Biology Reference
In-Depth Information
The following sections outline the methods for describing timeliness, pro-
vide examples from the literature, and identify the strengths and limitations.
Many of the case studies presented are based on the surveillance of influ-
enza. This disease has been used as a model infection for studies on respira-
tory syndrome and provides regular outbreaks to study.
1.2 Peak Comparison
Peak comparison involves calculating the temporal distance between the
peaks or maximal values observed in data sources, and considers them in
relation to one another (see Box 1.1).
Peak comparison has been used to measure the time difference between
two data sources (Davies and Finch 2003; Lenaway and Ambler 1995; Proctor,
Blair, and Davis 1998; Welliver et al. 1979). Welliver and colleagues (1979) com-
pared the weekly percentage change in nonprescription cold remedy sales in
a Los Angeles supermarket chain to the proportion of positive influenza iso-
lates from children presenting to a pediatric hospital. They found that sales of
nonprescription cold remedies peaked 7 days earlier than the peak in virus
isolation in one season (Welliver et al. 1979). It is unknown whether this time
difference would exist today if the study were repeated using modern labora-
tory methods.
Lenaway and Ambler (1995) analyzed a school-based influenza surveil-
lance system that included 44 schools in a Colorado county. This study com-
pared five influenza seasons and found varied results. In two of the 5 years,
absenteeism surveillance peaked 7 days earlier than sentinel surveillance.
For the other 3 years, there was no time difference between peaks (Lenaway
and Ambler 1995). Hence, over the 5-year period, the peak in school-based
absenteeism occurred on average 2.8 days earlier than the peak in ILI senti-
nel influenza surveillance.
One study investigated retrospective peaks associated with the 1993
Milwaukee
Cryptosporidiosis
outbreak (Proctor, Blair, and Davis 1998). The
first data source to peak was turbidity measurements of the water treatment
plant's effluent. This source was used as the reference for calculating timeli-
ness (see Table 1.1).
Customer complaints and rates of diarrhea among nursing home residents
were the most timely data sources with lags of 2 and 11 days, respectively.
The peak in emergency department presentations relating to gastrointestinal
illness and laboratory-confirmed cases of
Cryptosporidium
followed 15 days
beyond the reference date. Surrogate morbidity peaks based on telephone
surveys (35 days) and school absenteeism (64 days) were the least timely. In a
real-time surveillance system, data sources are only timely if the information
