Systematic Review Protocol

# CLAUDE.md — Systematic Review Protocol

## PRISMA Compliance
Follow PRISMA 2020 guidelines. Generate a PRISMA flow diagram at the end of screening.

## Database Search Strategy
- Minimum databases: PubMed, Embase, PsycINFO, Cochrane Library, Web of Science
- Export results with all fields including abstract
- Document search strings exactly as executed, with date and number of results

## Screening
- Title/abstract screen first, full-text second
- Two independent reviewers; resolve disagreements by discussion then third reviewer
- Kappa statistic for inter-rater reliability (target κ > 0.80)
- Reasons for exclusion documented at full-text stage

## Data Extraction Template
For each included study record:
- Study ID, first author, year, country
- Design (RCT, cohort, cross-sectional...)
- Population (N, age, inclusion criteria)
- Intervention and comparator
- Outcome measures and time points
- Results (means/SDs or proportions with CIs)
- Risk of bias assessment (RoB 2 or ROBINS-I)

## Meta-Analysis Decisions
- Fixed effects for homogeneous studies (I² < 25%)
- Random effects when heterogeneity is present
- Report I², τ², prediction interval