAncestor = FALSE
)
codes1

Let's take a moment to focus on the standardConcept and searchNonStandard arguments to clarify the difference between them. standardConcept specifies whether we want only standard concepts or also include non-standard concepts in the final candidate codelist. searchNonStandard determines whether we want to search for keywords among non-standard concepts.

In the previous example, since we set standardConcept = "Standard", we retrieved the code for Osteoarthrosis from the non-standard search. However, we did not obtain the non-standard code degenerative arthropathy from the initial search. If we allow non-standard concepts in the final candidate codelist, we would retireve both codes:

knitr::include_graphics("Figures/11.png")

codes2 <- getCandidateCodes(
  cdm = cdm,
  keywords = "Degenerative",
  domains = "Condition",
  standardConcept = c("Non-standard", "Standard"),
  searchNonStandard = FALSE,
  includeDescendants = FALSE,
  searchInSynonyms = FALSE,
  includeAncestor = FALSE
)
codes2