# The LWF Blog

## Fire Engineering for Healthcare Premises – Sensitivity Analysis – Part 58

October 25, 2021 12:50 pm

In LWF’s blog series for healthcare professionals, our aim is to give information on best practice of fire safety in hospitals and other healthcare premises. In part 57 of Fire Engineering for Healthcare Premises, LWF discussed the use of risk assessment in fire safety engineered healthcare environments. In part 58, we look at the use of sensitivity analysis and worst-case scenarios.

### Sensitivity Analysis

A fire engineering calculation depends on many variables, all of which may have some degree of uncertainty in their values. For the outcome to be relied upon, it may be necessary to undertake a sensitivity analysis (SA). A SA compares the outcomes of many calculations, altering one variable to enable a fuller understanding of the impact, for example, the fire growth rate could be altered, as an uncertain value from medium to fast, to ascertain the impact on the outcome.

The range of values tried for a given variable should depend on the level of uncertainty. If the outcome is not sensitive to the value of a given variable, it does not matter if that variable’s value has large uncertainty. Conversely, where the outcome is sensitive, measures should be taken to lower the level of uncertainty, even where it is already low.

The outcome may be a highly non-linear function of the variable’s values, therefore, just considering two or three possible values for a variable may omit regions where the sensitivity is high.

The robustness and reliability of decisions from a fire-engineered analysis are dependent on effective sensitivity analysis.

### Worst-case Scenarios

When undertaking a deterministic study (as opposed to a full probabilistic risk assessment), assumptions relating to fire size and growth rate, building occupancy etc., tend to be conservative, so as to define a worst-case scenario.

The consequences of the worst-case scenario can then be calculated and compared with the assessment threshold criteria.

This process results in a system which is designed to perform in a worst-case scenario and therefore, should perform to a required level in less severe scenarios; more severe scenarios are assumed to have minimal probability.

As the consequences of the scenarios are not known prior to calculation, it may be necessary to define several worst-case candidates and evaluate each. If extreme initial assumptions were made, this would be over-conservative – and the resulting scenario would have a very low probability of ever occurring. However, if typical or average values were used, it would result in a scenario that was not conservative enough.

Each scenario must be self-consistent. Limiting fire size in a sprinklered building would no longer be appropriate if the sprinklers were to fail, for example. A sensitivity analysis can estimate the consequences of uncertainties in the scenario, variable values etc.