Failure at the bedside takes several forms. Observations may not be taken, or taken incompletely, with respiratory rate the sign most often left out. They may be recorded without being interpreted, each set judged alone and the trend unseen. The nurse may sense that something is wrong before the numbers change and lack the words to justify a call; a systematic review identified a set of indicators behind such worry, among them changes in breathing, circulation, mental state and agitation (Douw et al., 2015). Or the nurse may recognise the problem and hesitate to escalate, for fear of being wrong or of a dismissive reply. A scenario intended to teach recognition has to give the learner the chance to fail in each of these ways.

Tanner (2006) describes clinical judgement as four linked processes: noticing, interpreting, responding and reflecting. Noticing is shaped by what the nurse expects to see, which in turn depends on knowledge of the typical patient and of this particular one. Endsley's model of situation awareness, developed in aviation, makes a parallel distinction between perceiving the elements of a situation, comprehending their meaning and projecting what will happen next (Endsley, 1995). Both models imply that recognition is not a single act. A learner may perceive a raised respiratory rate and fail to comprehend it, or comprehend it and fail to foresee where the patient is heading. The Lasater Clinical Judgment Rubric turns Tanner's four phases into observable descriptors at four levels of development and gives designers a ready vocabulary for objectives and debriefing (Lasater, 2007).

The framework proposed by Jeffries (2005) directs attention to objectives, fidelity, problem solving, student support and debriefing as the design features of a simulation. The current standard of best practice on simulation design sets out criteria that include a needs assessment, measurable objectives, a scenario that provides the context for the experience, an appropriate type of fidelity, a prebriefing, a debriefing and pilot testing before full use (Watts et al., 2021). These documents say what a good scenario must contain. They do not say how to write one for a particular learning problem, and that is the gap the present procedure is meant to fill for the case of deterioration.

Objectives for a recognition scenario should name the cognitive act and the cue. "Manages a septic patient" is too broad. "Identifies a rising respiratory rate across three sets of observations and states its possible significance" can be observed and discussed. Three or four objectives are enough, and at least one should concern escalation, since recognition that leads to no action protects no one. Tanner's phases supply a convenient frame: one objective each for noticing, interpreting and responding.

The clinical story should be one the new nurse will meet within months: a chest infection after abdominal surgery, bleeding after a procedure, an older patient with a urinary infection who becomes confused, opioid-related respiratory depression. Rare conditions invite learners to treat the exercise as a diagnostic puzzle. The designer writes a short summary of the patient, including the baseline observations and the details a nurse would learn at handover.

This step distinguishes the procedure from conventional scenario writing. The designer scripts four or five sets of observations, spaced as they would be on a ward, in which values move gradually. In the first set everything lies within the patient's normal range. In the second, one or two values shift without crossing a threshold. By the third the direction is clear to anyone who compares the sets. Only in the fourth does the aggregate score require an urgent response. Time is compressed so that the scenario fits a teaching session, and the compression is announced to learners. The scripted values should be checked by an experienced clinician for plausibility.

Numbers are not the only evidence. The patient, played by an actor or voiced through a manikin, should show the softer signs that experienced nurses rely on: a change in colour, new restlessness, shorter sentences, a relative who says that something is different (Douw et al., 2015). The scenario should also contain the ordinary demands of a shift, such as a call bell elsewhere or a medicine that is due. Cognitive load theory cautions against overdoing this. Working memory is limited, and material that does not serve the learning goal competes with material that does (Sweller, 1988). For nurses in their first months one or two distractions are sufficient.

The designer marks on the scripted chart the point at which the local track-and-trigger system calls for escalation. Where the National Early Warning Score 2 is in use, an aggregate score of 5 or more is the threshold for an urgent response and 7 or more for an emergency response (Royal College of Physicians, 2017). The expected action is then written down: whom the learner should call and what the call should contain. A structured tool such as SBAR (situation, background, assessment, recommendation) gives the content (Leonard et al., 2004). The person who takes the call follows a script and may be instructed to respond with mild reluctance, so that the learner practises persisting.

Fidelity should follow from the cues. Dieckmann et al. (2007) distinguish the physical, the conceptual and the experiential realism of a simulation and argue that the last two matter more for learning than equipment does. For recognition, conceptual realism is essential: the observations must change as they would in a real patient. A simulated patient played by a person conveys restlessness or confusion better than most manikins, while a manikin allows chest sounds and a monitor. A real observation chart of the kind used on the learners' own wards should always be provided, since reading the trend from it is the skill being taught.

The prebriefing sets expectations, explains what the simulator can and cannot show, and establishes that mistakes are material for learning and not for assessment, which Rudolph et al. (2014) call the safe container. The debriefing is where noticing is examined. Debriefing with good judgement pairs the facilitator's honest observation with real curiosity about the learner's reasoning (Rudolph et al., 2006), and the PEARLS framework offers a script that blends learner self-assessment, focused discussion and direct feedback (Eppich and Cheng, 2015). For this kind of scenario the facilitator's guide should include questions aimed at each of Tanner's phases: what the learner saw at each set of observations, what it was taken to mean, and what prompted or delayed the call.

A scenario should be run with a few nurses from the target group before general use. The designer watches for the point at which they notice the change. If everyone notices at the second set of observations the trend is too steep, and if no one notices before the trigger it is too flat or the soft cues are too faint. The script is adjusted until the scenario discriminates, and the facilitator guide is revised in the light of the questions that proved productive.