Background A Clinical Rule (CR) can be defined as an algorithm that combines different patient-related factors (drugs, medical history and laboratory values) and generates an alert when specific conditions are present (e.g. hypokalaemia without potassium supplements).

Purpose To optimise the sensitivity and specificity of the CR ‘Potassium’.

Materials and methods The ’Potassium’ CR has been revised four times since the implementation of the CR database:

1. At first it generated an alert when a potassium-affecting drug was used and the potassium level was abnormal (<3.0 mmol/L or > 5.0 mmol/L).

2. The CR was then divided into two separate rules:

3. A hypokalaemia rule generating alerts if a potassium-lowering drug was used and the potassium level was low (<3.0 mmol/L).

4. A hyperkalaemia rule generating alerts if a potassium-increasing drug was used and the potassium level was high (>5.5 mmol/L).

5. The clinical rules were further developed:

6. The hypokalaemia rule generated alerts if a potassium-lowering drug was used, the potassium level was <3.0 mmol/L, and no potassium supplements were used.

7. The hyperkalaemia rule generates alerts if a potassium-increasing drug is used, the potassium level is >5.5 mmol/L, and no potassium binder is used.

8. The hypokalaemia CR was adjusted so that alerts are generated if the potassium levels are <2.0 mmol/L, despite the use of potassium supplements.

Results The number of alerts generated by the ‘Potassium’ CR was reduced with every change to the CR. When a potassium rule was first used there were approximately 25 alerts concerning potassium per day, of which 2–3 required an intervention (adding or stopping a potassium-affecting drug).

With the latest version of the ‘Potassium’ CR 2–3 alerts are generated per day and all require an intervention.

Conclusions Daily practice and regular evaluation make it possible to optimise clinical rules by increasing the sensitivity and specificity.

No conflict of interest.