A risk estimate is information for a conversation, not the conversation's conclusion. DynaMed Decisions is designed to connect individualised estimates, treatment options and patient discussion. Its most useful distinction is also easy to miss: EBSCO currently states that the named product does not employ AI in its tools. Digital decision support and generative clinical answers are not interchangeable categories.
That statement comes from EBSCO's DynaMed Decisions page, checked on 26 September 2026. The page describes patient-specific decision tools, Option Grid patient decision aids and customisable chart notes. Those are product-design features, not proof that every input is correct or that every resulting decision is appropriate.
This article is published by iatroX and includes iatroX among the complementary tools discussed. The comparison concerns the work each type of resource performs, rather than declaring a universal winner between a calculator, a decision aid and a conversational reference.
Separate the calculator, the decision aid and the clinical answer
A calculator takes defined inputs and applies a specified model or rule. Before relying on the output, the clinician needs to know what the inputs mean, what the model predicts and whether the patient falls within its intended use.
A decision aid helps organise a choice. It can present alternatives, benefits, harms and practical burdens so that a patient can consider them in relation to what matters to them. It does not make the patient's priorities visible unless those priorities are actually discussed.
A generative reference answer addresses a question in prose, usually drawing on some combination of retrieved material and model behaviour. Its relevant inspection includes source support, context, uncertainty and whether the answer addresses the question asked. A fluent explanation is not automatically a validated risk calculation.
As checked on 26 September 2026, EBSCO's description of DynaMed Decisions includes predictive tools based on identified evidence and expert review, together with Option Grid materials. The product information should not be extended into an assertion about every product in the wider DynaMed family.
Audit the inputs before discussing the number
The first task is not interpretation of the result but verification of the inputs. A field labelled with a familiar clinical term may have a specific definition. A measurement may require a particular unit, a defined time point or an established method of collection. A historical diagnosis may not mean the same thing as a current finding.
Check whether information was entered, imported or assumed. An automatically populated value is still a value to verify. A missing item should not quietly become a reassuring default merely because the calculation can proceed without an obvious error message.
Then ask what outcome the model estimates. A risk of one event over a stated period is not interchangeable with lifetime risk, symptom severity or the chance of benefiting from a particular treatment. Even a well-calibrated risk model for a population does not tell you which individual will experience the outcome.
For a fictional consultation, the clinician might pause because the patient's recent circumstances differ from those represented by the available historical information. The right response is to clarify the discrepancy, not select whichever input generates the preferred result.
Read the model's population and limits
Look for the population in which the underlying model was developed and assessed. Consider age range, clinical setting, exclusion criteria and the outcome definition. A familiar disease label does not establish that a person with substantial additional complexity is adequately represented.
Separate a model's predictive performance from the benefits of acting on its output. A tool may estimate an outcome reasonably well without establishing that a particular implementation improves patient outcomes or that a chosen threshold is appropriate in every service.
Also distinguish uncertainty in the estimate from uncertainty about applicability. The first concerns the information provided by the model or supporting evidence. The second concerns whether that information transfers to this patient and setting. A neat number on the screen can conceal both unless the clinician deliberately asks about them.
These are practical appraisal questions rather than a criticism unique to DynaMed Decisions. They apply equally to a stand-alone calculator, a tool embedded in a record system or a numerical claim inside a generated answer.
An original example of absolute and relative effects
The following figures are deliberately invented for an arithmetic and communication exercise. They are not DynaMed Decisions output, a clinical trial result or advice about a real intervention.
Suppose an illustrative comparison describes 12 events among 100 similar people over five years without an intervention and nine events among 100 over the same period with it. The absolute difference is three events per 100 people, or three percentage points. The relative reduction is 25%, because three is one quarter of 12.
Those descriptions concern the same hypothetical difference, but they do not communicate it equally well in isolation. A patient hearing only "25% lower risk" has not been told the starting risk or the time period. A patient hearing only "three fewer events" still needs the denominator, the outcome and the uncertainty.
The Cochrane Handbook's discussion of interpreting results, checked on 26 September 2026, emphasises absolute effects, applicability and the distinction between evidence and decisions. The arithmetic above is an original illustration of those principles, not a summary of a Cochrane intervention finding.
From estimate to options, priorities and a recorded decision
The next step is to compare what the options would mean in practice. Treatment burden may include monitoring, appointments, inconvenience or effects the patient particularly wishes to avoid. Choosing not to proceed now, where clinically appropriate, may also require an explicit review plan rather than being treated as the absence of a decision.
| Risk estimate or evidence | Options to discuss | Patient priorities to establish | What the record should capture |
|---|---|---|---|
| The artificial 12-versus-nine example above | The intervention and the relevant alternative | How the possible benefit is weighed against its burden | The absolute comparison and the patient's understanding |
| Important uncertainty about applicability | Clarify missing information or obtain further advice | Whether the patient prefers more information before deciding | What remains unresolved and how it will be addressed |
| Similar expected outcomes with different burdens | The available approaches, without assuming one preference | Practical demands, concerns and acceptable trade-offs | Why the agreed option fits the discussion |
| A decision to reconsider later, where appropriate | A defined review rather than an indefinite deferral | What information or change would help a later choice | Responsibility, timing and reasons to revisit |
This is an original documentation framework, not a reproduction of an Option Grid or a claim about the exact fields in DynaMed Decisions. It shows why recording only the calculated score would miss much of the decision.
A shared decision can be well reasoned even when the patient values the options differently from the clinician. The purpose is not to obtain agreement with a predetermined choice. It is to support an informed discussion of the relevant options and the individual situation, consistent with the GMC's decision-making and consent guidance, reviewed on 26 September 2026.
What should the chart note say?
As checked on 26 September 2026, EBSCO describes customisable DynaMed Decisions notes that can be copied and pasted into the record. That is not the same as establishing an automatic write-back integration in every clinical system. The product page is the source of the documentation description.
Before saving any prepared note, check that it reflects the actual conversation. A template may describe options that were not discussed, assume a preference that was never expressed or omit a reason the estimate may not apply. Editing is part of the clinical task, not an optional cosmetic step.
A useful note distinguishes the evidence considered, material uncertainty, options discussed, the patient's priorities, the decision reached and arrangements for review. It should not imply that a generated or copied paragraph is a transcript of consent.
For example, "risk discussed" does not establish which risk, over what period or what the patient understood. Conversely, a long imported explanation can make the record harder to use if the patient's actual preference remains buried within it.
Where iatroX can help without duplicating the product
Per iatroX product information, September 2026, Ask-iatroX provides free source-linked clinical reference grounded in NICE, CKS, SIGN and SmPC information from emc. It can support a question about the surrounding evidence or guideline context; it should not be described as reproducing every DynaMed Decisions model or patient decision aid.
A separate learning need may emerge after the consultation. A clinician might recognise that they struggle to explain an absolute difference or distinguish a predicted risk from a treatment effect. Per the same September 2026 product information, iatroX's Socratic Tutor uses targeted questioning around an attempted question to explore a misconception. That is an educational use, not an independent validation of the clinical calculation.
For a clinician needing a named decision model and associated patient-facing comparison, inspect DynaMed Decisions' relevant tool and access arrangements. For someone seeking the broader guideline context, a suitable reference may answer the question. For someone who understands the number but not how to explain it, focused communication practice may add more than another calculator.
Frequently asked questions
Does DynaMed Decisions use generative AI?
EBSCO's DynaMed Decisions page checked on 26 September 2026 states that its tools do not currently employ AI. That statement concerns the named product, not every offering within the broader DynaMed family.
Is a risk score enough to document shared decision-making?
No: the score does not record the options discussed, the patient's priorities, material uncertainty or the agreed review arrangements. The note should represent the conversation rather than simply preserve a number.
Are the numerical examples in this article real treatment effects?
No: the 12-versus-nine example is deliberately artificial and illustrates arithmetic and communication only. It is not a clinical recommendation, product test or estimate for an actual patient.
Explore the evidence behind a clinical decision with Ask-iatroX →