About

How a Prizmatic review works.

Every review moves through the same nine stages, from the first database query to the finished synthesis, and every decision along the way is logged. To see the stages run, open the live demo.

The pipeline
01

Retrieve

Several queries per database, merged and deduplicated into one pool.

PubMed · Scopus · Europe PMC · Semantic Scholar · Google Scholar

02

Classify and fork

Trials and observational studies split into separate tracks, each judged with its own instruments.

03

Screen titles

A first relevance pass, with every exclusion logged and its reason kept on the record.

04

Screen abstracts

Full eligibility on title and abstract. Borderline calls, the dashed ones, route to a human reviewer.

05

Fetch and parse full text

Full articles are retrieved and parsed. Anything unreachable is reported, never silently dropped.

06

Screen full text

Eligibility confirmed on the complete article. Exclusions here feed the PRISMA flow diagram.

07

Assess risk of bias

RoB 2 for trials, ROBINS-I for observational studies. Every judgment carries its supporting quotes.

08

Extract

One row per finding, and each figure keeps the sentence it came from.

09

Synthesise

An evidence grid and a written synthesis, with knowledge gaps computed from the empty cells.

See the pipeline run.

The demo steps through a complete review, from the first query to the finished synthesis.