Forecasting incidence of extended spectrum β-lactamases-producing Escherichia coli by multiple seasonal ARIMA model / 预防医学

ABSTRACT

Objective To predict monthly incidents of extended spectrum β-Lactamases (ESBLs)-producing Escherichia coli in Zhejiang Hospital by establishing multiple seasonal autoregressive integrated moving average (ARIMA) model, so as to provide scientific evidence for reducing the incidents of nosocomial infection of ESBLs producing Escherichia coli. Methods Multiple seasonal ARIMA model was established by monthly records of ESBLs producing Escherichia coli from 2010 to 2016 in Zhejiang hospital. Monthly incidents of ESBLs producing Escherichia coli from 2017 to February 2018 were used to verify the predicted result. The predictions were evaluated by models of mean absolute percent error (MAPE) and bayesian information criterion (BIC) . Results The optional model for the monthly incidence from 2010 to 2016 was ARIMA (0, 1, 1) (0, 1, 1)12. The MAPE was 14.76, BIC was 2.01, and the Ljung-Box statistics value Q was 16.79 (P=0.40) . These parameters suggested a good model fitting. The average relative error between the predictive value and the actual value of the monthly incidents ESBLs producing Escherichia coli from 2017 to February 2018 was 14.08%.The actual values were within the 95% confidence interval. Conclusion The multiple seasonal ARIMA model of ARIMA (0, 1, 1) (0, 1, 1 )12 fits and can be used for short-term prediction and dynamic analysis of the incidents of ESBLs producing Escherichia coli in Zhejiang Hospital.