Studies

A Study in Parthenon is a container that groups related cohort definitions, concept sets, analyses, and data sources into a single reproducible research package. Studies provide the organizational structure, provenance tracking, and export capabilities needed for real-world evidence projects --- from single-site exploratory analyses to multi-database OHDSI network studies.

---

Why Use Studies?

Individual cohort definitions and analyses are useful on their own, but a real-world evidence study requires many coordinated components:

• Multiple target, comparator, and outcome cohorts
• Shared concept sets (exposure definitions, outcome definitions, covariate definitions)
• Multiple analysis types (characterization for Table 1, incidence for background rates, estimation for causal effects)
• Execution across multiple data sources for network-level evidence
• Documentation of design choices, protocol amendments, and analysis decisions
• Reproducible export for peer review, regulatory submission, or multi-site replication

Without a study container, these components are disconnected --- you cannot reliably reconstruct which cohort versions and analysis configurations were used together for a specific research question. Studies solve this by binding everything together with metadata and execution history.

---

Creating a Study

1. Navigate to Analyses and click Studies (or navigate directly to the Studies section).
2. Click New Study.
3. Fill in the study metadata:

Field	Description	Required
Study Title	Descriptive title (e.g., "Cardiovascular Safety of GLP-1 Agonists vs. DPP-4 Inhibitors")	Yes
Abstract	Brief summary of the research question, population, and methods	No
Protocol	Full study protocol document (plain text or Markdown)	No
Status	Current lifecycle stage	Yes

4. Set the initial Study Status (default: Design).
5. Click Save.

Study Status Lifecycle

Status	Description
Design	Study is being designed. Components can be added, removed, and modified freely.
Execution	Study is actively being executed. Cohorts are being generated and analyses are running. Modifications are discouraged but allowed.
Complete	All analyses have been executed and results reviewed. The study is finalized.
Archived	Study is preserved for reproducibility but no longer actively used. Read-only.

Protocol-first design

Write your study protocol (or at least a structured abstract) before adding components. This disciplines the analysis plan and prevents post-hoc fishing for significant results. Include the study protocol in the Protocol field so it is permanently associated with the study package.

---

Adding Components

Once a study is created, open it and use the component tabs to link the building blocks of your research:

Cohorts Tab

Attach cohort definitions to the study, categorized by their role:

Role	Description	Example
Target	The primary population(s) of interest	New users of GLP-1 RA
Comparator	The comparison population(s)	New users of DPP-4i
Outcome	The clinical event(s) being studied	Myocardial infarction, stroke, heart failure
Subgroup	Subpopulation definitions for stratified analysis	Age >= 65, patients with CKD

Click + Add Cohort to search and select from existing cohort definitions. Each cohort can be tagged with its role in the study.

Concept Sets Tab

Attach concept sets used across the study. While concept sets are embedded within cohort expressions, explicitly linking them to the study provides:

• A central inventory of all concepts used in the study
• Easy comparison of concept set versions
• Export of concept sets as standalone artifacts

Analyses Tab

Attach analyses of any type:

• Characterizations --- for baseline Table 1 and covariate profiling
• Incidence Rates --- for background event rates and comparative incidence
• Treatment Pathways --- for understanding real-world treatment patterns
• Estimations --- for causal effect estimates (CohortMethod)
• SCCS --- for self-controlled analyses
• Predictions --- for risk models
• Evidence Synthesis --- for meta-analysis across data sources

Each analysis is linked to the study with its configuration preserved. Changes to the analysis configuration after linking are reflected in the study (they are references, not copies).

Data Sources Tab

Select which data sources this study will execute against. For single-site studies, select your primary CDM. For network studies, select all participating sites.

Setting	Description
Available sources	All configured data sources in the Parthenon instance
Selected sources	Sources included in this study's execution plan

---

Study Execution

Execute All

Click Execute All to queue all analyses in the study across all selected data sources. Parthenon handles execution in dependency order:

1. Cohort generation: All cohort definitions linked to the study are generated on each selected source (if not already generated).
2. Analysis execution: Analyses that depend on generated cohorts are queued after cohort generation completes.
3. Evidence synthesis: Meta-analytic analyses are queued after individual source-level analyses complete.

A progress dashboard shows the real-time status of each component:

Component	Source 1	Source 2	Source 3
Target cohort (GLP-1 RA)	Completed	Running	Pending
Comparator cohort (DPP-4i)	Completed	Completed	Pending
Outcome cohort (MI)	Completed	Running	Pending
Characterization	Completed	Pending	Pending
Incidence rates	Running	Pending	Pending
Estimation (HR)	Pending	Pending	Pending
Evidence synthesis	Pending	Pending	Pending

Selective Execution

You can also execute individual components:

• Click Generate on a specific cohort to generate it on selected sources
• Click Execute on a specific analysis to run just that analysis
• Click Rerun on a failed component to retry it

Results Aggregation

The study's Results tab aggregates outputs from all analyses:

• Characterization tables per source
• Incidence rate tables per target-outcome-source combination
• Effect estimates per source with forest plots
• Evidence synthesis meta-analytic summaries

This aggregated view is the study's "answer" --- the complete evidence package that supports the research conclusion.

---

Exporting a Study Package

Click Export Study Package to download a comprehensive ZIP archive containing:

study-package/
  README.md                      # Study metadata, abstract, protocol summary
  protocol.md                    # Full study protocol
  cohorts/
    target-glp1ra.json          # CIRCE JSON for each cohort
    comparator-dpp4i.json
    outcome-mi.json
    outcome-stroke.json
    outcome-hf.json
  concept-sets/
    glp1-ra-concepts.json       # Concept set expressions
    dpp4i-concepts.json
    mi-concepts.json
  analyses/
    characterization.json       # Analysis configuration
    incidence-rates.json
    estimation.json
    evidence-synthesis.json
  results/                       # (Optional) Aggregated results
    characterization-results.csv
    incidence-results.csv
    estimation-results.csv
  r-analysis/
    StudySpecification.json     # HADES-compatible study spec
    RunStudy.R                  # Pre-configured R script
    requirements.R              # Required R packages

OHDSI Network Study Compatibility

The exported study package follows OHDSI conventions and can be:

• Submitted to OHDSI network studies: The r-analysis/ folder contains scripts that partner sites can run on their own OMOP CDM databases without access to your data.
• Shared with collaborators: The complete package provides everything needed to understand, replicate, or extend the study.
• Archived for reproducibility: The package captures the exact state of all study components at the time of export.

Version your exports

Export the study package at key milestones (protocol lock, after each data refresh, at final analysis). Name exports with dates or version numbers (e.g., glp1-safety-study-v2.1-2026-03-05.zip) for clear provenance.

---

Collaboration Features

Team Access

Study creators can manage access:

• Owner: Full control (edit, execute, delete, manage access)
• Contributor: Can add/remove components and execute analyses
• Viewer: Read-only access to study design and results

Activity Log

The study detail page includes an activity log showing:

• When components were added or removed
• When executions were started and completed
• When the study status was changed
• Who performed each action

This audit trail supports regulatory requirements for documenting analysis decisions and timeline.

---

Best Practices

1. One study per research question: Keep studies focused on a single research question. A study comparing GLP-1 RA vs. DPP-4i safety should not also include an unrelated T2DM pathway analysis.

2. Write the protocol first: Define your research question, target/comparator/outcome cohorts, analysis plan, and data sources in the protocol before building anything. This prevents HARKing (Hypothesizing After Results are Known).

3. Use consistent naming: Name cohorts and analyses within the study using a consistent scheme that makes their role clear (e.g., prefix target cohorts with "[T]", comparators with "[C]", outcomes with "[O]").

4. Link all dependencies: Ensure every concept set used in any cohort or analysis is explicitly linked to the study. This makes the export self-contained.

5. Execute incrementally: Do not wait until everything is configured to execute. Generate cohorts early to validate counts, run characterization to check population balance, then proceed to estimation.

6. Export before publication: Always export the study package before submitting results for publication or regulatory review. The export preserves the exact state of all components.

7. Archive completed studies: Change status to Archived after publication to prevent accidental modification while preserving full access for future reference.

Regulatory traceability

For studies supporting regulatory submissions (FDA, EMA), the study package serves as the analysis audit trail. Ensure the protocol, all cohort definitions, analysis configurations, and results are captured in the export. Missing components can delay or derail regulatory review.