How to Trace a Guideline Recommendation Back to Its Evidence

Tracing a guideline recommendation back to its evidence means reconstructing the chain from a specific recommendation statement to the studies that support it, then judging how well those studies fit the population and outcome the recommendation addresses. The practical goal is a defensible evidence brief: a short document that names the recommendation, lists the underlying studies, records how you found them, and states clearly where the evidence is direct, indirect, or limited.

This article walks through that task for clinicians and researchers preparing evidence briefs, research updates, or team discussions. It uses one focused example throughout: a recommendation about a treatment effect on a defined clinical outcome in a defined adult population. It does not give individual treatment advice. Instead it shows the decisions, artifacts, and checkpoints that let you finish the trace and know when the result is ready.

Before searching, establish your starting state. Most traces begin from one of three positions, and each needs a different first move.

• Recommendation only. You have the recommendation text and perhaps a citation number, but not the studies behind it. Your first job is to locate the guideline's reference list and any linked evidence review.
• Recommendation plus a reference list. You have candidate studies but no sense of which ones carry the recommendation. Your first job is to map each citation to the specific claim it supports.
• A partial evidence set. You have some studies but suspect gaps. Your first job is to define eligibility criteria so your search can find what is missing rather than re-finding what you have.

Write down your starting position in one sentence. This becomes the first line of your evidence brief and tells you which checkpoint you are working toward.

The single most useful artifact in a trace is a written PICO statement derived from the recommendation itself: population, intervention, comparator, and outcome. Cochrane guidance on planning searches begins from eligibility criteria, so build those criteria directly from the recommendation's wording rather than from your general impression of the topic.

For the worked example, suppose the recommendation states that a particular intervention improves a defined outcome in adults with a specified condition compared with usual care. Your PICO becomes:

• Population: adults with the specified condition, as defined in the recommendation.
• Intervention: the named treatment at the dose or format the recommendation describes.
• Comparator: usual care, as the recommendation frames it.
• Outcome: the specific outcome the recommendation claims to improve.

Keep this statement visible while you work. Every study you keep or discard should be judged against it, and every deviation should be recorded.

Cochrane's guidance on searching for studies recommends planning searches from eligibility criteria, using multiple sources, combining free-text terms with subject headings, and documenting the process. Apply each element deliberately.

Turn each PICO element into a concept block. For each block, combine free-text terms with controlled vocabulary. The NLM Medical Subject Headings system provides preferred headings, entry terms, scope notes, qualifiers, and tree positions, which you can use to expand a concept to related headings or focus it with a qualifier. Record which headings and which free-text terms you used for each block.

PubMed supports field tags, phrase searching, filters, Advanced Search, saved searches, Similar Articles, citation lookup, and links to full text when available. Its search details view shows how your terms were translated and combined, which is exactly the record you need for reproducibility. Save the search and note the date you ran it.

Multiple sources matter because a single database may not index every report of a study. Cochrane also notes that one study may have several reports, so plan to link multiple reports of the same study rather than counting them as separate studies. Deduplicate records before screening, and screen in stages so that title and abstract screening precedes full-text review.

If you want a scholarly search interface to run and refine these concept blocks, WisPaper search is the place to find academic papers; a public entry point for finding papers is also available through WisPaper Scholar Search.

Checkpoint 1: You have a written PICO, a documented strategy with concept blocks, at least two sources searched, a saved search record, and a deduplicated candidate list. If any of these is missing, do not proceed to screening.

Screening is where traces most often go wrong, because a study with several reports can look like several studies, and a secondary report can look like primary evidence.

1. Title and abstract screen. Apply your eligibility criteria. Record a reason for every exclusion.
2. Full-text screen. Confirm the population, intervention, comparator, and outcome actually match your PICO.
3. Link reports. Group multiple reports of the same study under one study identifier so your evidence count reflects studies, not papers.
4. Flag the study design. Note whether each included study is a randomized trial, an observational study, or another design, because design is part of how you will describe the evidence later.

Keep a screening log with three columns: record, decision, and reason. This log is the audit trail that makes your brief defensible.

These four judgments are related but distinct, and conflating them is the most common analytical error in an evidence brief. Treat them as four separate columns, not one verdict.

• Evidence level describes the kind of study and its design. It is a property of the study, not of your population.
• Applicability asks whether the study population, setting, and intervention match the recommendation's population and intervention.
• Indirectness asks whether the comparison or outcome in the study differs from the comparison or outcome in the recommendation. A study can be high-level evidence and still be indirect for your question.
• Limitations are the weaknesses within a study, such as how the outcome was measured or how the comparison was constructed.

A study can score well on one dimension and poorly on another. Your brief should say so explicitly rather than collapsing everything into a single quality rating.

Return to the example recommendation about an intervention improving a defined outcome in a specified adult population compared with usual care.

Step 1 — Anchor the claim. Copy the recommendation sentence verbatim into your brief. Underline the population, the intervention, the comparator, and the outcome.

Step 2 — Map citations to claims. For each citation attached to the recommendation, write one line stating what claim it supports. If a citation supports only background and not the effect claim, mark it as background.

Step 3 — Retrieve and screen. Run your documented strategy, deduplicate, and screen in stages. Link multiple reports of the same study.

Step 4 — Build the evidence table. For each included study, record design, population, intervention, comparator, outcome, applicability judgment, indirectness judgment, and limitations.

Step 5 — Write the synthesis paragraph. State what the body of evidence shows about the recommendation's specific outcome in the specific population, then state where the evidence is indirect or limited. Keep the paragraph narrower than the evidence: if the studies enrolled a broader population than the recommendation names, say the evidence is broader and note the applicability gap.

Step 6 — Record gaps. If your search found no study addressing the exact outcome in the exact population, record that as a gap rather than filling it with inference.

Checkpoint 2: Your evidence table has one row per study, your screening log explains every exclusion, and your synthesis paragraph names applicability, indirectness, and limitations separately. If the synthesis paragraph makes a claim that no row in the table supports, revise it.

A trace is only as good as the set of papers behind it. Save the studies you include, and keep the excluded ones logged separately so a colleague can re-run your decisions. If you want a place to save or upload papers and organize them, My Library supports saving and organizing papers in the WisPaper library. For a team effort, Research Projects is the WisPaper area for creating and working within a research project.

When your included papers are in your library, you can ask questions grounded in those uploaded or added papers using Library QA. Use it to interrogate the set you have already assembled, not to replace the search and screening steps that produced the set.

Checkpoint 3: A colleague can open your library or project, see the included studies, read the screening log, and reproduce your search from the saved record.

Before a brief goes to a team, check that every cited item resolves to a real record. Citation errors propagate quickly in guideline traces because recommendation reference lists are often long and secondary.

TrueCite supports BibTeX citation screening by checking BibTeX against real academic databases to flag hallucinated references. Run it on the reference list you plan to circulate, then reconcile any flagged items against your screening log before finalizing.

Checkpoint 4: Every citation in the brief resolves to a record you have screened, and every flagged item has been reconciled.

The trace is ready when five conditions hold:

1. The recommendation is quoted verbatim and its PICO is written out.
2. The search is documented well enough to re-run, with sources, terms, and date recorded.
3. Every included study has a row in the evidence table with evidence level, applicability, indirectness, and limitations judged separately.
4. The synthesis paragraph makes no claim that a table row does not support, and gaps are recorded as gaps.
5. Every citation resolves to a real record.

If any condition fails, the brief is a draft, not a finished trace.

Use these rules to decide when to stop working or hand the question to someone else.

• Stop when the chain is complete. If every claim in the synthesis maps to a table row and every citation resolves, stop. Further searching adds cost without changing the conclusion.
• Escalate when the recommendation's PICO cannot be pinned down. If the recommendation text does not specify the population or outcome clearly enough to build eligibility criteria, the ambiguity belongs with the guideline authors or the clinical team who owns the recommendation.
• Escalate when the evidence set is empty for the exact outcome. Record the gap and route the question to whoever can decide how to handle an evidence gap; do not substitute a broader outcome silently.
• Escalate when applicability is contested. If team members disagree about whether a study's population matches the recommendation's population, document both judgments in the table and let the team resolve it rather than choosing one silently.
• Re-run the search before circulation. Cochrane recommends rerunning searches before publication. If your brief has been in draft for a while, re-run the saved search and log what changed.