Hospital Operations Simulation: Why a Good Metric Is a Band, Not a Direction

TL;DR: In hospital management no single metric can be improved on its own; the decision that raises clinical quality usually raises cost as well. A simulation makes that trade-off visible, but only if the metric is designed correctly. Bed occupancy is the clearest example: the scenario's own options raise occupancy most in the decisions that strain capacity — so a metric built as "the higher the better" rewards precisely the opposite of what the scenario teaches. We examine why, in a healthcare operations simulation, a metric is defined as a band rather than a direction.

Healthcare organisations do not suffer from a shortage of indicators. Infection rate, average length of stay, patient satisfaction, bed turnover, unit cost: the dashboard that lands on a hospital manager's desk usually carries dozens of lines. What is missing is not the indicator but the way indicators should be read against one another. Everyday reporting rarely shows that a single decision moves all of those lines at once and pushes some of them in the opposite direction. The healthcare scenario in SimAna Field Operations was built for exactly this gap: over eight weeks the participant makes the operational decisions of a private hospital and sees the effect of each decision on four metrics in the same view. In this article we examine not the content of the scenario but its metric design, because what a business simulation teaches is determined largely by what it scores.

Four metrics, four different directions

The Healthcare Operations scenario carries four business metrics. Three count as good when they rise, one when it falls. The weights determine the composite ranking and they are not equal: clinical quality alone carries more than a third of the total weight.

Metric	Starting value	Direction	Ranking weight
Clinical Quality	70 points	Should rise	35%
Patient Satisfaction	78%	Should rise	25%
Operating Cost	₺12,000,000	Should fall	20%
Bed Occupancy	72%	Should stay within a band	20%

The last row of that table is the subject of this article. The other three are directional metrics in the classical sense; bed occupancy is not. To see why it is not, we first have to look at how the decisions are built.

Eight weeks, eight decisions

The scenario runs across eight weeks and puts a single operational decision in front of the participant each week. The decisions are drawn from a hospital's real agenda and follow on from one another:

The infection-control approach in response to rising hospital-acquired infection rates.
The staffing policy to follow when the number of patients per nurse increases.
The management of patient flow in the face of lengthening emergency-department waits.
The investment decision for an MRI machine whose breakdowns are becoming frequent.
The management of the elective-surgery schedule given idle operating-room capacity.
The level and budget of preparation for the JCI accreditation audit six weeks away.
The management of winter-season emergency admissions and bed-occupancy pressure.
The management of patient experience in the face of rising complaints and negative online reviews.

The eight weeks are divided into three phases: situation assessment, improvement and sustainability. The phase structure reflects the fact that the outcome of a decision is not always visible in the week it is made; the clinical effect of a cost-saving decision taken in the first phase is read in the third.

The anatomy of one decision

The second week's decision is the most legible example of the trade-off structure. The scenario tells the participant this: the number of patients per nurse has risen, the risk of burnout and medical error is growing, you must set your staffing policy. The four options and their effects on the metrics are as follows:

Option	Clinical Quality	Patient Satisfaction	Operating Cost
Hire additional nursing staff to improve the ratio permanently	+4	+4	+₺600,000
Support the peak hours with flexible shifts and part-time reinforcement	+2	+2	+₺250,000
Freeze the open positions and manage with the current staff	−2	−3	−₺100,000
Cut staffing costs aggressively by raising patients per nurse	−4	−4	−₺300,000

There is no hidden right answer here; in a hospital whose budget has tightened, the third option is a defensible decision. What the simulation measures is not the choice itself but whether the participant noticed the trade-off and held a consistent line across eight weeks. This is exactly what is meant by modelling the decision process rather than the decision's outcome.

Why a metric has to be a band, not a direction

Now back to bed occupancy. Intuitively, the higher the occupancy the better: an empty bed is idle capacity and does not cover its fixed cost. A metric built on that intuition rewards the decision that raises occupancy the most. The problem is that in this scenario the decisions that raise occupancy the most are the decisions that damage the hospital the most.

Two options raise bed occupancy by five points, and both are the path the scenario warns against:

Option	Bed Occupancy	Clinical Quality	Patient Satisfaction
Increase elective surgeries to the maximum and fill the beds	+5	−2	−1
Exceed capacity with corridor and extra beds and admit every patient	+5	−4	−3
Expand the elective schedule moderately with an extra preparation team	+3	+1	+1

Under a one-way metric the first two rows would score higher than the third. In other words the model would rank the decision that lowers quality and satisfaction together above the decision that raises both. This is not an arithmetic error but a definition error: the metric has been asked the wrong question.

The right approach is to accept that occupancy has a good band. Below the band capacity sits idle; above it there is no buffer left for emergency admission, isolation and cleaning, which means the hospital looks efficient on paper while it has spent its safety margin. The scoring curve therefore peaks at the target and, past the upper threshold, does not keep rising but falls. The rule of metric design is this: it must not be possible to score highly by pushing a single metric to its limit — if it is possible, the metric has been defined wrongly.

This choice is recorded as a rationale in the platform's own codebase: the band curve was written on the observation that, because the scenario's own options raise occupancy most when the participant exceeds capacity, a one-way metric would award the highest score to the very answer the scenario teaches against. The thresholds are not guessed; they are set together with the healthcare organisation using the scenario and explained to the participant before the session.

What the score says: relative rank or absolute level

The second question, as important as the metric's direction, is how the raw value is turned into a score. There are two options and they say different things.

The first is to position the participant against everyone else in the same session. This method is easy but makes a weak statement: even within a cohort that is doing badly there is a participant in first place, and someone playing alone cannot be ranked at all. The second is to define two anchors for the metric: the expected level a competent manager reaches, and the organisation's own target. The expected level corresponds to 50 points, the target to 100. In that case the score is absolute; a 78 in March means what a 78 in October means, and it is meaningful for a single participant too.

The healthcare scenario uses the second method and is the only Field Operations scenario with all four of its metrics anchored. The practical consequence is this: the number a participant receives answers "how well did you do" rather than "where are you among these people", and it is comparable across periods. Rank is separate information, and it is not shown at all while the cohort is below five participants; in a group of three, the sentence "you are third" carries no information and only misleads.

The same hospital, a different lens

The Field Operations scenario measures what a manager chooses: week by week, one decision, four metrics. The other half of a manager's job is putting dozens of simultaneous demands in order, and that cannot be measured by a weekly decision cycle. In SimAna's healthcare portfolio this second lens is carried by the Decision Box Healthcare Management scenario: the inbox of a 200-bed hospital, sixty minutes, twenty emails from eight different roles. The two products are not rivals but complements; one measures the content of the decision, the other its ordering.

Conclusion

A simulation's teaching power comes not from the realism of the scenario but from the honesty of the scoring. A realistic hospital case, combined with a badly defined metric, teaches the participant a bad habit and presents it as measurable success. The bed-occupancy example is the concrete form of that risk: the metric definition that fits intuition best is the one that contradicts the scenario's own content. The question to ask when evaluating a simulation is therefore not "which indicators does it measure" but "what happens to the participant who pushes one indicator to its limit". If the answer is "they win", what is being measured is not management.

Because occupancy has a good range rather than a good direction. Low occupancy means idle capacity; very high occupancy removes the buffer needed for emergency admission, isolation and cleaning. The distribution of options in the scenario confirms it: the two decisions that raise occupancy the most are the decisions that lower clinical quality and patient satisfaction together. A one-way metric rewards those two decisions, which puts what the model teaches at odds with what the scenario says.

It is comparable to the extent that the metrics are anchored. If the score is calculated only against the other participants in the same session, comparison across periods is meaningless, because the scale is rebuilt in every session. When expected and target values are defined in advance the scale is fixed and the same number expresses the same level in every period. In the healthcare scenario all four metrics are defined this way; rank is separate information and is not shown in small groups.

The aim is not to compress a year but to make the link between decisions visible. Eight decision points are long enough to show that the clinical consequence of a cost-saving choice emerges weeks later, and short enough to complete in a single session. The phase structure carries that delay as well: a decision made in the first phase is read in the third. The full complexity of a real hospital is not modelled at this scale, nor is it meant to be; what is meant is that the trade-off be experienced.