|
|
Research
Estimating Foodborne Gastroenteritis,
Australia
Gillian Hall,*
Martyn D. Kirk,† Niels Becker,* Joy E. Gregory,‡ Leanne Unicomb,§ Geoffrey
Millard,¶ Russell Stafford,# Karin Lalor,‡ and the OzFoodNet Working Group
*Australian National University, Canberra, Australian Capital Territory
(ACT), Australia; †OzFoodNet, Canberra, ACT, Australia; ‡Department of
Human Services, Melbourne, Victoria, Australia; §Hunter Population Health
Unit, Wallsend, New South Wales, Australia; ¶ACT Analytical Laboratory,
Weston Creek, ACT, Australia; and #Queensland Health, Archerfield, Queensland,
Australia
Appendix
Data and
Adjustment Factors
Raw data sources and adjustments required are shown in Appendix
Table 1.
Adjustment for Underreporting
In the calculations for determining the underreporting factor from the
community, the sensitivity of laboratory tests was estimated to be ≈90%
(on the basis of estimates for Salmonella and Shigella testing
in Australia), and laboratory reporting was estimated at 100%. The underreporting
factors were based on information from various sources as outlined in
Appendix Table 2.
Adjustment for Overseas-acquired
Cases
Notifiable data from the states of Victoria and South Australia were
used to estimate the proportion of cases that were acquired overseas for
different pathogens as shown in Appendix Table 1.
The relevant proportion of cases was subtracted from the estimates.
Adjustment for Underreporting
of Outbreaks Compared with Surveillance
The outbreak factor describes the relationship between the number of
cases identified in outbreaks, and the number of cases that would have
been identified by surveillance had the disease been a notifiable illness.
Since the pathogens of interest are not actually reported to surveillance,
the outbreak factor was based on Salmonella data that were reported
to both surveillance and a database of outbreaks in Victoria from 1998
to 2002. The plausible distribution of the outbreak factor was deduced
from a comparison of the number of Salmonella notifications and
the number of Salmonella cases identified in outbreaks.
On average, we found 14 times as many notifications as cases identified
in outbreaks in Victoria, with variability each year. The outbreak factor
was simulated as a normal distribution with a mean of 14 and a standard
deviation of 4. The illness severity for the 3 pathogens, based on outbreak
data, was moderate, so the underreporting factor for moderate illness
was used.
Simulation Technique to Account
for Uncertainty
Wherever uncertainty existed for a factor used in the calculations, a
simulated distribution of plausible values was used to model the uncertainty
in that factor, rather than a point estimate. In the absence of definitive,
statistically sound data, decisions about the plausible distribution of
values are based on a reasonable interpretation of real data. In other
words, the parameters of the plausible distribution are not necessarily
based on a statistically derived value but on judgments guided by the
best available data. The different distributions are used to simulate
1,000 plausible values, and the most likely values have the greatest frequency.
The median was used as the central estimate, and the credible interval
was taken as the 2.5–97.5 percentile of the simulated distributions. The
width of the credible interval of the final estimate is therefore determined
by the precision with which each of the component probabilities is estimated.
An example of the simulation technique is shown in the Appendix
Figure. It shows the estimation of the number of community cases of
campylobacteriosis by using Notifiable Diseases Surveillance data, leading
to the estimate of the number of foodborne cases of Campylobacter
infection for Australia.
|
Appendix Table 1. Sources
of raw data and adjustments for each microorganism
|
|
|
Source/pathogen
|
Raw data
|
|
|
Water Quality Study 1998–1989, Melbourne
|
No. positive/no. stools tested
|
|
E. coli, other (pathogenic,
non-STEC)†
|
53/791
|
|
Calicivirus†
|
75/703
|
|
Rotavirus†
|
11/791
|
|
Astrovirus/adenovirus†
|
9/791
|
|
Cryptosporidium parvum†
|
13/791
|
|
Giardia lamblia†
|
20/791
|
|
Queensland and South Australian laboratory data,
various years 1995–2001
|
No. positive/no. stools tested
|
|
Aeromonas spp. †
|
248/107,600
|
|
Vibrio parahaemolyticus†
|
2/30,880
|
|
Victorian outbreaks, 4-year data, 1998–2001
|
Mean no. cases/y (range)
|
|
Bacillus cereus‡§
|
12 (0–37)
|
|
Clostridium perfringens‡§
|
60 (28–73)
|
|
Staphylococcus aureus‡§
|
15 (0–40)
|
|
National Notifiable Diseases Surveillance System,
5-year data, 1996–2000
|
Mean no. cases/y (range)
|
|
Campylobacter spp. (AOA
4%)‡§
|
Not including NSW, 12,756 (11,829–13,528)
|
|
Salmonella spp. (AOA 8%)§
|
6,801 (5,791–7,712)
|
|
Shigella spp. (AOA 40%)§
|
626 (487–797)
|
|
STEC (AOA 21%)‡§
|
3-y data, South Australia only, 37
(18–51)
|
|
Yersinia spp. (AOA 2%)§
|
157 (74–212)
|
|
|
*STEC, Shiga toxin–producing Escherichia coli;
AOA, adjusted for overseas acquired; NSW, New South Wales.
|
|
†Proportion applied to estimate of total gastroenteritis
in Australia.
|
|
‡Population factor applied.
|
|
§Underreporting factors for outbreaks and moderate
illness applied.
|
|
Appendix Table 2. Underreporting
factors for moderate, bloody, and serious illness
|
|
|
Severity of illness
|
Information used for estimation
|
Estimate (credible interval)
|
|
|
Moderate illness
|
Australian surveillance data, Melbourne
Water Quality Survey (17), Victorian
Salmonella outbreak data, National Gastroenteritis Survey,
Hunter Salmonella case-control study (24)
|
1 in 15 (5–25)
|
|
Bloody diarrhea
|
National Gastroenteritis Survey
|
1 in 9 (1–17)
|
|
Serious illness
|
Mead et al., 1999 (6)
|
1 in 2 (1–3)
|
|
|
