John D Bullock1. 1. Department of Ophthalmology, Wright State University Boonshoft School of Medicine, Dayton, OH 45420-4006, USA. john.bullock@wright.edu
Abstract
PURPOSE: The Poisson distribution can be used to model discrete events that occur infrequently in time, and it was applied, retrospectively, to recently reported clusters of Fusarium keratitis. METHODS: This distribution was utilized with data reported from 6 geographically diverse ophthalmic centers during the worldwide Fusarium keratitis epidemic of 2004-2006. The expected numbers of cases, determined from published historical data, were compared with the observed and probabilities were calculated. The Poisson probabilities were also used to determine the numbers of cases in each cluster required to reach statistical significance. RESULTS: The probabilities that the numbers of cases observed in the various clusters were a chance variation from the expected, and not due to an outbreak, were between 3.83 x 10(-5) and 4.72 x 10(-47). The recognition of the first 2 or 3 cases in each cluster was sufficient to establish that an outbreak was present. CONCLUSIONS: Because these probabilities are all much less than 0.05, multiple Fusarium keratitis outbreaks have thus been documented. Future use of the Poisson distribution may serve as an "early warning system" that a new, potentially modifiable, factor may be operating to increase the occurrence of a disease above its historic baseline endemic rate.
PURPOSE: The Poisson distribution can be used to model discrete events that occur infrequently in time, and it was applied, retrospectively, to recently reported clusters of Fusarium keratitis. METHODS: This distribution was utilized with data reported from 6 geographically diverse ophthalmic centers during the worldwide Fusarium keratitis epidemic of 2004-2006. The expected numbers of cases, determined from published historical data, were compared with the observed and probabilities were calculated. The Poisson probabilities were also used to determine the numbers of cases in each cluster required to reach statistical significance. RESULTS: The probabilities that the numbers of cases observed in the various clusters were a chance variation from the expected, and not due to an outbreak, were between 3.83 x 10(-5) and 4.72 x 10(-47). The recognition of the first 2 or 3 cases in each cluster was sufficient to establish that an outbreak was present. CONCLUSIONS: Because these probabilities are all much less than 0.05, multiple Fusarium keratitis outbreaks have thus been documented. Future use of the Poisson distribution may serve as an "early warning system" that a new, potentially modifiable, factor may be operating to increase the occurrence of a disease above its historic baseline endemic rate.