Screening features: complete breakdown

What the Screening workspace is for

The Screening workspace moves records from imported citations to included, excluded, or full-text eligible status. It combines data review, assignments, title/abstract screening, full-text screening, conflict resolution, and PRISMA tracking.

Review data and imports

• Search/import records from supported sources and file imports.
• Use DOI and citation metadata where available.
• Review recent import batches.
• Undo or redo import batches when supported by the batch history.
• Use a 5,000-record per-source import request limit to prevent accidental or abusive oversized imports.
• Narrow overly broad search strings before importing large result sets.

Search strategy and search logs

Coventra can build suggested Boolean strings from the project PICO lens and log searches/imports. The log helps support PRISMA-S reporting by preserving source, query, filters, result counts, imported counts, search date, notes, and living-review markers where entered.

Deduplication

• Detect duplicate candidates after imports.
• Review held duplicate candidates before they enter the active screening queue.
• Use Keep left to keep the incoming import record or the left-side study.
• Use Keep right to keep the existing study or the right-side study.
• Use Keep both when two records represent distinct reports or linked publications.
• Use exact auto-resolution when one active record should remain; Coventra keeps the more complete record and deletes or discards the other side rather than merging fields.
• Use bulk dismissal for large groups of non-duplicate candidates.
• Export deduplication audit details for manuscript appendices or audit review.

Labels

Labels help organize screening work. They are best used for routing, source grouping, priority marks, special populations, or operational tags.

• Create project-specific labels.
• Apply labels to studies.
• Filter or scan records using labels where the UI exposes them.
• Avoid using labels as final eligibility decisions; use screening decisions for that.

Title/abstract decisions

• Review title, abstract, year, journal, authors, identifiers, and notes.
• Choose include, exclude, or maybe.
• Add exclusion reasons and notes where relevant.
• Use maybe for uncertainty that should be revisited.
• Use undo/delete of your own decision when a correction is needed.

Full-text decisions

Full-text screening is the second phase. It focuses on records that survived title/abstract review or otherwise need full report assessment.

• Work from eligible/full-text records.
• Use the same include/exclude/maybe decision model.
• Record full-text exclusion reasons carefully for manuscript appendices and PRISMA reporting.
• Full-text conflicts are tracked separately from title/abstract conflicts.

Screening settings

• Blind mode: hides peer decisions, notes, and decision discussions until an include/exclude disagreement opens.
• Reviewers required: sets how many independent reviewers are expected when no explicit assignments override the workflow.
• Auto-advance: helps reviewers move to the next record after a decision.
• Exact auto-resolution: can keep the more complete exact duplicate and remove the other side where configured.
• Highlight terms: helps reviewers spot protocol-relevant words.
• PICO terms: supports search strategy construction and review focus.

Assignments

• Create assignments for individual reviewers.
• Bulk assign selected records, first N records, random N records, or divide equally.
• Assign by phase: title/abstract or full-text.
• Prevent duplicate assignments unless intentionally including already assigned records.
• Reviewers can filter to work assigned to them.

Conflicts

An open conflict exists when include and exclude decisions disagree for the same study and phase. Maybe decisions show uncertainty but do not create the include/exclude conflict item by themselves.

• Conflicts include title/abstract and full-text phases.
• Decision discussions sit beside each visible judgment.
• Lead resolution stores the final include or exclude outcome without overwriting original reviewer decisions.
• Resolved conflicts are removed from open conflict counts.
• Changing or removing an underlying decision can cause the project state to be re-evaluated.

Reviewer reliability

Reliability views summarize reviewer agreement by phase when enough data are available. Agreement and kappa-style values are helpful quality signals, but they should be interpreted in context.

Active learning and screening ML

Screening ML ranks or prioritizes records. It never makes include/exclude decisions. Conflicted labels are excluded from model training, and the system reports when there are not enough non-conflicting labels to train.

PRISMA flow

The PRISMA tab uses project counts to help visualize identified, screened, eligible, included, and excluded records. Leads can enter overrides where manuscript reporting requires protocol-specific adjustments.