Full-Text Screening

What Full-Text Screening Does

Full-text screening is the second pass of study selection. Studies that passed abstract screening are now reviewed against the full PDF — methods, results, supplementary information — so you can make a final inclusion or exclusion decision based on the complete article rather than the abstract alone.

Fulltext screening typically excludes another 50–80% of studies that passed abstract screening. The most common reasons are intervention details that didn't match the protocol, study designs that weren't clear from the abstract, and outcomes that turned out not to be reported in extractable form.

How It Starts

When abstract screening completes, the assistant transitions into the Fulltext phase and the Full-text screening tab unlocks in the screening drawer. The tab opens to a list of every study that was included or marked uncertain at the abstract stage.

Scholara automatically attempts to fetch open-access PDFs for each study. Studies whose PDFs couldn't be retrieved appear under the Flagged filter in the toolbar — those are the ones you'll need to upload manually or exclude.

Adding PDFs

Bulk Upload

The fastest way to add missing PDFs is bulk upload. Open the bulk-upload modal from the toolbar and drop in a folder of PDFs — Scholara matches each file to the corresponding study and queues unmatched files for manual assignment.

Bulk Exclude Studies Without PDFs

If you cannot retrieve a PDF for a study and don't intend to chase it, the toolbar's "Exclude flagged PDFs" action lets you exclude all studies under the Flagged filter in one click. The exclusions are recorded with a "PDF not available" reason that flows through to your PRISMA flow diagram.

Supported format is PDF only. Files must be text-based — scanned PDFs need OCR before upload, otherwise the AI cannot read the article body.

The Paper Review Sheet

Click any study row to open the paper review sheet — a full-screen view with the PDF on the left and the AI's assessment on the right. From this sheet you can review the AI decision, see exactly which criteria it relied on, override it, and add notes.

PDF Viewer

  • Zoom in or out using the zoom control above the PDF
  • Scroll through pages, or jump to a specific page number
  • When you arrive at a paper from a chat citation, the relevant text is highlighted automatically so you don't have to search for it

AI Assessment Panel

The right panel shows everything the AI used to make its decision:

  • Decision banner — Include, Exclude, or Uncertain, with a confidence percentage and a badge showing whether the decision was made by the AI or overridden by a human
  • Paper metadata — authors, journal, year, study design, sample size where available
  • Abstract — collapsible, with MeSH terms surfaced
  • Eligibility criteria assessment — each inclusion and exclusion criterion checked individually against the paper, with pass/fail and the supporting evidence
  • AI decision cards — the model's reasoning broken into discrete checks
  • Notes — your free-text record for the paper, persisted across sessions

Overriding the AI's Decision

Click the Override button at the top of the assessment panel and pick the new decision — Include, Exclude, or Uncertain — from the dropdown. The decision banner switches to a "Human Override" badge and the change is recorded immediately.

As with abstract screening, the override is also passed back to the chat as screening context on your next message — so the assistant knows about your decision when it answers questions about the study or runs subsequent steps.

You can also override from the chat directly. Ask the assistant something like "exclude Smith 2021 — the study didn't randomise allocation despite calling itself an RCT" and it will apply the override and update the table.

Choosing the Right Decision

Include

Mark the study as Include if it meets every inclusion criterion, doesn't meet any exclusion criterion, reports the outcomes you care about, and provides extractable data. The eligibility criteria assessment in the side panel makes this easy to verify — if every criterion is satisfied, the study is includable.

Exclude

Mark the study as Exclude if it fails any inclusion criterion, meets any exclusion criterion, or has fatal methodological flaws. The AI's reasoning records which criterion was the primary cause, which then populates your PRISMA diagram. If you override an inclusion to an exclusion, write a sentence in the notes panel — that's what reviewers and readers will see in your final flow diagram.

Uncertain

Mark the study as Uncertain when there's missing or ambiguous information you need to resolve before making a final call — for example, awaiting clarification from a corresponding author, needing a translation, or wanting to discuss the case with a co-reviewer. Uncertain decisions stay flagged so you can come back to them.

Common Full-Text Exclusion Reasons

  • Wrong population — participants don't match the criteria (age range, disease severity, comorbidities)
  • Wrong intervention — wrong dose, duration, or delivery method, or co-interventions that confound the comparison
  • Wrong comparator — no appropriate control group, or comparator doesn't match the protocol
  • Wrong outcomes — the paper measures or reports different outcomes than yours
  • Wrong study design — design only became clear in the methods section (often happens with observational studies labelled as trials)
  • Insufficient data — outcomes are mentioned but not reported in extractable form
  • Duplicate publication — same data already reported in another paper; keep the most complete report

Quality Tips

  • Read the methods section first — that's where most full-text exclusions originate
  • Use the eligibility criteria assessment as your map — satisfied criteria are quick to skim, the unsatisfied or partial ones are where you need to look closely
  • When you override, write a sentence in the notes panel explaining why. Future reviewers and your future self will thank you
  • If you're a team reviewer, run the screening twice (once independently, then compare). The override audit trail makes disagreements easy to surface

What Comes Next

When fulltext screening completes, the assistant transitions into the Analysis phase. From there you'll be able to extract structured data, generate characteristics tables and outcome plots, run meta-analyses, and produce the final PRISMA flow diagram — all from the same chat.

Excluding 50–80% of studies at full text is normal and expected. The first round of abstract screening is intentionally liberal; the full-text round is where you tighten.