cSearch

Systematic reviews, end to end - protocol through PRISMA, with every decision traceable. Two reviewers, blinded, conflicts resolved.
Start now!

[cSearch @cSearch_io]

Tell us what you are working on.

If you are starting a systematic review, or stuck in the middle of one, we would like to know which part is costing you the most time.

Signup is open and there is no waitlist. https://t.co/OLbanpZSid

[cSearch @cSearch_io]

Meta-analysis in cSearch runs on the extraction table you already filled in.

Mean differences for continuous outcomes, risk ratios for events, heterogeneity surfaced rather than buried in an appendix. No second file, no re-keying, and no drift between the numbers you extracted and the numbers you pooled.

https://t.co/JJY4RkOehS

[cSearch @cSearch_io]

A PRISMA flow diagram is not a figure you draw at the end.

It is a reconciliation. Records in, duplicates out, screened, excluded with reasons, included.

If the arithmetic does not close, something upstream was never written down.

[cSearch @cSearch_io]

Two short walkthroughs, if you are starting out.

One on creating your first project, one on finding your way around. No narration, no sign-up wall, about half a minute each.

[cSearch @cSearch_io]

Screening four hundred abstracts in a spreadsheet means four hundred chances to scroll past the wrong row.

cSearch screens one study at a time, full screen and keyboard first. You never touch the mouse and you never lose your place.

[cSearch @cSearch_io]

A systematic review that can explain itself is worth more than one that merely arrived at the right answer.

The first can be checked, corrected and built on. The second has to be taken on trust, and trust is not what the method was for.

[cSearch @cSearch_io]

Finding a second reviewer is usually treated as a social problem, so it gets solved with favours, group chats and a lot of waiting.

We treat it as a workflow problem. Communities and collaborator matching sit inside the project, next to the work the second person would actually be doing.

[cSearch @cSearch_io]

Cochrane asks for two people screening independently, with the process for resolving disagreements defined in advance.

In cSearch that is the screening mode, not a discipline you maintain: two reviewers blinded to each other, conflicts flagged as they appear, and a guided step to resolve them.

[cSearch @cSearch_io]

We pointed extraction at a network meta-analysis and it drafted nothing at all.

Every empty field came back with a reason rather than a number. Events not reported in the main text, check the study tables. Subgroup not reported, check the supplementary material.

An empty cell with a reason attached is a lead. A confident number with nothing behind it is a liability.

[cSearch @cSearch_io]

At step one cSearch will suggest exclusion criteria for your review.

Suggest. They arrive as candidates next to the ones you wrote, and none of them is applied until you say so.

The eligibility criteria of a systematic review are an author's decision. They should read like one.

[cSearch @cSearch_io]

A detail we think matters: in cSearch, drafted values are stored separately from the values you and your co-extractor entered.

They never enter the human channel, so dual-review conflict detection compares two people, not a person and a model.

[cSearch @cSearch_io]

A drafted extraction value in cSearch does nothing on its own.

You apply it, one value at a time. Then it sits there as an unsaved edit until you save, exactly like something you typed yourself.

Two deliberate acts before a number enters your review.

[cSearch @cSearch_io]

When a paper does not report something, cSearch says it did not find it.

That sounds small. It is the hard part. A model asked to fill a table will fill the table, and a plausible number in an extraction sheet is worse than an empty cell, because the empty cell is honest.

[cSearch @cSearch_io]

The practice project in cSearch is not a sandbox with invented studies in it.

It is a real published review with 57 real records already imported, and the interface says so: nothing here is simulated.

You can reset it and start again whenever you like. Nobody is marking your work.

[cSearch @cSearch_io]

Most teams can answer who screened a study and why. Only by reconstructing it afterwards: a spreadsheet, an email thread, someone's memory of a call in March.

The reconstruction is the part that takes weeks, and it is the part a reader is least able to check.

[cSearch @cSearch_io]

PRISMA item 16b asks you to cite the studies that looked eligible and were excluded, and explain why.

Not a count. The actual papers, named, with reasons.

It is the item people find last, and the one deciding whether a reader can check your judgement or must trust it.

[cSearch @cSearch_io]

screening_final_v3_FINAL.xlsx

Every review has one. The problem is not the filename.

It is that two reviewers spent a fortnight in different copies and nobody can say which decisions survived the merge.

Version confusion is not a filing problem. It is a data integrity problem.

[cSearch @cSearch_io]

"Does drug X improve outcomes?"

That question has a comparator whether you state one or not, and the one a reader will assume is no treatment, placebo, or usual care.

If that is not what you meant, the review answers a question you did not ask.

[cSearch @cSearch_io]

Signup is open, and there is no waitlist.

Create an account and start your first review. https://t.co/OLbanpZSid

[cSearch @cSearch_io]

cSearch will suggest search vocabulary, including Emtree candidate terms for Embase.

The interface labels them as suggestions and asks you to verify each one before use, because a term you did not check is a term you cannot defend when someone asks how the search was built.