Risk Epidemiology

The standard process of risk analysis, as described above, may use data from a number of sites, for example on the failure rates of specific fittings or equipment, but the results of the overall risk analysis process are usually quite site-specific. Recently, it has been suggested that standardized data from across a large number of sites could be used to provide more aggregate assessments of accident rates and severities. This approach, using data from a number of facilities, has been called accident (or risk) epidemiology, since it is similar to its medical cousin in attempting to use cohort cross-sectional or panel data to determine underlying factors of undesirable outcomes.

The best known example of this type of study is associated with the Risk Management Plan Rule under section 112(r) of the Clean Air Act Amendments in the USA, which is now briefly described. In the European Union, similar actions were undertaken under the Seveso II Directive (Orts and Deketelaere 2000). Section 112(r) sets forth a series of requirements aimed at preventing and minimizing the consequences associated with chemical accidental releases. These requirements are the basis of the US Environmental Protection Agency's rule on 'Risk Management Programs for Chemical Accidental Release Prevention' (hereafter the 'Rule'). The federal regulations promulgated under 112(r) apply to facilities (both public and private) that manufacture, process, use, store or otherwise handle regulated substances at or above specified threshold quantities (which range from 200 to 10000 kilograms). With some exceptions, the Rule requires all regulated facilities to prepare and execute a risk management program (RMP) which must describe the facility's approach to hazard assessments, worst-case scenarios and prevention programs. The Rule also specifies the requirement that regulated facilities maintain a five-year history of accidental releases and submit this history to the EPA no later than 21 June 1999, with further five-year filings expected in the future (probably repeated in 2004). The number of facilities that initially filed under the Rule was 14 500, with 1145 of these facilities (7.9 per cent) reporting 1913 accidents over the five-year period of interest. The database itself has been named RMP*Info and, except for sensitive data on worst-case scenarios, has been available to the public since August 1999.

The initial data were first analyzed in Kleindorfer et al. (2000). The basic approach followed was the epidemiologic methodology known as (retrospective) cohort study design. Epidemiology is the study of predictors and causes of illness in humans. Its use in studying industrial accidents had been proposed in a number of quarters (for example, Saari 1986; Rosenthal 1997). The motivating idea is to study the demographic and organizational factors of those facilities whose accident histories are captured in RMP*Info to determine whether any of these factors have significant statistical associations with reported accident outcomes, positive or negative, just as one might use demographic or lifestyle data for human populations to determine factors that might be associated with the origin and spread of specific illnesses.

To provide some sense of what accident epidemiology is, consider the basic demographics of the facilities that filed under RMP*Info. There are 14 500 facilities in RMP*Info and there are 1913 reported accidents in RMP*Info, with 1145 facilities reporting at least one accident. However, the sample size for various statistics will not remain constant at 14 500 and 1913, since some sites have multiple processes and some processes use multiple listed chemicals. For example, the average facility size among facilities reporting to RMP*Info, as measured in employee FTEs (full time equivalents), is 163 FTEs, ranging from facilities with less than 0.5 FTEs (recorded as 0 FTEs in RMP*Info) to 48 000 FTEs. Half of the facilities have 11 FTEs or fewer.

Various accident rate statistics can be computed from the information in RMP*Info. For example, Table 37.1 reports accidents by listed chemical involved in the accident for chemicals involved in 10 or more accidents. The numbers of accidents by chemical ranged

Table 37.1 Accidents reported in RMP*Info by chemical involved in the accident, 1994-9 (results for chemicals with 10 or more accidents during the five-year reporting period)

Chemical name


Number of

Was this article helpful?

0 0

Post a comment