Predictive input parameters for enteric virus presence at the inlet of a potable water supply

Kentucky river water samples were collected at the input of a potable water supply weekly, for a period of two years. Eighteen parameters were analysed and/or recorded for each weekly observation. The variables were of a physical, chemical and biological nature. Enteric virus presence was detected by cell culture analysis from approximately 200 L samples of river water. A database of 108 observations was assembled. Of these only 100 complete observations were used for the modelling analysis. Model input parameters were selected solely based on their statistical significance in forward, backward and stepwise selection procedures. Multivariate logistic regression (MLR) was used to determine the relative significance of the input parameters for predicting the presence or absence of enteric virus. The final, best-fit MLR model that used log-transformed values for enterococci, the AC/TC ratio, faecal coliforms (FC) and temperature demonstrated 82.5 and 81.4% sensitivity and specificity, respectively.