Year: 2014 Vol.: 63 No.: 2
Authors: Lorraine Christelle B. Angkico; Priscilla A. Diaz; Robert Neil F. Leong; Frumencio F. Co
This paper aims to explore early outbreak detection methods for measles. Two methods adopted from statistical process control were modified and used to fit biosurveillance, namely Shewhart and Exponentially Weighted Moving Average (EWMA) charts. Seven variations of such control charts are proposed: two under Shewhart chart (normal-based and zero-inflated Poisson (ZIP)-based) and five under EWMA charts (?s of 0.05, 0.10, 0.15, 0.20, and 0.25). To study the proposed charts, daily counts of laboratory confirmed cases of measles in the National Capital Region from 2009 until 2014 were utilized to characterize both the disease background and outbreak equations. During this time span, three measles outbreaks have transpired. The proposed charts, set at average time between false signals (ATFSs) of both one and two months, were evaluated and compared using performance metrics such as conditional expected delay (CED), proportion of true signals (PTS), proportions of detections in an outbreak (PDO), and probability of successful detection (PSD), computed from 500 sets of simulated data. It was found that ZIP-based Shewhart and EWMA with a ? of 0.05 work best for ATFSs of one and two months, respectively. Health-governing bodies may seek to explore the possible utilization of these charts to improve measles surveillance.
Keywords: control charts, measles, early event detection, biosurveillance