PICO Extraction
PICO extraction uses AI to read the full text of uploaded PDFs and automatically identify the Population, Intervention, Comparator, and Outcome elements described in each study. This saves reviewers significant time compared to manually reading each paper to locate PICO information.
What PICO extraction does
When you upload a PDF for a study and trigger extraction, the AI:
- Parses the full text of the PDF into structured sections
- Identifies passages that describe the study's population (patient group or sample)
- Identifies the intervention under investigation
- Identifies the comparator (control or comparison group, if any)
- Identifies the primary and secondary outcomes measured
The extracted elements are stored against the study record and can be viewed alongside the PDF.
Prerequisites
PICO extraction requires a PDF to be uploaded for the study. If a study does not have an attached PDF, PICO extraction cannot run.
Upload PDFs for studies before triggering PICO extraction. See Full-text screening for guidance on attaching PDFs to studies.
Running PICO extraction
Open a study
Navigate to the study record that has a PDF attached.
Trigger PICO extraction
In the study details panel, click Extract PICO. The extraction job is queued.
Wait for results
Extraction runs asynchronously. A status indicator shows when the job is running. Results appear in the PICO panel once extraction completes - typically within a few minutes per document.
You can trigger PICO extraction for multiple studies at once from the study list view using bulk selection.
Viewing and editing extracted PICO
Once extraction completes, the extracted PICO elements appear in the study's detail panel. Each element shows the AI-extracted text and the original passage it was derived from, so you can verify accuracy in context.
If the AI extraction is incorrect or incomplete, you can edit the extracted values manually. Manual edits are preserved and will not be overwritten by re-running extraction unless you explicitly reset the field.
Accuracy considerations
PICO extraction accuracy depends on PDF quality and study structure. Well-formatted journal articles with clearly labeled sections (Methods, Population, Outcomes) yield the best results. Conference abstracts or poorly scanned PDFs may produce partial or inaccurate extractions.
Always review AI-extracted PICO values before using them in data synthesis. The AI provides a starting point - researcher verification is required for accuracy.