How to use the lab

From idea to decision

A structured way to test what would need to be true for an intervention to create value.

How the sandboxes are used

Each sandbox helps structure assumptions, test value, and clarify what needs validating next.

What this process does

Intervention idea

Start with a plausible service change or intervention.

Translate assumptions

Turn the idea into explicit assumptions about scale, risk, effect, and cost.

Observe model signal

Review the implied effect on events, cost, and value.

Test uncertainty

Change assumptions to see where the case is robust or fragile.

Clarify what to validate

Identify what matters most and what needs testing next.

In practice

Two example decision questions

Prevent Need

Falls prevention (SafeStep)

Question

Can a falls prevention programme reduce admissions and bed use enough to create value?

Set the base case

population sizebaseline riskintervention reachexpected risk reductioncost per participant

Review what the model suggests

falls avoidedadmissions avoidedbed days avoidednet costcost per QALY

Check what matters

effectivenessengagementdelivery costtargeting of higher-risk groups

Decision signal

The programme appears valuable if it meaningfully reduces falls in a sufficiently high-risk population at a controlled delivery cost.

Improve Access

Waiting list optimisation (WaitWise)

Question

Can waiting list interventions reduce pressure without simply shifting activity elsewhere?

Set the base case

size of waiting listproportion suitable for removal or redirectiondownstream activity assumptionsdelivery cost

Review what the model suggests

waiting list reductionescalations avoidedadmissions avoidedcost impactoverall value signal

Check what matters

avoidable activitydownstream pathway costssustainability of change

Decision signal

Value is created only if unnecessary activity is genuinely reduced, not redistributed.

What this process does not do

  • It does not produce a final answer.
  • It does not replace formal evaluation.
  • It does not remove uncertainty.

What it gives instead

A clearer decision.

This is likely to work if these conditions hold.

One consistent question

Across all sandboxes, the same question sits underneath the model:

What would need to be true for this to create value?