Estimating Risk

Minimum risk estimates are sometimes used to quantify either maximum exposure in monetary terms or, in the case of an annual work plan containing multiple projects, to help determine the proportion of firm projects. Firm projects are those which have budget cover even if costs overrun. A minimum risk estimate is one with little or no probability of overrun, and can be used to reflect the risk associated with very complex or novel projects. 

Risk Estimation References Health and Safety Executive, Canvey—An... 

Risk Estimation There are a number of risk measures which can be estimated. The specific risk measures chosen are generally related to the study objective and depth of study, and any preferences or requirements established by the decision makers. Generally, risk measures can be broken down into three categories risk indices, individual risk measures, and societal risk measures. 

QRA is fundamentally different from many other chemical engineering activities (e.g., chemistry, heat transfer, reaction kinetics) whose basic property data are theoretically deterministic. For example, the physical properties of a substance for a specific application can often be established experimentally. But some of the basic property data used to calculate risk estimates are probabilistic variables with no fixed values. Some of the key elements of risk, such as the statistically expected frequency of an accident and the statistically expected consequences of exposure to a toxic gas, must be determined using these probabilistic variables. QRA is an approach for estimating the risk of chemical operations using the probabilistic information. And it is a fundamentally different approach from those used in many other engineering activities because interpreting the results of a QRA requires an increased sensitivity to uncertainties that arise primarily from the probabilistic character of the data. 

The core function of QRA is to provide information for decision making. QRA results in and of themselves cannot prove anything. However, decision makers can compare QRA risk estimates to their own risk tolerance criteria to decide whether a plant or operation is safe enough. The same QRA results can support both the plant manager s contention that the plant is safe, as well as the community activist s claim that the plant is unsafe. The difference lies in the individual s risk tolerance, not the QRA. 

Relative risk results show only the difference between the levels of safety of one or more cases of interest and a reference, or baseline, case. Relative risk estimates can be used (as can absolute estimates) to determine the most efficient way to improve safety at a facility. But, the use of relative risk estimates alone does little to help ensure that the most efficient way is safe enough unless one of the cases meets qualitative safety criteria (e.g., compliance with relevant codes, standards, and/or regulations consistency with current industry practice). 

If both frequency and consequence values are calculated and reported on an absolute basis, then they may be reported graphically in combination with one another (Chapter 3), or simply as the product of frequency and consequence. Table 5 contains some examples of typical risk estimates (frequency and consequence products). Based on absolute risk estimates, you can decide whether the risk of a specific activity exceeds your threshold of risk tolerance (risk goal). If so, analysts can estimate the reduction in risk, given that certain improvements are made, assumptions changed, or operating circumstances eliminated. 

TABLE 7. Converting Absolute Risk Estimates to Relative Risk Estimates... 

System Absolute Risk Estimate Relative Risk Estimate... 

Whenever possible, relative comparisons of risk should be made (Step 8). Comparing relative risk estimates for alternative strategies avoids many of the problems associated with interpreting and defending absolute estimates. Table 9 contains examples of typical conclusions you can reach using relative risk estimates. In some cases, however, absolute estimates may be required to satisfy your needs. Table 10 contains a list of examples of typical conclusions possible using absolute risk estimates. 

TABLE 10. Examples of Possible Conclusions Using Absolute Risk Estimates... 

The QRA project team can select the appropriate technical approach once you specify the study objectives, and together you can define the scope. A variety of modeling techniques and general data sources (discussed in Section 3.2) can be used to produce the desired results. Many computer programs are now available to aid in calculating risk estimates, and many automatically give more answers than you will need. The QRA team must take care to supply appropriate risk characteristics that satisfy your study objectives—and no more. 

If there is a lack of specific, appropriate data for a process facility, there can be considerable uncertainty in a frequency estimate like the one above. When study objectives require absolute risk estimates, it is customary for engineers to want to express their lack of confidence in an estimate by reporting a range estimate (e.g., 90% confidence limits of 8 X 10 per year to 1 X 10 per year) rather than a single-point estimate (e.g., 2 X 10per year). For this reason alone it may be necessary for you to require that an uncertainty analysis be performed. 

The F-N curve, the risk profile, and the risk contour are the three most commonly used methods of graphically presenting risk results. Normally, you will elect to use more than one of these methods when evaluating risk estimates for decision making. 

Another way to evaluate risks is to calculate the sensitivity of the total risk estimates to changes in assumptions, frequencies, or consequences. Risk analysts tend to be conservative in their assumptions and calculations, and the cumulative effect of this conservatism may be a substantial overestimation of risk. For example, always assuming that short-term exposure to chemical concentrations above some threshold limit value will cause serious injury may severely skew the calculated risks of health effects. If you do not understand the sensitivity of the risk results to this conservative assumption, you may misallocate your loss prevention resources or misinform your company or the public about the actual risk. 

Risk sensitivity results are also very useful in identifying key elements in your existing loss prevention program. For example, suppose your fire protection system was assumed to have a very low probability of failure because you test it weekly. Fire protection failures may not show up as an important contributor to your total risk (because failure is so unlikely), but your total risk estimate may be extremely sensitive to any change in the probability of fire protection failures. Flence you should not divert resources away from testing the fire protection system unless the alternate use of funds will decrease risk more than the reduced testing will increase risk. 

Absolute risk estimates can be difficult to use when there is no apparent human experience against which to calibrate them. By definition, there never exists enough experience about catastrophic rare events (fortunately) with which to calibrate the thinking about their significance. If there were enough data, you would not have elected to do the QRA in the first place. So, now thatyou have a bottom line estimate... 

Consider the following example in which the worker risk from a catastrophic accident has been calculated to be 2 X 10 fatalities per year. It is possible to interpret this number in many ways, but one of the most common ways is the following there is one chance in 5000 per year that a worker will be fatally injured at the plant. However, you should be cautious when interpreting single risk estimates that are the sums of products of frequency and consequence of many accidents. The way you believe (and act) may be affected by the frequency/consequence profile that the number represents (see Sections 3.2.4 and 4.2.5.) That is, your reaction to an accident that occurs once every 100 years and kills 1 person (Risk = 10 fatalities per year) and your reaction to an accident that occurs once every 10,000 years and kills 100 people (Risk = 10 fatalities per year) are likely to be very different. 

There are several widely used approaches for developing perspective about the significance of absolute risk estimates (Figure 16). The first approach is to compare the risk estimates to historical experience within your company, looking for similar events. Most companies have safety and loss recordkeeping programs that date back many years. But if directly related data are sparse, you may widen your comparison to extrapolate from near-miss incidents that could have caused the event of interest. You will not, however, frequently find solace from the company data—or even comparable industry data. 

Another approach is to use government and private mortality and injury statistics. Calculated absolute risk estimates (the probability per year of a worker being injured or killed) can be compared to those de facto worker risk standards. For example, in the United Kingdom, industry and government alike are using the fatal accident rate (FAR, see Glos-... 

FICURE 16. Means of establishing perspective with absolute risk estimates. 

Another way of interpreting absolute risk estimates is through the use of benchmarks or goals. Consider a company that operates 50 chemical process facilities. It is determined (through other, purely qualitative means) that Plant A has exhibited acceptable safety performance over the years. A QRA is performed on Plant A, and the absolute estimates are established as calibration points, or benchmarks, for the rest of the firm s facilities. Over the years, QRAs are performed on other facilities to aid in making decisions about safety maintenance and improvement. As these studies are completed, the results are carefully scrutinized against the benchmark facility. The frequency/consequence estimates are not the only results compared—the lists of major risk contributors, the statistical risk importance of safety systems, and other types of QRA results are also compared. As more and more facility results are accumulated, resources are allocated to any plant areas that are out of line with respect to the benchmark facility. 

Having a numeric criterion for tolerable risk would be everyone s choice when making decisions using absolute risk estimates. Unfortunately, no universally accepted or mandated criterion exists. Nevertheless, when attempting to establish risk guidelines satisfying the requirements described, an organization has a number of resources avail-able. 3 Some particularly valuable sources of information include ... 

However, because of the diversity of short-term and long-term hazards to the public and workers in the CPI, no single criterion will ever meet everyone s needs. Even if such a criterion could be developed, there would still be controversy over its use. Absolute risk estimates are not... 

In communicating risk information, trust and credibility are imperative. If you do not know an answer, say so, then get back to those people when you do have an answer. Discuss data uncertainties, strengths, and weaknesses, including ones identified by other credible sources. Identify worst-case estimates as such, and cite ranges of risk estimates when appropriate. 

When managers are faced with the necessity of using QRA results on an absolute basis, they must respect the potentially large uncertainties associated with the numbers and use prudent and conservative interpretations of these results for their decisions. Absolute risk estimates in... 

The process by which the results of a risk analysis (i.e., risk estimates) are used to make decisions, either through relative ranking of risk-reduction strategies or through comparison with risk targets... 

---