Remove Temporal Bias from CohortContrast Analysis

This function identifies and optionally removes concepts that may represent temporal bias in a CohortContrast analysis. It works by creating age/sex matched controls from the general population for the same time periods as the target cohort, then using a proportion test to identify concepts where the matched cohort has greater or equal prevalence compared to the target. These concepts likely represent temporal trends (e.g., seasonal effects, healthcare changes) rather than condition-specific features.

Usage

removeTemporalBias(
  data,
  cdm,
  ratio = 1,
  alpha = 0.05,
  domainsIncluded = NULL,
  removeIdentified = FALSE
)

Arguments

data

A CohortContrast result object (returned from CohortContrast function)

cdm

Connection to the database (package CDMConnector)

ratio

Matching ratio for control cohort generation (default: 1)

alpha

Significance level for the proportion test before Bonferroni correction (default: 0.05)

domainsIncluded

Domains to analyze for temporal bias (default: same as original analysis)

removeIdentified

If TRUE, automatically remove identified temporal bias concepts from the data (default: FALSE)

Value

A list containing:

temporal_bias_concepts

A data frame of concepts identified as potential temporal bias

data

The original or filtered CohortContrast data object (if removeIdentified = TRUE)

matched_control_prevalences

Prevalence data from the matched control cohort

Details

The function applies Bonferroni correction for multiple testing, adjusting the significance level by dividing alpha by the number of concepts being tested.

Examples

if (FALSE) { # \dontrun{
# Run CohortContrast analysis
data <- CohortContrast(cdm, targetTable, controlTable, pathToResults, ...)

# Identify temporal bias concepts
result <- removeTemporalBias(data, cdm, ratio = 1)

# View identified concepts
print(result$temporal_bias_concepts)

# Remove identified concepts and get filtered data
result_filtered <- removeTemporalBias(data, cdm, removeIdentified = TRUE)
filtered_data <- result_filtered$data
} # }