Characterization is not supporting temporal covariate settings object

[gowthamrao]

Reproducible code. the code below gives the error
Error: Columns time_id not found

While uploading covariates

This is because timeId is absent in these tables

Characterization/inst/sql/sql_server/ResultTables.sql

Lines 99 to 123 in fa8c856

	CREATE TABLE @my_schema.@table_prefixcovariates (
	run_id int NOT NULL,
	database_id varchar(100) NOT NULL,
	cohort_definition_id int NOT NULL,
	covariate_id int NOT NULL,
	sum_value int NOT NULL,
	average_value float NOT NULL
	);
	CREATE TABLE @my_schema.@table_prefixcovariates_continuous (
	run_id int NOT NULL,
	database_id varchar(100) NOT NULL,
	cohort_definition_id int NOT NULL,
	covariate_id int NOT NULL,
	count_value int NOT NULL,
	min_value float,
	max_value float,
	average_value float,
	standard_deviation float,
	median_value float,
	p_10_value float,
	p_25_value float,
	p_75_value float,
	p_90_value float
	);

# Set up
remotes::install_github("OHDSI/DatabaseConnector")
remotes::install_github("OHDSI/FeatureExtraction")

outputLocation <- file.path("D:", "test")

# cohort definition set
targetCohortIds <- 40 # diabetes mellitus
outcomeCohortIds <- 23 # inpatient
cohortDefinitionSet <-
  PhenotypeLibrary::getPlCohortDefinitionSet(cohortIds = c(targetCohortIds, outcomeCohortIds))

database <- "eunomia"
resultsLocation <- file.path(outputLocation, database)
cohortDatabaseSchema <- "main"
cdmDatabaseSchema <- "main"
databaseId <- database
cohortTableNames <-
  CohortGenerator::getCohortTableNames(cohortTable = "cohort")
targetCohortDatabaseSchema <- cdmDatabaseSchema
targetCohortTable <- cohortTableNames$cohortTable
outcomeCohortDatabaseSchema <- cdmDatabaseSchema
outcomeCohortTable <- cohortTableNames$cohortTable
connectionDetails <- Eunomia::getEunomiaConnectionDetails()
CohortGenerator::createCohortTables(
  connectionDetails = connectionDetails,
  cohortDatabaseSchema = cohortDatabaseSchema,
  cohortTableNames = cohortTableNames
)
CohortGenerator::generateCohortSet(
  connectionDetails = connectionDetails,
  cdmDatabaseSchema = cdmDatabaseSchema,
  cohortDatabaseSchema = cohortDatabaseSchema,
  cohortTableNames = cohortTableNames,
  cohortDefinitionSet = cohortDefinitionSet
)

# feature extraction specification from CohortDiagnostics as it has temporalCovariateSettings
cohortDiagnosticsCovariateSettings <-
  CohortDiagnostics::getDefaultCovariateSettings()


# Setting for characterization package
aggregateCovariateSettings1 <-
  Characterization::createAggregateCovariateSettings(
    targetIds = targetCohortIds,
    outcomeIds = outcomeCohortIds,
    riskWindowStart = 0,
    startAnchor = 'cohort start',
    riskWindowEnd = 0,
    endAnchor = 'cohort end',
    covariateSettings = cohortDiagnosticsCovariateSettings
  )

timeToEventSettings1 <- Characterization::createTimeToEventSettings(targetIds = targetCohortIds,
                                                                    outcomeIds = outcomeCohortIds)
dechallengeRechallengeSettings <-
  Characterization::createDechallengeRechallengeSettings(
    targetIds = targetCohortIds,
    outcomeIds = outcomeCohortIds,
    dechallengeStopInterval = 30,
    dechallengeEvaluationWindow = 31
  )


characterizationSettings <-
  Characterization::createCharacterizationSettings(
    timeToEventSettings = list(timeToEventSettings1),
    dechallengeRechallengeSettings = list(dechallengeRechallengeSettings),
    aggregateCovariateSettings = list(aggregateCovariateSettings1)
  )

Characterization::runCharacterizationAnalyses(
  connectionDetails = connectionDetails,
  cdmDatabaseSchema = cdmDatabaseSchema,
  targetDatabaseSchema = targetCohortDatabaseSchema,
  targetTable = targetCohortTable,
  outcomeDatabaseSchema = outcomeCohortDatabaseSchema,
  outcomeTable = outcomeCohortTable,
  characterizationSettings = characterizationSettings,
  saveDirectory = resultsLocation,
  tablePrefix = 'c_',
  databaseId = databaseId
)

[jreps]

This edit will require adding migration SQL scrips to add in the time reference. I'll look into adding those.

[jreps]

Actually - enabling temporal features would make things difficult, especially for the shiny viewer (you would have to check for timeId and do a different view). I think the best solution would be to add a new function that does temporal description where the user specifies cohortIds and a temporal features setting. I'll add that in the future.