Clinical Decision-Making and Evidence-Based Practice
This domain tests whether you can carry evidence into an actual decision for a specific patient, team, or service. HS 326 wants the five-step process applied rather than recited: ask, acquire, appraise, apply, and audit the result. Questions often hinge on baseline risk, patient values, and the difference between a statistically valid finding and a warranted action.
The applied chain
- Frame the question with PICO and specify the population that matches your patient.
- Convert relative effects to absolute: ARR = baseline risk x (1 - RR); NNT = 1/ARR, and NNH for harms.
- Compare NNT and NNH before recommending; a favourable NNT loses value if harms are frequent or severe.
- Check applicability: is my patient represented in the trial, is the intervention feasible here, do the outcomes matter to the patient?
- Integrate patient values through shared decision-making; a decision aid showing natural frequencies beats a relative-risk slogan.
- Close the loop with audit and feedback or PDSA cycles, measuring whether practice changed and whether outcomes moved.
Guidelines and prediction rules
Appraise guidelines with AGREE II and look for a dated systematic review, explicit grading, and disclosed conflicts. Validate clinical prediction rules in your own population before use; a rule derived in one setting often loses discrimination elsewhere, and a rule is a probability statement, not a diagnosis.
Worked example
A trial reports RR 0.80 for a cardiovascular event with a new agent. In a patient whose 10-year baseline risk is 20 percent, ARR = 0.20 x 0.20 = 4 percent, so NNT = 25. In a lower-risk patient at 5 percent, ARR = 1 percent and NNT = 100. The relative benefit is identical; the decision is not. If the drug causes serious bleeding in 1 per 200 treated (NNH 200), the higher-risk patient still benefits clearly, while the lower-risk patient is close to equipoise, and that is precisely where the patient's own weighting of stroke against bleeding decides.
Traps that cost marks
- Applying a trial's relative risk reduction to a patient with a very different baseline risk without recomputing the absolute effect.
- Quoting an NNT without stating the baseline risk and the time horizon.
- Using a prediction rule as if it were a diagnostic test.
- Treating strength of recommendation as identical to certainty of evidence; they are separate GRADE axes.
- Skipping the audit step so the evidence-based practice cycle never closes.