Alerts B: Thresholds 1-4
Peak comparisons to alert with varying thresholds.
the basis of timing alone, as these may identify an unusually early seasonal
increases or the emergence of an unknown agent.
Algorithms will not replace the need for local knowledge and experience.
The definition of an outbreak requires the description and interpretation of
relationships between cases, and algorithms will only alert one to the possi-
bility of such an event based on past experience. O'Brien and Christie (1997)
explained that the investigator must decide whether changes might be real
or artifactual, and that algorithms may therefore be regarded as an adjunct
to other methods of interrogating surveillance data.
As with peak comparison, correlations produce a preliminary measure
of potential. In some studies, authors used the cross-correlation function as
part of an initial, exploratory analysis (Lazarus et al. 2002; Lewis et al. 2002;
Sebastiani et al. 2006) while for others, the result of the correlation analyses
was the main outcome (Davies and Finch 2003; Hogan et al. 2003; Najmi and
Magruder 2004; Tsui et al. 2002).
Correlation is not biased by algorithm selection and provides a measure
of the relationship between two data sources. However, a limitation of
this method is that it is sensitive to large variations in the amplitude of the
time series (Magruder 2003), such as long and short wavelengths (Bloom,
Buckeridge, and Cheng 2007). Bloom and colleagues (2007) suggested that
researchers should stipulate the feature scale length of interest and filter
the data appropriately prior to the application of the CCF or the results may
be ambiguous and misleading. CCFs are useful in showing that one data
source is more timely than another; however, they cannot be used to define a
change or level that might be indicative of a disease occurrence or outbreak
as with aberration detection (Suyama et al. 2003).