Comparing Directionally Sensitive MCUSUM and MEWMA Procedures with Application to Biosurvei.pdf

Comparing Directionally Sensitive MCUSUM and MEWMA Procedures with Application to Biosurveillance Ronald D. Fricker, Jr., Matthew C. Knitt, and Cecilia X. Hu Naval Postgraduate School June 18, 2007 Abstract This paper compares the performance of two new directionally-sensitive multivariate methods, based on the multivariate CUSUM (MCUSUM) and the multivariate exponen- tially weighted moving average (MEWMA), for syndromic surveillance. While neither of these methods is currently in use in a biosurveillance system, they are among the most promising multivariate methods for this application. Our analysis is based on a detailed series of simulations using synthetic syndromic surveillance data that mimics various types of disease background incidence and outbreaks. We apply the MCUSUM and the MEWMA to residuals from an adaptive regression that accounts for the systematic effects normally present in syndromic surveillance data. We find that, similar to results for the univariate CUSUM and EWMA in classical statistical process control applications, the directionally- sensitive MCUSUM and MEWMA perform very similarly. 1 Introduction Biosurveillance is the process of monitoring health data in order to assess changes in disease incidence. Traditional methods have been focused on retrospectively analyzing medical and public health data, such as hospital admittance or mortality rates, to determine the existence of a disease outbreak (Shmueli, 2006) and/or to conduct epidemiological investigations (Stoto, 2007). Via traditional biosurveillance, the pro