Use of environmental parameters to model pathogenic Vibrios in chesapeake bay

Annual reports show that human infections caused by Vibrio spp. have nearly doubled over the past decade in the Virginia and Maryland waters of the Chesapeake Bay. Vibrio spp. are autochthonous to estuarine and coastal waters and follow a seasonal cycle attributed mainly to fluctuations in water temperature and salinity. This study presents the development of empirical algorithms for predicting the probability of Vibrio vulnificus and Vibrio parahaemolyticus likelihood and abundance in the upper Chesapeake Bay. To model likelihood of occurrence, a set of binary classification models was developed, employing a suite of geophysical predictor variables and statistical methods. Accuracy of results was ~ 68% at 0.40 prediction for V. vulnificus and ~ 70% at 0.60 prediction for V. parahaemolyticus. For Vibrio spp. abundance, regression methods were applied to samples positive for Vibrio, showing Vibrio abundance can be predicted as a function of sea surface temperature and salinity in Chesapeake Bay, with mean absolute error (MAE) of 3.9 cells 10 ml-1 for V. vulnificus and 5.8 cells 10 ml-1 for V. parahaemolyticus. Additionally, for the purpose of operational potential in the Chesapeake Bay, we developed a two-step classification/regression hybrid approach was used to generate estimates of abundance in the absence of bacteriological data on presence of Vibrio spp. This hybrid approach predicted Vibrio abundance with MAE of 2.8 cells 10 ml-1 for V. vulnificus and 4.4 cells 10-1 ml for V. parahaemolyticus. Since the risk of human infection is a function of Vibrio spp. pathogenicity and abundance, extending available predictive modeling capabilities to provide concentration, in addition to presence/absence, advances the public health utility of these models significantly.