Cover Image


View/Hide Left Panel

by the committee’s definition almost everyone in the U.S. population is a stakeholder in BTRA information, it is important to develop strategies to reconcile differences among subpopulations. These subpopulations will perceive risk on the basis of their own goals and objectives. Techniques such as value-tree analysis (von Winterfeldt, 1987) may be useful in bringing out and reconciling these differences.

Risk Assessment

Risk assessment is the process of identifying hazards and targets and quantifying the risks that the hazards pose (magnitude, spatial scale, duration, and intensity) and the associated probabilities, including the uncertainties surrounding these estimates.2 The primary goal of risk assessment is to produce information to improve risk management decisions by identifying and quantifying cause-and-effect relationships between alternative risk management decisions and their consequences and by identifying decisions that may increase the probabilities of preferred outcomes. Risk assessment may include a description of the cause-and-effect links between different hazards, and the nature of the interdependencies, vulnerabilities, and consequences.

Once the problem has been formulated, risk assessment begins with hazard identification: the process of specifying the scope of the assessment and summarizing the available empirical evidence showing that a specific “hazard” (such as exposure to a specific pathogen in a specific environment) causes specified adverse health effects. Hazard identification can serve the following purposes:

  • Rapid screening of potential hazards by identifying whether available data support the hypothesized relationship between the hazard and specific health effects, possibly using formal statistical methods of causal analysis (Shipley, 2000);

  • Identification of causal relationships between identified hazards and specific adverse human health effects; and

  • Identification of risk factors, behaviors, and exposure conditions that increase risks to specific exposed populations (e.g., the old, the young).

Studies to identify specific hazards, their probability of occurrence, and the probability of occurrence of their associated consequences are a part of risk assessment. In these studies, experts can provide insight into terrorists’ values and objectives—along with their assessments of associated risks—but the experts need to take special care not to filter these estimates through their own values.

Health risk assessments are specializations of the methods described above. They typically use explicit analytic models (e.g., statistical models, probabilistic simulation) of causal relationships between actions and their probable health effects. Exposure models describe the transport and distribution of hazardous materials through different media and pathways (e.g., air, foods, drinking water) leading from their source(s) to members of the exposed population. Because different exposures lead to different health outcomes, a successful exposure assessment should describe the frequency distribution of exposures of different parts of the population.

Dose-response models ideally quantify the conditional probability of illness caused by each level of exposure as well as the degree of uncertainty surrounding these estimates. For some biological agents, it may be necessary to fit separate dose-response models to “normal” and “susceptible” subpopulations at risk and to account for interindividual variability in dose-response relations. In general, risk assessment requires a description of the severities as well as the frequencies of adverse health outcomes caused by exposures and the potential value of gathering additional information to reduce the uncertainty surrounding these risk estimates.

One useful graphical way to capture the extent of expert knowledge about a particular risk is to construct an exceedance-probability (EP) curve. An EP curve specifies the probability that a certain level of losses will be exceeded. The losses can be measured in terms of dollars of damage, fatalities, illness, or some other unit of analysis. If one views the loss as a random variable, the EP is simply the complementary cumulative distribution of the loss.

For example, suppose one were interested in constructing an EP curve for direct dollar losses from the first bioterrorism attack described in the aerosol anthrax scenario employed in this report (see Chapter 1). Event trees and fault trees,3 used as part of probabilistic risk assessments, would identify the set of conditions and subsequent events that could produce a given dollar loss, determine the resulting probabilities of exceeding losses of different magnitudes, and combine the results. Based on these estimates, the mean EP curve, depicted in Figure 2.1, could be constructed. Suppose that one focuses on a specific loss, Li. One can see from Figure 2.1 that the likelihood that losses will exceed Li is given by pi. The x axis measures the loss in dollars and the y axis depicts the probability that losses will exceed a particular level.4

It is much easier to construct an EP curve for natural disasters and chemical accidents than for bioterrorist activities. But even for those more predictable accidents or disasters, there may be considerable uncertainty regarding the occurrence of certain risks and the resulting damage. Providing information on the range of this uncertainty asso-


See Haimes (1998) for a comprehensive summary of recent work in risk assessment.


See the lexicon in Appendix A for definitions of event tree and fault-tree analysis.


A detailed discussion of how one constructs an EP curve and incorporates elements of uncertainty on these estimates appears in Grossi and Kunreuther (2005, Chapters 2 and 4).

The National Academies | 500 Fifth St. N.W. | Washington, D.C. 20001
Copyright © National Academy of Sciences. All rights reserved.
Terms of Use and Privacy Statement