AI-Powered Screening

Transform the most time-consuming phase of systematic review with intelligent screening that learns from your decisions.

The Screening Challenge

Traditional screening means:

Reading 8,000+ titles and abstracts
Hours of tedious manual work
High risk of human error
Inconsistent decisions

EvidAI's Intelligent Approach

Multi-Agent Consensus

Our advanced AI system uses multiple perspectives to ensure accuracy:

Different analytical approaches evaluate each study
Consensus algorithms combine assessments
Confidence scores indicate certainty
Disagreements flag for human review

Accuracy You Can Trust

96%+ sensitivity: Rarely misses relevant studies
Validated against experts: Tested on gold-standard datasets
Transparent reasoning: See why each decision was made
Human final say: You approve all critical decisions

Screening Workflow

Title/Abstract Screening

Step 1: Configure Criteria

Define your inclusion and exclusion criteria:

Population characteristics
Intervention requirements
Study design filters
Publication requirements

Step 2: AI Pre-Screening

The system automatically:

Analyzes all citations against your criteria
Assigns relevance predictions
Prioritizes likely includes
Flags uncertain cases

Step 3: Human Review

You see studies organized by:

High confidence include: Quick verification needed
High confidence exclude: Spot-check recommended
Uncertain: Requires your judgment

Step 4: Active Learning

As you make decisions:

AI learns your preferences
Predictions improve in real-time
Remaining studies re-prioritize

Full-Text Screening

For studies passing title/abstract:

PDF Retrieval: Automatic where available
Criteria Application: Detailed assessment
Exclusion Documentation: Required reason capture
Arbitration Tools: Resolve disagreements

Dual Screening Support

Blind Review Mode

Reviewers work independently
Decisions hidden until both complete
Prevents bias in assessments

Conflict Resolution

When reviewers disagree:

Side-by-side comparison view
Discussion thread for each study
Third reviewer assignment option
Consensus documentation

Inter-Rater Reliability

Automatic calculation of:

Cohen's Kappa
Percent agreement
Disagreement patterns
Performance metrics

PICO Highlighting

Our intelligent text analysis:

Highlights relevant PICO elements in abstracts
Color-codes by category (Population, Intervention, etc.)
Shows AI confidence for each highlight
Helps you scan abstracts faster

Productivity Features

Keyboard Shortcuts

I - Include study
E - Exclude study
M - Maybe (uncertain)
→ - Next study
← - Previous study

Batch Operations

Include/exclude multiple studies
Apply tags in bulk
Export selections

Best Practices

For Accuracy

Pilot test your criteria on 50 studies
Review AI uncertain cases carefully
Check a sample of AI confident excludes
Document all inclusion decisions

For Efficiency

Screen in 30-minute focused sessions
Use keyboard shortcuts
Trust high-confidence AI predictions
Focus human attention where it matters