Abstract
To reduce damage caused by chemical accidents, it is important to establish a prevention system for chemical accidents. The first step in the prevention of chemical accidents is to screen the high-risk chemical plants. Risk index, one of the screening methods, can indirectly estimate the risk at each chemical plant. For calculating the risk index, the probability of an accident needs to be estimated, which requires complex calculation and confidential data from plants that are difficult to obtain. Therefore, we developed a new index, the accident probability index, to estimate accident probability in chemical plants using readily accessible data. We conducted a literature survey on the existing risk indices and interviewed chemical experts and government chemical managers to select surrogate indicators related to chemical accidents, and four indicators were chosen: hazardous characteristics of chemicals, handling volume, records of accident frequency, and national accident frequency of chemicals. We calculated the accident probability index for 4520 chemical plants, and index value means was 5.324 (95% confidence interval (CI): 3.156, 7.493). An increase by 10 in the index value denoted a 1.06-fold (95% CI: 1.04, 1.08) increase in the odds ratio for actual accident occurrence. The accident frequency of the fourth quartile of the index value was 4.30 times (95% CI: 1.72, 10.75) higher than those of the first quartile.