Search within specified journals and dates

Worked example Mindfulness intervention trials in two journals, 2019–2025

A focused search needs explicit limits and a record of what was screened. This example uses PubMed to retrieve mindfulness-related publications in JAMA Psychiatry and Behaviour Research and Therapy, then separates candidate randomized trial reports from other publications. It produces a complete screening table and a search note, not treatment recommendations or a systematic review.

1. Set the question and eligibility rule

Open a project and select a connected model. Confirm the PubMed Connector is available in Settings → Connectors; configure contact information there if requested. No downloaded papers are required to start this example.

Use publication dates 2019-01-01 through 2025-12-31, not the date on which a database record was added. Include primary randomized reports with a defined mindfulness intervention arm and reported participant outcomes. Keep a separate uncertain category for mixed interventions, unclear publication status or conflicting dates. Mechanistic outcomes alone do not make a report a secondary analysis.

Search PubMed for mindfulness intervention randomized trial reports
published in JAMA Psychiatry or Behaviour Research and Therapy from
2019-01-01 through 2025-12-31. Preserve the exact query, date field,
search date, total count, retrieved count and truncation status.
Inspect complete returned abstracts and metadata, not titles alone.
Keep every candidate in mindfulness-search-audit.csv with title,
journal, date, PMID, DOI, source links, included/excluded/uncertain
and a study-specific reason. Include primary randomized reports with
a defined mindfulness arm and participant outcomes. Flag mixed
interventions and unclear primary/secondary status as uncertain.
Save mindfulness-search-notes.md and the raw returned records.
Do not retrieve full text or infer clinical recommendations.
Write in English, do not delegate, and reopen the saved files.

The screenshot shows the initial request. The operational screening rule above incorporates the subsequent review of that run.

2. Check the query and coverage

The run submitted this concept-and-journal query, with the publication-date filter supplied separately:

("mindfulness"[Title/Abstract] OR "mindfulness-based"[Title/Abstract])
AND ("JAMA Psychiatry"[Journal] OR "Behaviour Research and Therapy"[Journal])

Open Notebook or expand the PubMed activity. Confirm the journal field, both dates and the returned count. On September 16, 2026, the search returned 62 records; all 62 were retrieved, with has_more = false. Counts may change as PubMed updates its index. If your result is truncated, retrieve the remaining pages before claiming that all search hits were screened.

The broad concept query deliberately retains non-trials. Trial eligibility is a separate screening decision. Publication-type indexing alone can be incomplete, and a paper mentioning a previous randomized trial is not necessarily reporting a new one.

3. Review and correct the screening table

When the response completes, open mindfulness-search-audit.csv under Generated. It should retain every retrieved PMID, including excluded and uncertain records. Check the title, journal, DOI and date against the linked PubMed record, then compare the decision with the abstract.

Review found inaccurate reasons in the initial draft: for example, PMID 38837133 is a primary randomized trial of a broader psychotherapy, not a secondary analysis. The revised table marks its mindfulness eligibility uncertain. PMID 34009273 is a meta-analysis and remains excluded with the corrected reason. Naming the record and the specific error gives the Agent a concrete revision request; then reopen the saved CSV to confirm the change.

The reviewed example table contains 20 included, 37 excluded and 5 uncertain records, accounting for all 62 hits. These are abstract-level screening decisions, not a declaration that 20 distinct trials have been fully appraised. Multiple publications can concern the same underlying trial.

4. Keep uncertainty visible

PMID 41418645 was returned by the 2019–2025 PubMed publication-date filter, while its metadata reports a 2026-01 print date. Keep the discrepancy and verify the publication history before making a final date decision. Do not silently replace the year to fit the search window.

Open mindfulness-search-notes.md and verify that its counts, eligibility rule and limitations match the CSV. This run revised both files and retained the unchanged raw metadata snapshot.

Download the reviewed candidate table and search note. Full abstracts are not redistributed here; follow the source links to inspect them.

For a formal evidence review, resolve the uncertain records, obtain full texts, link companion reports to their trials and arrange appropriate independent screening. Use the reading-list workflow to build a collection, or evidence extraction after the source set and access are established.