Table 1 Aiming for perfection — characteristics of an ideal severity score, and practical limitations

Simple

Includes routinely recorded data

Limitations of demographic and physiological data

Easy to calculate

All systems require training at roll-out and later reinforcement.

Memorable or computer-based tool

Paper and computer systems are limited by availability

Observer independent

Consistency and reliability

Training is required for reliable physiological measurements

Functioning medical equipment is needed for some variables

Systematic

Comprehensively applied

Scores may be validated for unrealistically well-defined circumstances

Useful in varied populations

Dissimilar environments and populations require revalidation of existing scores to ensure utility

Specifically applied

Appropriately used in a validated population e.g. suspected pneumonia (CURB-65), gastrointestinal bleed (Blatchford Bleeding Score)

Disease specific scores are quickly unreliable where diagnoses are uncertain, unconfirmed, or over-generalised

Indicates a scale of response

Scores quantitatively reflect outcomes, or urgency. Linearity is ideal i.e. doubling the score indicates the patient is twice as ill

Most trigger scores are calibrated to “all or nothing” outcomes Triage systems are more finely graded and responsive but more complex

For triggering scores:

Trigger early

Early intervention is a key factor in improving outcome

Timely action in hospital systems requires significant human resources

Identifying patients too late to alter outcome is not clinically relevant

Trigger threshold in “Goldilocks” zone

Insensitive trigger misses the opportunities to act

Triggering too easily increases workload

High discrimination power is often practically unachievable

“Alarm fatigue” leads to reduced staff compliance with procedures