| Abstract: | Successful identification of unnatural epidemics relies on a sensitive risk assessment tool designed for the differentiation between unnatural and natural epidemics. The Grunow–Finke tool (GFT), which has been the most widely used, however, has low sensitivity in such differentiation. We aimed to recalibrate the GFT and improve the performance in detection of unnatural epidemics. The comparator was the original GFT and its application in 11 historical outbreaks, including eight confirmed unnatural outbreaks and three natural outbreaks. Three steps were involved: (i) removing criteria, (ii) changing weighting factors, and (iii) adding and refining criteria. We created a series of alternative models to examine the changes on the parameter likelihood of unnatural outbreaks until we found a model that correctly identified all the unnatural outbreaks and natural ones. Finally, the recalibrated GFT was tested and validated with data from an unnatural and natural outbreak, respectively. A total of 238 models were tested. Through the removal of criteria, increasing or decreasing weighting factors of other criteria, adding a new criterion titled “special insights,” and setting a new threshold for likelihood, we increased the sensitivity of the GFT from 38% to 100%, and retained the specificity at 100% in detecting unnatural epidemics. Using test data from an unnatural and a natural outbreak, the recalibrated GFT correctly classified their etiology. The recalibrated GFT could be integrated into routine outbreak investigation by public health institutions and agencies responsible for biosecurity. |
