When two clinical references appear to disagree, compare the decisions they are addressing before choosing a side. Different patient groups, stages of care, jurisdictions or publication dates can produce different advice without a direct contradiction. If the conflict remains, identify the consequential claim and investigate its supporting evidence rather than accepting the longer or more confident answer.
This article uses an original fictional exercise, not a documented disagreement between UpToDate, BMJ Best Practice or a particular NHS service. It is published by iatroX and applies the same source-checking expectations to Ask-iatroX. Product descriptions and source routes were checked on 20 September 2026.
Write the disagreement in a single sentence
"The websites disagree" is too vague to investigate. A useful statement identifies the decision, the alternatives and the circumstances under which they appear to differ. For example: one document seems to recommend additional assessment, while another appears to support routine follow-up, but their intended populations have not yet been compared.
Do not begin by ranking the publishers. Find the exact paragraph and read the surrounding section. A recommendation may depend on an earlier definition, an exclusion or a different stage of the pathway. Search results and isolated excerpts often leave those qualifications behind.
Record the version or revision information available, the date accessed and the relevant source links. That makes the question answerable for a supervisor, librarian or specialist. It also prevents a disagreement about an older screenshot from being mistaken for a conflict between current guidance.
Three fictional source cards
Consider a clinician reviewing follow-up after an initial assessment. The following cards are invented teaching summaries and are not quotations, recommendations or paraphrases from named products.
| Card | What it addresses | Qualification that matters |
|---|---|---|
| A | General follow-up after an uncomplicated assessment | Assumes the person is stable and the initial assessment is complete |
| B | Reassessment when the presentation remains unexplained | Includes people with unresolved or changing features |
| C | A local referral service's acceptance process | Specifies the information needed for that service to process a request |
If Card A is read as applying to everyone, it appears to conflict with Card B. If Card C is mistaken for a national clinical standard, an operational condition can appear to contradict both. The first task is therefore classification, not voting on which card is best.
The exercise deliberately supplies no clinical thresholds or treatment instructions. Its purpose is to show how a reader can separate a general recommendation, a conditional recommendation and a service process.
Compare the patient population before the conclusion
For each source, write who is included and who is excluded. Consider age group, disease stage, comorbidities, previous treatment, severity and the point in the clinical pathway. Only include factors that matter to the actual question rather than creating an indiscriminate checklist.
In the fictional exercise, stability and completeness of the initial assessment are the key distinctions. A reader who overlooks those assumptions could learn the wrong rule even while accurately copying the conclusion.
A useful question for a colleague is: "What fact would need to be true for this recommendation to apply?" The answer may reveal that the case has not been sufficiently characterised. Missing information should remain missing, not be filled in to make one source fit.
If the populations genuinely match and the recommendations still differ, move to the evidence and date rather than repeatedly searching for a preferred answer.
Check what changed over time
Record both the publication information and the date of the evidence being discussed where available. A newly updated webpage can describe an older study. A long-standing review can remain relevant but omit a subsequent development. These are questions to investigate in the source, not reasons to reject either automatically.
When a recommendation changes, ask what changed with it: the evidence, the population, the interpretation, the available intervention or a practical constraint. A new date alone does not explain the difference.
For a learner, keep a correction note that identifies the old assumption and the current source. Do not retain two contradictory flashcards without their conditions. The goal is to learn a qualified decision, not memorise whichever answer was encountered last.
Evidence interpretation can differ even when the studies overlap
Two references may address similar evidence but give different weight to outcomes, uncertainty or practical considerations. The responsible next step is to inspect how each conclusion was reached and seek appropriate expertise where the decision is consequential.
UpToDate's published editorial process describes an authored and reviewed clinical-reference approach. BMJ Best Practice's public offering includes point-of-care guidance and its Comorbidities Manager. These descriptions, checked on 20 September 2026, explain aspects of the products' design; they do not establish which would be correct in an untested disputed case.
Nor should disagreement be resolved by counting citations. Several references may concern the same study population, while one carefully chosen source addresses the precise exception at issue. Relevance and support matter more than the length of the bibliography.
Local guidance answers an additional question
A local pathway can specify where to refer, which information accompanies a request and who is responsible for follow-up. Those operational details may not appear in an international reference. Their absence is not necessarily a defect in a product that does not claim to maintain that service's current arrangements.
HealthPathways, checked on 20 September 2026, is an example of a platform organised around localised clinical and referral pathways. Right Decisions hosts guidance collections whose relevant scope must be checked. Do not treat one region's collection as applicable everywhere.
Local instructions should also not be used to hide a substantive clinical uncertainty. If a pathway appears inconsistent with a relevant national source, raise the specific issue through the appropriate clinical route. Distinguish a capacity constraint from a statement about what is clinically indicated.
For the fictional cards, establish whether Card C describes service eligibility or a clinical recommendation. That single clarification can prevent an unnecessary argument about sources that were answering different questions.
A source-reconciliation record
An original working record can contain the disputed decision, the patient's relevant characteristics, the source versions, the conditions attached to each recommendation and the unresolved question. Add the person or service consulted and the basis of the final decision where appropriate.
For teaching, keep patient information fictional or appropriately governed. A useful discussion might show that both recommendations are coherent within their intended populations. Alternatively, it may reveal a genuinely unresolved conflict requiring further advice. Both outcomes are more honest than forcing agreement.
As described on 20 September 2026, Ask-iatroX's methodology includes source retrieval and citation grounding. It can support clarification of the question, but it should not be treated as an independent arbiter whose synthesis automatically overrides the underlying sources.
Ask a bounded question such as which assumptions differ between two permitted summaries. Inspect any sources returned. Do not upload protected content or identifiable patient material without the necessary permission and governance.
Choose the next step by the type of disagreement
For a population mismatch, refine the case and apply the appropriate source. For a date mismatch, establish what has changed. For an operational mismatch, consult the relevant local pathway. For a genuine evidence or interpretation conflict, seek appropriate expert review and document the uncertainty.
A broad clinical reference, a local service resource and a learning tool can each contribute. None becomes the universal winner because its answer is easiest to read. The useful result is a traceable explanation of which source applies to which part of the decision.
Frequently asked questions
Does this article show that UpToDate and BMJ Best Practice disagree on a clinical topic?
No, the source cards are fictional teaching material. A claim about an actual product disagreement would need the exact dated sources and a documented comparison.
Should local guidance always override a general reference?
Local operational instructions and general clinical evidence answer different questions. Clarify the nature of the conflict and seek appropriate advice when clinical recommendations remain inconsistent.
Can an AI summary reconcile the sources for me?
It can help organise the comparison, but its interpretation still needs checking. Do not let the summary erase differences in population, uncertainty or service scope.