For this example we are going to generate a candidate codelist for osteoarthritis, looking at the impact of alternative search strategies.
library(DBI)
library(RPostgres)# postgres database connection details
serverDbi <- Sys.getenv("server")
user <- Sys.getenv("user")
password <- Sys.getenv("password")
port <- Sys.getenv("port")
host <- Sys.getenv("host")
db <- dbConnect(RPostgres::Postgres(),
dbname = serverDbi,
port = port,
host = host,
user = user,
password = password
)
# name of vocabulary schema
vocabularyDatabaseSchema <- "vocabulary"
# create cdm reference
cdm <- CDMConnector::cdm_from_con(
con = db,
cdm_schema = vocabularyDatabaseSchema,
cdm_tables = tidyselect::all_of(c(
"concept",
"concept_relationship",
"concept_ancestor",
"concept_synonym",
"drug_strength",
"vocabulary"
))
)To start we will search for “osteoarthritis”, while excluding “post-infection” and “post-traumatic”, but without searching synonyms, without searching via non-standard codes, without fuzzy matching, and without including descendants or the direct ancestor of the included concepts.
oaCodes1 <- getCandidateCodes(
cdm = cdm,
keywords = "osteoarthritis",
domains = "Condition",
searchInSynonyms = FALSE,
searchViaSynonyms = FALSE,
searchNonStandard = FALSE,
fuzzyMatch = FALSE,
exclude = c(
"post-infection",
"post-traumatic"
),
includeDescendants = FALSE,
includeAncestor = FALSE
)What is the candidate codelist?
datatable(oaCodes1,
rownames = FALSE,
options = list(
pageLength = 10,
lengthMenu = c(10, 20, 250)
)
)Now we will also include the descendants of included concepts.
oaCodes2 <- getCandidateCodes(
cdm = cdm,
keywords = "osteoarthritis",
domains = "Condition",
searchInSynonyms = FALSE,
searchViaSynonyms = FALSE,
searchNonStandard = FALSE,
fuzzyMatch = FALSE,
exclude = c(
"post-infection",
"post-traumatic"
),
includeDescendants = TRUE,
includeAncestor = FALSE
)What new codes do we pick up?
newCodes1To2 <- compareCodelists(oaCodes1, oaCodes2) %>%
filter(codelist == "Only codelist 2") %>%
select(-"codelist")
datatable(newCodes1To2,
rownames = FALSE,
options = list(
pageLength = 10,
lengthMenu = c(10, 20, 50)
)
)Now we will search the observation domain as well as the condition domain.
oaCodes3 <- getCandidateCodes(
cdm = cdm,
keywords = "osteoarthritis",
domains = c("Condition", "Observation"),
searchInSynonyms = FALSE,
searchViaSynonyms = FALSE,
searchNonStandard = FALSE,
fuzzyMatch = FALSE,
exclude = c(
"post-infection",
"post-traumatic"
),
includeDescendants = FALSE,
includeAncestor = FALSE
)What new codes do we pick up?
newCodes1To3 <- compareCodelists(oaCodes1, oaCodes3) %>%
filter(codelist == "Only codelist 2") %>%
select(-"codelist")
datatable(newCodes1To3,
rownames = FALSE,
options = list(
pageLength = 10,
lengthMenu = c(10, 20, 50)
)
)Now we will search the concept synonym table to identify concepts to include.
oaCodes4 <- getCandidateCodes(
cdm = cdm,
keywords = "osteoarthritis",
domains = "Condition",
searchInSynonyms = TRUE,
searchViaSynonyms = TRUE,
searchNonStandard = FALSE,
fuzzyMatch = FALSE,
exclude = c(
"post-infection",
"post-traumatic"
),
includeDescendants = FALSE,
includeAncestor = FALSE
)What new codes do we pick up?
newCodes1To4 <- compareCodelists(oaCodes1, oaCodes4) %>%
filter(codelist == "Only codelist 2") %>%
select(-"codelist")
datatable(newCodes1To4,
rownames = FALSE,
options = list(
pageLength = 10,
lengthMenu = c(10, 20, 50)
)
)Now we will search the concept synonym table to identify concepts to include.
oaCodes5 <- getCandidateCodes(
cdm = cdm,
keywords = "osteoarthritis",
domains = "Condition",
searchInSynonyms = FALSE,
searchViaSynonyms = FALSE,
searchNonStandard = TRUE,
fuzzyMatch = FALSE,
exclude = c(
"post-infection",
"post-traumatic"
),
includeDescendants = FALSE,
includeAncestor = FALSE
)What new codes do we pick up?
newCodes1To5 <- compareCodelists(oaCodes1, oaCodes5) %>%
filter(codelist == "Only codelist 2") %>%
select(-"codelist")
datatable(newCodes1To5,
rownames = FALSE,
options = list(
pageLength = 10,
lengthMenu = c(10, 20, 50)
)
)Now we will use an approximate search. See https://www.rdocumentation.org/packages/base/versions/3.6.2/topics/agrep for further details (as this is what is running under the hood). Note, using the defaults of 0.1 for for substitutions, deletions, and insertions.
oaCodes6 <- getCandidateCodes(
cdm = cdm,
keywords = "osteoarthritis",
domains = "Condition",
searchInSynonyms = FALSE,
searchViaSynonyms = FALSE,
searchNonStandard = FALSE,
fuzzyMatch = TRUE,
maxDistanceCost = 0.1,
exclude = c(
"post-infection",
"post-traumatic"
),
includeDescendants = FALSE,
includeAncestor = FALSE
)What new codes do we pick up?
newCodes1To6 <- compareCodelists(oaCodes1, oaCodes6) %>%
filter(codelist == "Only codelist 2") %>%
select(-"codelist")
datatable(newCodes1To6,
rownames = FALSE,
options = list(
pageLength = 10,
lengthMenu = c(10, 20, 50)
)
)Here we widen the approximate searching, increasing the maximum distance allowed for a match to 0.2 for substitutions, deletions, and insertions.
oaCodes7 <- getCandidateCodes(
cdm = cdm,
keywords = "osteoarthritis",
domains = "Condition",
searchInSynonyms = FALSE,
searchViaSynonyms = FALSE,
searchNonStandard = FALSE,
fuzzyMatch = TRUE,
maxDistanceCost = 0.2,
exclude = c(
"post-infection",
"post-traumatic"
),
includeDescendants = FALSE,
includeAncestor = FALSE
)What new codes do we pick up?
newCodes1To7 <- compareCodelists(oaCodes1, oaCodes7) %>%
filter(codelist == "Only codelist 2") %>%
select(-"codelist")
datatable(newCodes1To7,
rownames = FALSE,
options = list(
pageLength = 10,
lengthMenu = c(10, 20, 50)
)
)Now we include the direct ancestor of included terms.
oaCodes8 <- getCandidateCodes(
cdm = cdm,
keywords = "osteoarthritis",
domains = "Condition",
searchInSynonyms = FALSE,
searchViaSynonyms = FALSE,
searchNonStandard = FALSE,
fuzzyMatch = FALSE,
maxDistanceCost = 0.1,
exclude = c(
"post-infection",
"post-traumatic"
),
includeDescendants = FALSE,
includeAncestor = TRUE
)What new codes do we pick up?
newCodes1To8 <- compareCodelists(oaCodes1, oaCodes8) %>%
filter(codelist == "Only codelist 2") %>%
select(-"codelist")
datatable(newCodes1To8,
rownames = FALSE,
options = list(
pageLength = 10,
lengthMenu = c(10, 20, 50)
)
)