Published
- Patients checking health headlines
- Students
- Journalists
- Clinicians
Prepared · sources accessed (UTC) · physician review still to be recorded
A strong claim can travel far from its evidence. A study becomes a headline, a preliminary result becomes a recommendation, or an agency announcement becomes a statement that a rule already applies. The most useful reading habit is to rebuild the connection between the exact claim and the original record.
This workshop extends the site's health-study headline guide and rule-status decoder. It supplies a paper worksheet that works without uploading medical records or using a new app. The library and methodology page provide additional reading context; they do not replace examination of the cited evidence.
1. Find the original study behind the sentence
Copy the claim in a short sentence and identify what would make it true. Does it say a treatment improved a measured outcome, caused a change, reduced risk in a particular population, or merely showed promise? Search using the intervention, condition, author, journal, and year rather than repeating only the headline's dramatic wording.
The PubMed user guide explains citation searches, filters, full-text links, and records linking corrections or retractions. Use a database record to locate the article and check its status. Read the methods and results when available; an abstract may omit details needed to assess selection, missing outcomes, harms, or limitations.
For a clinical trial, look for its registration identifier and compare the registered population, interventions, and outcomes with the publication. ClinicalTrials.gov explains that sponsors or investigators submit and update study information and that government listing does not mean every study's safety and science have been approved. Registration is a useful record, not a quality certificate or proof of a favorable result.
2. Put effects back into denominators and time
Ask for the event counts or rates in both groups, the number of people or amount of follow-up, and the observation period. Determine whether the outcome is one that matters directly to patients or a laboratory measure that may require further interpretation. Record benefits and harms separately; an improvement in one outcome does not erase another outcome's uncertainty.
For simple risks over the same period, an absolute difference subtracts one risk from the other. A relative reduction divides that difference by the comparison risk. The arithmetic is straightforward, but the underlying estimate still depends on study design, uncertainty, missing data, and whether the groups are comparable. Do not apply a simple risk calculation to a hazard ratio or a rate without understanding the measure.
Ask whether the study population resembles the person or setting in question. A result in a narrowly selected trial group is not automatically a prediction for every patient. The medicine-access comparison adds a separate check: evidence, regulatory authorization, payment, and individual suitability answer different questions.
3. Read the instrument behind an agency release
An agency release is often a useful locator. Follow it to the actual document: proposed text, final regulation, order, guidance, report, or dataset. Identify the jurisdiction, issuing authority, date, affected activity, and stated legal or administrative status.
California OAL's rulemaking explanation distinguishes the processes used to adopt regulations and the role of procedural review. A proposal's official appearance does not make it a current requirement. Nor does a guidance document necessarily settle every question about statutory scope.
For a legal claim, preserve the exact instrument and any later amendment or status notice. A court decision requires attention to the issue decided and the court's authority; a party's filing describes that party's position. Use the California implementation guide to follow an enacted provision into a service process. This worksheet organizes questions; it does not determine the legal outcome of a personal dispute.
4. Preserve the earlier record when a page changes
Record publication date, stated update date, and your access date as separate fields. A new access date does not mean the underlying information is new. Save the relevant public document or a dated excerpt with its URL and title, subject to appropriate use of the material. Keep an explanatory note when a later version supersedes it.
For a research article, check corrections, expressions of concern, and retractions through the publisher and linked bibliographic records. For an agency page, look for a superseding notice, revised attachment, or explicit withdrawal. Do not assume a changed headline means every table in the document changed.
The site's existing Intelligence app describes a frozen snapshot. Treat it as a historical exploration surface rather than proof of a live current corpus. This guide adds a printable method; it does not announce a replacement workbench or restore any withdrawn data files.
5. Label a proposal, finding, interpretation, and forecast
Write a status label beside each sentence. A proposal describes something someone wants to do. A finding reports what an investigation or study observed within its methods. An interpretation explains what the observation may mean. A forecast estimates a future outcome under assumptions.
One article can contain all four. Keep the transitions visible: “The program proposes expansion”; “the report recorded completed visits”; “the authors attribute the change to better outreach”; “they project future growth.” Each statement needs different support. A descriptive increase alone does not establish what caused it.
When an authoritative source uses cautious language, retain that uncertainty in your own summary. When you disagree with an interpretation, state the evidence gap or alternative explanation instead of dismissing the entire record. The public-service tracker applies this distinction to announcements, expenditure, and delivered services.
6. Compare sources against one defined question
Do not count how many websites repeat a claim. Several summaries may all depend on one study or one press release. Trace that shared origin and identify genuinely independent evidence. Two sources can agree in wording while addressing different populations or time periods.
Use the worksheet to ask what each source directly supports, what it leaves unresolved, and whether it contradicts the claim or simply examines another question. Prefer a conclusion that fits the evidence: supported within stated limits, partially supported, unresolved, or unsupported by the records examined. These are reading judgments, not universal grades of an author or institution.
End with a next action proportionate to the decision. A classroom discussion may need a better citation. A treatment change needs clinical assessment. A disputed legal deadline needs the actual notice and qualified advice. A policy comparison may need a narrower question or additional data. The goal is a defensible next step, not an illusion that all uncertainty has disappeared.
Fictional worked example: a twenty-percent claim with a smaller absolute change
A fictional one-year study follows one hundred people in each of two groups. Ten in the comparison group and eight in the intervention group experience the defined event. A headline says the intervention “cuts risk by twenty percent.”
The arithmetic describes a relative reduction from ten percent to eight percent. The absolute difference is two percentage points, or two fewer observed events per hundred people over that year. Neither calculation alone establishes a precise treatment effect: the reader still needs the design, uncertainty, losses to follow-up, harms, and whether the outcome was prespecified.
Amir writes both numbers and checks the full paper and trial record. He finds that the headline concerns one endpoint, while the paper reports other outcomes separately. He summarizes the limited finding without calling the intervention a cure or changing a relative's treatment. All numbers and people in this example are invented for arithmetic instruction.
Printable claim-to-source comparison sheet
Claim in one sentence: ________
Decision this claim might change: ___________
| Check | Source A | Source B |
|---|---|---|
| Original title, issuer/author, URL | ____ | ____ |
| Publication / update / access dates | ____ | ____ |
| Study, proposal, final instrument, or commentary | ____ | ____ |
| Population, jurisdiction, setting, and period | ____ | ____ |
| Outcome, counts, denominator, effect measure | ____ | ____ |
| Uncertainty, missing data, exceptions | ____ | ____ |
| Direct support for the exact claim | ____ | ____ |
| Later correction, withdrawal, or superseding record | ____ | ____ |
Conclusion and limits: _________
Next source or qualified review needed: _____
- [ ] Repeated summaries were traced to their original source.
- [ ] Relative and absolute quantities are labeled correctly.
- [ ] A proposal or forecast is not presented as an observed result.
- [ ] I preserved the source version used for the conclusion.
Questions readers ask
Does a PubMed or trial-registry entry mean the conclusion is established? No. Use the entry to locate and examine the relevant research and its status; evaluate methods, results, and limitations separately.
Is a relative effect misleading? It can be informative when accompanied by the baseline, absolute quantities, time period, and uncertainty. The problem is presenting it without the context needed for a decision.
What if I can only access the abstract? State that limitation, avoid claims requiring unseen methods or results, and seek the full report or an appropriate assessment before relying on a consequential conclusion.
Sources and scope
Primary sources: NLM PubMed user guide, ClinicalTrials.gov explanation, and California OAL process. Prepared November 5, 2026 Pacific; accessed November 6 UTC. The worksheet is an educational reading aid. It is not a completed systematic review, individualized treatment calculator, current software workbench, or professional determination of a claim.