Skip to contents

Covariate assembly using FeatureExtraction

buildMRCovariates.Rd

Leverages the OHDSI FeatureExtraction package to assemble a rich covariate object from OMOP CDM data. This object serves two purposes: (1) providing covariates for the outcome model to control for confounders and population stratification, and (2) exposing the full phenome for the instrument PheWAS diagnostic that checks for pleiotropic associations.

If ancestry principal components (PCs) are available, they are merged into the covariate matrix. Ancestry PCs are critical for controlling population stratification in genetic association studies.

Usage

buildMRCovariates(
  connectionDetails,
  cdmDatabaseSchema,
  cohortDatabaseSchema,
  cohortTable,
  outcomeCohortId,
  covariateSettings = NULL,
  ancestryPCsTable = NULL,
  ancestryPCsSchema = NULL,
  numAncestryPCs = 10
)

Arguments

connectionDetails

A DatabaseConnector::connectionDetails object.

cdmDatabaseSchema

Character. Schema containing OMOP CDM tables.

cohortDatabaseSchema

Character. Schema containing the cohort table.

cohortTable

Character. Name of the cohort table.

outcomeCohortId

Integer. Cohort definition ID for the outcome.

covariateSettings

A FeatureExtraction covariate settings object. If NULL (default), uses createDefaultMRCovariateSettings.

ancestryPCsTable

Character or NULL. Name of a table containing person_id and ancestry principal components (PC1 through PC_K). If NULL, ancestry PCs are not included.

ancestryPCsSchema

Character or NULL. Schema containing the ancestry PCs table.

numAncestryPCs

Integer. Number of ancestry PCs to include (1 through this value). Default is 10.

Value

A list with class "medusaCovariateData" containing:

covariateData

The FeatureExtraction covariate object returned by getDbCovariateData(). Its main tables are covariateData$covariates and covariateData$covariateRef.

ancestryPCs

Data frame of ancestry PCs if provided, NULL otherwise.

settings

The covariate settings object used.

Details

Assemble Covariate Matrix for Mendelian Randomization

Default covariate settings include: conditions in 365-day lookback (binary), drug exposures in 365-day lookback (binary), most recent measurement values, and demographics (age group, sex, index year). These are assembled using standard FeatureExtraction covariate setting objects.

See also

createDefaultMRCovariateSettings, buildMRCohort, runInstrumentDiagnostics

Examples

if (FALSE) { # \dontrun{
covData <- buildMRCovariates(
  connectionDetails = connDetails,
  cdmDatabaseSchema = "cdm",
  cohortDatabaseSchema = "results",
  cohortTable = "cohort",
  outcomeCohortId = 1234
)
} # }