Find Clinical Research Authors and Affiliations in PubMed

Publication metadata can reveal who is active in a therapeutic area and which institutions collaborate. It can also create false confidence: common names, changing affiliations, and incomplete deposits make author identity resolution a separate problem.

Search a narrow topic with the PubMed Research Scraper, then expand the authors and affiliations arrays into a working table. Keep the PMID and publication date on every row. Count papers only after deduplicating the source records.

Treat affiliations as evidence, not identity

Normalize obvious formatting differences, but retain the original affiliation text. Use institution, topic overlap, co-authors, and publication timing to form candidate matches. Confirm important conclusions against the linked PubMed record and institutional sources.

For landscape analysis, useful measures include papers per institution, recurring co-author pairs, journal distribution, and topic-specific activity over time. Avoid ranking researchers on publication count alone. The dataset is best used to find where closer review should begin, not to automate consequential judgments about individuals.

Build an auditable affiliation table

Create one row per PMID, author position, author display name, and published affiliation. Preserve the original text before adding a normalized institution column. Record the rule or lookup used for each normalization; “Massachusetts General Hospital” and “Mass General” may be the same institution, while similarly named university systems may not be.

rows = []
for paper in papers:
    for position, author in enumerate(paper.get("authors") or [], start=1):
        rows.append({
            "pmid": paper["pmid"],
            "position": position,
            "author": author,
            "publication_date": paper.get("publication_date"),
            "source_url": paper["url"],
        })

Do not assume the author and affiliation arrays align positionally unless the source record explicitly provides that relationship. When the deposit supplies only article-level affiliation strings, label the relationship as article-level rather than assigning every institution to every author.

Resolve names conservatively

Use several signals together: exact name variants, institutional overlap, topic similarity, recurring co-authors, publication dates, and persistent identifiers when available from an authoritative source. Keep ambiguous matches separate. A false merge can turn two ordinary publication histories into one misleadingly prolific profile.

For outreach or expert discovery, verify the current role and public contact route on the institution's own site. PubMed describes authorship at publication time; it is not a current employment directory and should not be used to infer private contact details.

Report coverage alongside rankings

Show the percentage of papers with any affiliation, the share with author-linked affiliations, and the number of ambiguous names. Rank institutions only after deduplicating PMIDs, and label publication count as activity rather than quality or influence. Citation metrics, trial leadership, study design, and clinical impact require separate evidence.

These controls keep the analysis useful for collaboration mapping while preserving uncertainty. Every graph node should trace to a PMID and source URL, and every identity merge should be reversible when new evidence arrives.