Execution

Executing the study

Now as we have initiated database connection and created the targetTable as well as the cohortTable we are ready to initiate the execution function for the study.

################################################################################
#
# Execute
#
################################################################################

data = CohortContrast::CohortContrast(
  cdm,
  targetTable = targetTable,
  controlTable = controlTable,
  pathToResults = getwd(),
  domainsIncluded = c(
    "Drug",
    "Condition",
    "Measurement",
    "Observation",
    "Procedure",
    "Visit",
    "Visit detail"
  ),
  prevalenceCutOff = 2.5,
  topK = FALSE, # Number of features to export
  presenceFilter = 0.2, # 0-1, percentage of people who must have the chosen feature present
  complementaryMappingTable = FALSE, # A table for manual concept_id and concept_name mapping (merge)
  getSourceData = FALSE, # If true will generate summaries with source data as well
  runZTests = TRUE,
  runLogitTests = FALSE,
  createOutputFiles = TRUE,
  safeRun = FALSE,
  complName = "CohortContrastStudy")

The parameters

There are multiple parameters we can tweak for different outcomes:

Mandatory:

cdm Connection to the database

targetTable Table for target cohort

controlTable Table for control cohort

pathToResults Path to the results folder, can be project’s working directory

domainsIncluded list of CDM domains to include, choose from Drug, Condition, Measurement, Observation, Procedure, Visit, Visit detail

complName Name of the output file

Customization:

runZTests boolean for running Z-tests, tests is ran between the prevalence metrics

runLogitTests boolean for logit-tests on the prevalence, builds a model for predicting whether the patient is in target or control

runKSTests boolean for Kolmogorov-Smirnov tests on the occurrence time (vs uniform distribution)

getAllAbstractions boolean for creating abstractions’ levels for the imported data, this is useful when using GUI and exploring data

maximumAbstractionLevel Maximum level of abstraction allowed, if getAllAbstractions is TRUE, for hierarchy the concept_hierarchy table is used

getSourceData boolean for fetching source data, the data abstraction level which is used to map to OMOP CDM

lookbackDays FALSE or an integer stating the look-back period for cohort index date. This can be used inside the GUI which has a slider for adding look-back data.

prevalenceCutOff numeric or FALSE, if set, removes all of the concepts which are not present (in target) more than prevalenceCutOff times. Eg if set to 2, only concepts present double in target are exported.

topK numeric or FALSE, if set, keeps at maximum this number of features in the analysis. Maximum number of features exported.

presenceFilter numeric or FALSE, if set, removes all features represented by fewer target cohort subjects than the given percentage

complementaryMappingTable data frame or FALSE. Mappingtable for mapping concept_ids if present, columns CONCEPT_ID, CONCEPT_ID.new, CONCEPT_NAME.new,

numCores Number of cores to allocate to parallel processing, by default max number of cores - 1

createOutputFiles Boolean for creating output files, the default value is TRUE

safeRun boolean for only returning summarized data

Notes:

When using the GUI prevalenceCutOff, presenceFilter can be changed on a slider.

The effect of runZTests and runLogitTests can be toggled as a filter.

The function will output a file with complName, in this case CohortContrastStudy.rds to the pathToResults path. The function also outputs a metadata file CohortContrastStudy.csv for quicker overview inside the GUI.