Changelog

Source: NEWS.md

SelfControlledCohort 2.1.0

Changes:

  • Performance: MDRR, pre-exposure gain, event-dependent observation, and EASE diagnostics now use batched SQL reads and ParallelLogger-based parallelism, and no longer materialize full result sets in memory.
  • Study diagnostics are now configurable via Strategus through createRunSelfControlledCohortArgs() (runDiagnostics, diagnostics, diagnosticThresholds).
  • Added a CohortGenerator dependency; execute() now validates that all requested exposure/outcome cohorts were generated (present in the cohort checksum table).

Bug Fixes:

  • Fixed a bug where exposure-based negative controls (controlType = "exposure") were treated as outcomes, contaminating outcome_cohort_id in results.
  • Fixed scc_outcome_exposure to reflect the full input exposure-outcome list (rather than only pairs with results) and to correctly accumulate scc_result across all calibration groups/batches.
  • scc_result, scc_stat, scc_diagnostics_summary, and scc_outcome_exposure are now always exported (as empty files when an analysis yields no effect estimates), so the export folder and manifest stay consistent.
  • scc_outcome_exposure no longer emits exposure-outcome pairs with missing cohort IDs, which previously violated the NOT NULL primary key constraints on results upload.

SelfControlledCohort 2.0.02026-06-17

This is a major release and breaks backwards compatibility with version 1.x.x series code.

Changes:

  • Introduced a results data model definition csv file in line with other OHDSI HADES analytics packages
  • Removal of use of remotes for Eunomia
  • Creation of results data model sql for analysis in database
  • Export of large result sets from database with export managers
  • Calibration of effect estimates now computed with EmpiricalCalibration package
  • Storing risk window statistics is no longer an optional parameter
  • Removed support for CDM versions < 5.0 (and no longer includes check or parameter)
  • This release implements revised diagnostics aligned with SelfControlledCaseSeries package standards.

Breaking Changes:

  • Diagnostic names and thresholds have changed to match SCCS standards
  • Old diagnostics (counts, window_balance, cohort_stability, pre_exposure) have been removed
  • getDefaultDiagnosticThresholds() now returns different threshold names

New study Diagnostic functions

  • MDRR - Minimum Detectable Relative Risk (power analysis)
  • PRE_EXPOSURE_PROPORTION - Tests for outcomes before exposure start
  • PRE_EXPOSURE_P_VALUE - Statistical test for pre-exposure gain
  • EASE expected absolute systematic error, observed residual bias
  • EVENT_DEPENDENT_OBSERVATION - what percentage of events occur just before the observation period ends?

Changes:

  • Revised diagnostic framework to align with SCCS package
  • Added power analysis (MDRR) diagnostic
  • Added pre-exposure gain detection
  • Removed design-related diagnostics (window balance, cohort stability)
  • Updated diagnostic thresholds to match SCCS standards
  • Improved diagnostic result formatting with values in failure messages

SelfControlledCohort 1.6.0

Changes:

  • Removed support for older versions of DatabaseConnector (< 5.0.0)
  • Deprecated usage of oracleTempSchema in favour of tempEmulationSchema
  • Allow calls to run with a DatabaseConnector connection object
  • No longer supports undocumented use of connectionDetails$conn for storing connections when calling runSelfControlledCohort
  • Added function computeSccRiskWindows to allow computation of exposure risk windows without having to run full SCC analysis
  • Storage of results (before rate ratios are computed) now in a results table.
  • Storage of time at risk results no longer merged with estimates
  • Added tests on sqlite for faster unit testing while working on development
  • Added testing on redshift
  • Use of DatabaseConnector::renderTranslateQueryApplyBatched allowing optional callbacks for working with large data sets (too large for memory)
  • Added computation of conventional p-values to results

SelfControlledCohort 1.5.1

Changes:

  • Added support for github actions and removed travis configuration
  • Added computeTarDistribution flag to runSelfControlledCohort, which computes the distribution of exposure time windows for subjects experiencing the outcome and average absolute time between exposure and outcome.
  • Added vignette for usage with some examples

Bug Fixes: - Fixed issue with DatabaseConnector v 4.0.0 connectionDetails object on osx and linux caused by usage of connectionDetails$conn for storage of a connection