Explore original research, methods papers, and technical reports on AI-assisted systematic reviews, meta-analysis automation, and evidence synthesis.
A practical guide to what is a scoping review: when it beats a systematic review, which evidence it includes, and how to plan, screen, chart and report one.
A practical guide to the rapid review: types, step by step process, published guidance, and how to weigh speed against rigour before you commit.
See how a PRISMA diagram tracks records found, screened, excluded and included, plus the templates, tools and checks that keep your counts sound.
Study screening is the process of evaluating identified records against predefined inclusion and exclusion criteria to determine which studies enter a systematic review. This phase typically consumes 40 to 60 percent of total review time and is where most methodological errors occur. Title and abstract screening removes clearly irrelevant records, while full-text screening applies detailed eligibility criteria. Dual independent screening with conflict resolution is the gold standard recommended by the Cochrane Handbook. This guide walks through each stage of the screening process, explains how to write effective inclusion criteria, and compares manual, semi-automated, and AI-assisted screening approaches.
The PRISMA flow diagram is a standardised visual summary of how studies move through a systematic review, from identification to final inclusion. This guide covers the PRISMA 2020 update, explains all four phases of the flow diagram, compares tools and templates, and addresses common mistakes. Whether you are creating your first flow diagram or updating to the 2020 standard, this resource walks through every box and decision point with practical advice.
Systematic reviews and meta-analyses represent the highest levels of evidence in evidence-based medicine, yet many researchers conflate the two approaches. A systematic review is a comprehensive, protocol-driven synthesis of evidence addressing a research question, whilst meta-analysis applies statistical techniques to combine quantitative data into a single estimate of treatment effect. Not all systematic reviews include meta-analysis, but all meta-analyses must be conducted within a systematic review framework. This guide clarifies key differences between these complementary methods and explains how they work together to provide robust evidence for clinical practice decisions.
Background: Manual title and abstract screening remains the most time-intensive phase of systematic reviews, often consuming 40–60% of total project hours. AI-assisted screening tools promise to reduce this burden, but independent validation across diverse clinical topics is limited. Objective: To evaluate the sensitivity, specificity, and workload reduction of Systematicly's AI screening module across 14 completed systematic reviews spanning cardiovascular medicine, oncology, mental health, and public health. Methods: We retrospectively applied the AI screening algorithm to 14 published systematic reviews comprising 28,417 unique records. Each record had a human-determined gold-standard inclusion decision. We measured sensitivity (recall), specificity, F1 score, and the proportion of records that could be safely excluded without manual review. Results: The AI module achieved a pooled sensitivity of 97.3% (95% CI: 95.8–98.4%) and specificity of 82.1% (95% CI: 79.6–84.3%). Median workload reduction was 63% of records safely auto-excluded. No included study was missed in 12 of 14 reviews. Conclusions: AI-assisted screening demonstrates high sensitivity suitable for integration into systematic review workflows, with substantial time savings for research teams.