In a world where resources are finite, our ability to provide rescue services is limited by the number and types of resources we commit to other functions, such as fire suppression. In addition, we live in a nation where personal liberty is highly valued, which means we can't legislate risk away.

Finally, as a species we have a remarkable ability to come up with new and exciting ways to trap and/or injure ourselves. Simply put, for the foreseeable future, we'll need to allocate resources to rescue services carefully.

Because we must provide for the well-being of individuals, and it's extremely difficult to further one worthy objective without foregoing another, we must continuously balance competing interests. This is made more difficult because we're operating in an uncertain environment: We never know when or what the “big one” is going be.

It's clearly beneficial for a fire department to provide services such as technical rescue response, but these services aren't universally provided and it isn't clear that they're universally needed, which leads to a few questions. How do we determine what rescue services to provide? How do we provide them? What are the costs and benefits of providing a specific level of service? Answers to these questions hinge on our ability to handle uncertainty.

In dealing with uncertainty, it's helpful to use a rational process that identifies our goals and uses logical methods to explore the best way to meet these goals in light of the relative risks. In an uncertain environment, predicting the frequency and consequences of an event and valuing its outcomes depend on our analysis of risk.

Real risk is the actual combination of chance and loss in the real world. Because of the complexity of the universe, we'll never know what the real risk is for any given loss. Furthermore, we can't anticipate all of the potential loss scenarios. Therefore, when we think about the concept of risk, we typically think in terms of loss and uncertainty.

To understand risk, we begin by looking at two possible states of nature, loss and no loss, and the probability of being in either state. For example, if you flew on a commercial airline in 1999, there were two possible states that you could have arrived in — alive or dead — and the probability of arriving dead was 0.018 per 100,000 flights.

Figure 1 is a graphical representation of our theoretical model of risk for airline travel for a single flight. This is an example of the observed risk. That is, we have a simple theoretical model of the real world — arrive alive or arrive dead — and a measure of the probability of each of the possible states occurring.

Observed risk is the place to start when evaluating your rescue needs. To understand the observed risk, however, we need good data, and one of the greatest obstacles facing rescue-related policy-making is the lack of reliable data about rescue incidents and their related mortality and morbidity. The lack of rescue information in our incident reporting systems often handicaps our ability to assess our relative risks. Decisions regarding rescue services are often restricted, therefore, to insights based on limited information or the idiosyncratic experiences of other jurisdictions.

However, not all is lost. We can start collecting the appropriate data now and with a little effort, we can mine our existing data to better define the risk and need for a particular rescue service. For example, we can identify all of the automobile accidents where extrication was needed, and we can get an idea of the severity of the incidents by looking at how long our crews were on scene.

We can expand this concept to look at other rescue scenarios. We can tally how many water rescues we performed and whether or not they were swift-water rescues. What we're developing is data on the frequency of occurrences, a simple count of the number of times a given service is needed in a period of time.

What do you do when you have no response data available, but you know a risk to be present? Here we're dealing with perceived risk, which is an estimate of risk made in the absence of a theoretical model. Perceived risk tends to be subjective and can be based on a variety of sources: expert opinion, public perception or gut reaction. The problems associated with perceived risk are either underestimating or overestimating the risk.

We can make perceived risk more objective, however, by developing models of risks based on observations of other jurisdictions. For example, suppose your community is experiencing a growth spurt with new construction occurring rapidly. Although you haven't had any previous experience with trench rescues, the recent construction causes you concern. To determine your community's risk, you can look at aggregate incident data from other similar jurisdictions and count the number of trench rescues and number of building permits issued in those jurisdictions to build a model for your anticipated frequency of trench rescues.

A similar process can be used to estimate the risk of natural disasters. We can look at the historical frequency of natural events and project relative risks for our communities using expertise provided by climatologists, oceanographers and seismologists.

While estimating the probabilities of large-scale incidents is important, I want to emphasize the importance of thoroughly analyzing your response to “routine” emergencies. It's too easy to get stuck in the trap of preparing for the “big one” while the collective impact of routine emergencies goes unchecked. The absolute number of lives lost and dollar value of property lost tends to be the highest from the cumulative impact of routine emergencies, so be sure to consider incremental improvements to your existing services before taking on a whole new level of service.

Once we have an estimate of the probability of a particular rescue event happening, how do we go about determining the best course of action to mitigate or deal with the possible rescue scenarios? I'd like to suggest that you look at using decision trees.

Decision trees are flowcharts that can be used to calculate the “expected monetary value” of actions in an uncertain environment. These emvs are derived from best estimates of the probabilities of various outcomes and what the costs and benefits of those outcomes are.

Let's take trench rescue as an example. Suppose our trench-rescue frequency is one incident per 1,000 construction projects, with a fatality rate of 100%. In addition, we believe we can cut our fatality rate in half with a trench-rescue team. The cost of a trench-rescue team is $20,000 per year, and we anticipate the construction rate to be 200 projects per year.

Finally, we need to assign a value to the lives we'll save. According to litigation tables, the average life is worth $1.5 million. While assigning a dollar value to human life is distasteful, it's necessary when trying to decide which to choose among a variety of programs that save lives. Simply put, we need to maximize the number of lives we save, and only by assigning a value to human lives can we compare projects in monetary terms.

To determine the best way to go, we create a tree and determine the emv of each choice by multiplying the net costs/benefits by the probability of their occurrence and adding them to each other. In this example, the emv for the trench-rescue team is:

$-170,000 = (-[$1,500,000 + $20,000] ¥ 0.1) + ($-20,000 ¥ 0.9)

and the emv for not having a team is:

$-300,000 = ($-1,500,000 ¥ 0.2) + ($-0 ¥ 0.8).

Note that in each equation, the probability terms (0.1 and 0.9, or 0.2 and 0.8) add up to one, since we want to include all possibilities in the model. Since both of the answers are costs to the community, we choose the one with the lower cost, which is having a team. This is graphically represented in Figure 2.

What about mitigating the risk? Rather than create a trench-rescue team, we could have our building inspectors enforce shoring requirements instead of waiting for osha to do so. Suppose having local inspectors enforce shoring requirements would cost an additional $50 per project ($10,000 per year) and cut trench-rescue incidents in half.

In this case, the emv for the trench-rescue team is still $-170,000, while the emv for not having a team, but having more stringent enforcement is:

$-160,000 = ([$-1,510,000 ¥ 0.1] + [$-10,000 ¥ 0.9]).

Again, since both of these are costs to the community, we choose the one with the lower cost: better enforcement. We can also consider additional possibilities, such as trench-rescue capability plus tighter enforcement of shoring requirements. And in any of these examples, we can alter the outcome by changing the value we place on a human life.

While this process of evaluating uncertain outcomes may appear daunting on first glance, it's really quite simple. You can use this process to evaluate much more complex scenarios with hundreds of branches. To make this analysis easier, several companies make decision-tree software. (See sidebar.)

The process gives us a framework to discuss the relative merits and potential outcomes of different programs and ultimately to maximize the protection we provide to the community, in the most cost-effective ways.

For further information

• To learn more about decision trees, see Stokey, Edith and Richard Zeckhauser, “A Primer for Policy Analysis.” New York: W.W. Norton and Co., 1978.

• An online introduction to decision trees and other types of logic trees can be found at <www.echelon.ca/tonywild/logtree.htm>.

• Manufacturers of stand-alone decision-tree software include:

  • TreeAge Software Inc. <www.treeage.com> and
  • Vanguard Software Corp. <www.vanguardsw.com>.

• Two companies that make Microsoft Excel add-ons for creating decision trees are:

  • Decision Support Services <www.treeplan.com> and
  • Palisade Corp. <www.palisade.com>.

• For other articles on assessing the need for special rescue, see “Building specialized rescue capability on a budget,” November 1992; “Keeping up with the Joneses,” April 1997; and “A tale of two villages,” November 1998, all available at <www.firechief.com>.

A member of the fire service for 19 years, Kevin Klein owns Public Safety Consultants in Boulder, Colo., and manages the Colorado State Fire Chiefs' Association. He served as fire chief and chief financial officer for the Cherryvale Fire Protection District, Boulder, Colo., for eight years. A skydiver, Klein has lectured on analysis of risk for skydivers and on public safety doctrine development for the Republic of the Philippines. He has a master's in public administration from Harvard University.