Screen with confidence
Screening is where most reviews stall. mapped runs blinded dual-reviewer workflows — or lets a validated AI act as your second reviewer when you screen solo — with the Classification Engine accelerating both. Rigorous methodology, RAISE-aligned, with a human always in control.
Updated June 2026
Screening
Study Selection Process
Ding, D., Zhang, J., Wu, P., Wang, Z., Shi, H., Yu, X., Wu, X., Kang, J., Hahn, J.-Y., Nam, C.-W., Doh, J.-H., Lee, B.-K., Kim, H.-Y., Huang, J., Jiang, F., Zhou, H., Chen, P., Tang, L., Jian…
Background: Coronary physiology measured by fractional flow reserve (FFR) is superior to angiography for assessing the efficacy of percutaneous coronary intervention (PCI). Yet, the clinical adoption of post-PCI FFR is limited. Murray law-based quantitative flow ratio (μQFR) may represent a promising alternative, as it can quickly compute FFR from a single angiographic view. Objectives: The authors aimed to investigate the potential role of post-PCI μQFR in predicting clinical outcomes. Methods: This was a post hoc blinded analysis of the FLAVOUR trial. Patients with angiographical…
Read moreWrong comparator: compares FFR-guided vs IVUS-guided PCI; lacks an angiography-only control arm.
Key Capabilities
Two-Phase Screening
Phase 1 is title/abstract — typically 90–95% of records are excluded here. Phase 2 is full-text against detailed eligibility, with required exclusion reasons that auto-populate the PRISMA flow diagram. mapped tracks both phases independently with separate progress dashboards and IRR calculations.
Blinded Dual-Reviewer Workflow
PRISMA and Cochrane treat independent dual review as the conservative default. mapped enforces it structurally: each reviewer sees the same records but cannot see the other's decisions until both are done. Disagreements are flagged automatically; a third reviewer or the lead resolves them with a documented, timestamped rationale.
AI as a Validated Second Reviewer
Screening solo? A validated AI can act as your independent second reviewer. It re-screens every record you exclude and flags likely false exclusions for a second look — the compensating control the 2025 Cochrane, Campbell, JBI & CEE position endorses for single-reviewer screening, which otherwise misses ~13% of relevant studies (Gartlehner et al. 2020). The AI ranks and flags; you confirm every decision. AI is never the sole reviewer, and full-text screening stays dual.
RAISE-Aligned & Human-in-Control
Every mode follows the RAISE (Responsible AI in Evidence Synthesis) recommendations and the 2025 Cochrane/Campbell/JBI/CEE position: independent validation of the AI at your recall threshold, human decisional authority on every include/exclude, and full reporting. mapped records the model version, calibration, and AI–human override rate, and auto-fills the RAISE checklist on manuscript export. See mappedresearch.com/validation.
Classification Engine — Confidence-Tiered AI
The Classification Engine pre-screens records in bulk and assigns High / Medium / Low confidence tiers. High-confidence exclusions stack at the bottom of the queue; uncertain cases surface first. The AI handles obvious exclusions (wrong species, wrong design, conference abstracts) so human attention concentrates on borderline judgement calls.
Conflict Resolution
When reviewers disagree, mapped flags the record, hides both decisions, and routes it to a designated resolver (third reviewer or lead). The resolver sees both reviewers' rationales, the record itself, and a guided reconciliation interface. Every resolution is timestamped and attributed in the audit trail.
Inter-Rater Reliability + Real-Time Progress
Cohen's kappa is computed automatically per phase as decisions accumulate — human–human in dual mode, and AI–human when AI is your second reviewer, so the override rate stays visible and reportable. Live dashboards show records screened, pending, and in conflict, plus per-reviewer throughput and projected completion dates.
Non-RCT Pre-filter
For RCT-only reviews, mapped's Non-RCT pre-filter excludes case reports, editorials, conference abstracts, and observational designs before reviewers see them. Excluded records remain auditable; the filter is reversible per record. Typically removes 15–40% of the queue before screening even starts.
Frequently asked questions
- What is mapped's Screening workflow?
- Two-phase PRISMA-compliant screening (title/abstract then full-text) in two reviewer modes — blinded dual-human review, or a single reviewer with a validated AI as the second reviewer — plus automatic conflict detection, third-reviewer resolution, Cohen's kappa IRR, the Classification Engine for confidence-tiered AI pre-screening, and an optional Non-RCT pre-filter.
- How does mapped's Screening compare to Covidence?
- Covidence is a strong, focused screening tool with a long track record. mapped covers the same dual-reviewer methodology plus seven other workflow steps in the same project — research-question validation, search, extraction, RoB, meta-analysis, GRADE, and manuscript. If you only need screening and extraction, Covidence is solid; if you want the whole pipeline in one tool, mapped wins on breadth.
- Does mapped enforce blinded dual review?
- Yes — structurally, in dual-reviewer mode. Each reviewer sees the same records but cannot see the other's decisions until both are finished. You can't accidentally un-blind; the system won't let you. Conflicts are flagged the moment they appear and routed to the configured resolver.
- Can I use AI as my second reviewer instead of a second person?
- Yes. When you screen solo, mapped's validated AI acts as your independent second reviewer — re-screening your exclusions to catch false excludes. This is the compensating control the 2025 Cochrane, Campbell, JBI & CEE position endorses for single-reviewer screening. You confirm every decision; the AI never decides alone, and full-text stays dual.
- Does mapped adhere to RAISE and the 2025 AI position statement?
- Yes. Screening follows RAISE (Responsible AI in Evidence Synthesis) and the 2025 Cochrane/Campbell/JBI/CEE position: AI validated at your recall threshold, human decisional authority on every include/exclude, and full reporting — model version, calibration, and AI–human override rate. mapped auto-fills the RAISE checklist on export. Details: mappedresearch.com/validation.
- Who is mapped's Screening for?
- Research teams running PRISMA- or Cochrane-style reviews with 1,000+ records. Solo PhD students benefit from the AI confidence tiers; teams of 3–7 benefit from blinded dual-review, conflict resolution, and the Cohen's kappa transparency that peer reviewers ask about.
- How much does Screening cost?
- The free tier includes AI-assisted screening on one active project. The Mapped Project tier (list $119/project, currently $79 launch pricing) unlocks unlimited reviewers, full Classification Engine processing, and conflict resolution. Custom Enterprise plans add unlimited projects and team members. See mappedresearch.com/pricing for current details.
- What is the Non-RCT pre-filter?
- An optional filter that excludes case reports, editorials, conference abstracts, and observational designs before reviewers see them — useful for RCT-only reviews. It's reversible per record, fully audited, and typically removes 15–40% of the screening queue. PRISMA flow numbers are populated automatically.
Comparing tools? See how mapped stacks up against Covidence on the workflow you actually run.
Mapped vs Covidence