Glossary and Acronyms - Department of HealthFile/wFoodborne-illness-Australi… · Web viewBy applying this proportion of episodes due to food to the ... which is one of the most

FOODBORNE ILLNESS IN AUSTRALIA:

ANNUAL INCIDENCE CIRCA 2010

Foodborne illness in Australia: Annual incidence circa 2010

Authors: Martyn Kirk, Kathryn Glass, Laura Ford, Kathryn Brown and Gillian Hall,

National Centre for Epidemiology and Population Health, Australian National University.

Prepared for the Australian Government Department of Health, New South Wales Food

Authority and Food Standards Australia New Zealand by the National Centre for

Epidemiology and Population Health, Australian National University.

Online ISBN: 978-1-74186-170-9

Publications approval number: 10768

Copyright Commonwealth of Australia 2014

This work is copyright. You may download, display, print and reproduce the whole or part of this work in unaltered form for your own personal use or, if you are part of an organisation, for internal use within your organisation, but only if you or your organisation do not use the reproduction for any commercial purpose and retain this copyright notice and all disclaimer notices as part of that reproduction. Apart from rights to use as permitted by the Copyright Act 1968 or allowed by this copyright notice, all other rights are reserved and you are not allowed to reproduce the whole or any part of this work in any way (electronic or otherwise) without first being given the specific written permission from the Commonwealth to do so. Requests and inquiries concerning reproduction and rights are to be sent to the Communication Branch, Department of Health, GPO Box 9848, Canberra ACT 2601, or via e-mail to [email protected].

Foodborne illness in Australia circa 2010i

GLOSSARY AND ACRONYMS

Asymptomatic

GBS

CI

CrI

DALY

Delphi process

Foodborne illness

HUS

IBS

ICD-10-AM

IID2

Incidence

Monte Carlo simulation

Notifiable

NGSI

NGSII

A state where a person who is infected does not show any symptoms.Guillain-Barré syndrome – a disorder where the body’s immune system attacks the peripheral nervous system and may be the result of a preceding infectious event.Confidence interval – represents a range of values that act as good estimates for an unknown parameter using a frequentist distribution. Credible interval – represents a range of values where the most likely estimate might lie using a posterior probability distribution. It may be interpreted similar to confidence intervals.Disability adjusted life year – a metric to describe the burden of disease that takes into account the morbidity and mortality of a condition.A method for structuring a group communication process so that the process is effective in allowing a group of individuals, as a whole, to deal with a complex problem.Any illness resulting from the consumption of contaminated food, pathogenic bacteria, viruses or parasites that contaminate food.Haemolytic uraemic syndrome – a disorder where blood cells are destroyed injuring the kidneys; often occurring following infection with toxin-producing bacteria.Irritable bowel syndrome – chronic abdominal pain, bloating, constipation and diarrhoea; often triggered as the result of bacterial gastroenteritis.International Statistical Classification of Diseases and Related Health Problems, Tenth Revision, Australian Modification.The second longitudinal study of infectious intestinal disease in the United Kingdom by Tam et al.1

The rates at which new cases occur in a population in a specific time period.

A computerised mathematical technique that performs risk analysis by building models of possible results by substituting a probability distribution for any factor that has inherent uncertainty and producing distributions of possible outcome values.

An infection that doctors, laboratories, or other health professionals must report or notify to health departments for the purpose of prevention and control.National Gastroenteritis Survey I – a nationally representative cross-sectional survey conducted in Australia in 2001–2002.National Gastroenteritis Survey II – a nationally representative cross-sectional survey conducted in Australia in 2008–2009.

Foodborne illness in Australia circa 2010ii

NNDSS

OzFoodNet

OzFoodNet outbreak register

Prevalence

ReA

STEC

WQS

National Notifiable Diseases Surveillance System – the national system of surveillance for infectious diseases in Australia.

An Australian network for enhanced surveillance of foodborne disease established by the Australian Government Department of Health in 2000 with Australia’s state and territory health authorities. OzFoodNet investigates multi-jurisdictional outbreaks of disease, provides understanding of the causes and incidence of foodborne disease in the community, and provides evidence for policy formulation.A register of data on outbreaks of gastrointestinal and foodborne diseases that was established in 2000.

The proportion of the population that has a condition at a given point in time, including new and chronic cases of disease.Reactive arthritis – arthritis following bacterial infection; previously known as Reiter’s syndrome.Shiga toxin-producing Escherichia coli – strains of E. coli producing Shiga toxins, which may result in haemolytic uraemic syndrome.Water Quality Study – A randomised controlled trial conducted in Melbourne in 1997–1999 to examine whether reticulated water meeting national guidelines resulted in gastroenteritis by Hellard et al.2

Foodborne illness in Australia circa 2010iii

EXECUTIVE SUMMARYFoodborne illness causes significant morbidity and occasional mortality in Australia.

Reports of disease outbreaks linked to contaminated food are common and may result in

damage to specific food businesses, related food businesses and whole industries. Lost

productivity, impacts on lifestyle and medical expenses from foodborne illness can result in a

substantial burden for Australia. The costs of foodborne illness highlight the need to improve

efforts to prevent disease and strengthen food safety.

Understanding the epidemiology of diseases that occur as a result of contaminated food

is complicated, as there are many different agents that can cause illness. While the majority

of foodborne pathogens cause gastroenteritis, there are some that result in different illnesses,

such as meningitis and hepatitis. Only a fraction of cases of foodborne illness are reported to

health departments and investigated, and for many diseases it is not mandatory for doctors

and laboratories to report cases to health departments for investigation. It is necessary to use

novel methods and analyse multiple datasets to estimate incidence and outcome of foodborne

illness, including notifiable disease reports, laboratory data, outbreak surveillance reports,

expert opinion and the literature.

This report estimates the number of cases of illness and common sequelae acquired in

Australia from contaminated food, circa 2010. It also estimates the number of hospitalisations

and deaths due to foodborne illnesses and sequelae. Uncertainty in the data is accounted for

by reporting 90% Credible intervals (CrI) and comparing estimates using different sources of

data and over time.

Circa 2010, there were an estimated annual 4.1 million (90% CrI: 2.3–6.4 million)

cases of foodborne gastroenteritis acquired in Australia, along with 5,140 (90% CrI: 3,530–

7,980) cases of non-gastrointestinal illness and 35,840 (90% CrI: 25,000–54,000) cases of

sequelae. Norovirus, pathogenic Escherichia coli, Campylobacter spp. and non-typhoidal

Salmonella spp. were the most common known causes of foodborne gastroenteritis, although

approximately 80% of illnesses are of unknown pathogens. Approximately 25% (90% CrI:

13%–42%) of the 15.9 million episodes of gastroenteritis that occur in Australia were

estimated to be transmitted by contaminated food. This equates to an average of

approximately one episode of foodborne gastroenteritis every five years per person. Data on

the number of hospitalisations and deaths represent the occurrence of serious foodborne

illness. Including gastroenteritis, non-gastroenteritis and sequelae, there were an estimated

Foodborne illness in Australia circa 2010iv

annual 31,920 (90% CrI: 29,500–35,500) hospitalisations due to foodborne illness and 86

(90% CrI: 70–105) deaths due to foodborne illness circa 2010.

A main aim of this study was to compare if foodborne illness incidence had increased

over time. In this study, similar methods of assessment were applied to data from circa 2000,

which showed that the rate of foodborne gastroenteritis had not changed significantly over

time. Two key estimates were the total number of gastroenteritis episodes each year, and the

proportion considered foodborne. In circa 2010, it was estimated that 25% of all episodes of

gastroenteritis were foodborne. By applying this proportion of episodes due to food to the

incidence of gastroenteritis circa 2000, there were an estimated 4.3 million (90% CrI: 2.2–7.3

million) episodes of foodborne gastroenteritis circa 2000, although credible intervals overlap

with 2010. Taking into account changes in population size, applying these equivalent

methods suggests a 17% decrease in the rate of foodborne gastroenteritis between 2000 and

2010, with considerable overlap of the 90% credible intervals.

For certain specific foodborne illnesses data were compared from 2000 and 2010 to

examine whether they are increasing or decreasing. For example, foodborne salmonellosis

was estimated to have increased from 28,000 annual infections circa 2000 to 39,600 annual

infections circa 2010; a rate increase of 24% from 1,500 cases per million to 1,850 cases per

million annually. Similarly, for foodborne campylobacteriosis there were 139,000 infections

circa 2000, rising to 179,000 infections circa 2010; representing a 13% increase in incidence

from 7,400 cases per million to 8,400 cases per million annually. Illnesses from hepatitis A

decreased from 245 cases circa 2000 to 40 cases circa 2010, representing a rate decrease of

85% from 13 cases per million to two cases per million annually.

The findings of this study are similar to recent estimates in the United States of

America (USA), Canada and other countries. These estimates take into account

improvements in understanding foodborne illnesses and the agents responsible, improved

information sources, and advances made in methodological approaches. Where possible,

contemporary Australian data collected at or around 2010 were used.

The results of this study will improve the understanding of the epidemiology of specific

pathogens and foodborne causes of gastroenteritis in Australia. This will assist in prioritizing

foodborne illness for intervention. It is important that Australian governments and industry

work together to reduce the incidence of preventable foodborne illness and educate

consumers about good hygiene and food safety strategies.

Foodborne illness in Australia circa 2010v

ACKNOWLEDGEMENTSThe Australian Government Department of Health, New South Wales Food Authority

and Food Standards Australia New Zealand would like to thank the many people who helped

with the conduct of this study and the production of this report. This report was authored by

Dr Martyn Kirk, Dr Kathryn Glass, Ms Laura Ford, Dr Kathryn Brown and Dr Gillian Hall of

the National Centre for Epidemiology and Population Health, under contract to the

commissioning agencies. The Communicable Diseases Network Australia provided data

from the National Notifiable Diseases Surveillance System on notifications of infectious

causes of gastroenteritis. State and territory health departments, public health laboratories

and OzFoodNet epidemiologists also provided data used in this study. The commissioning

agencies and the authors thank Drs Martha Sinclair and Karin Leder for providing Water

Quality Study data for this study. This work was funded by the Australian Government

Department of Health, New South Wales Food Authority and Food Standards Australia New

Zealand.

Foodborne illness in Australia circa 2010vi

INTRODUCTIONIn 2000, it was estimated that every Australian would experience an episode of

gastroenteritis due to food every four years.3 The economic costs of foodborne illness are

substantial from medical practitioner visits, antibiotic prescriptions and days of work lost

each year. The total cost of foodborne illness in Australia is largely attributable to

productivity and lifestyle costs, premature mortality and health care services.4 This, largely

preventable, burden of foodborne illness to the Australian community highlights the need to

continue to improve food safety in Australia.

This report presents an updated estimate of the incidence of foodborne illness in

Australia, circa 2010. It is important to regularly update estimates for foodborne illnesses,

due to improvements in methods of estimation and changing incidence of disease. There are

many different influences on the incidence of foodborne illness, including:

new regulatory measures aimed at preventing infections;

changing agricultural and manufacturing practices;

trends in the way food is prepared, along with consumers’ food choices and changes in eating patterns;

international distribution of food;

emergence of new pathogens; and

identification of new and emerging strains of common infectious agents, such as those that are resistant to antibiotics.

In order to ensure that the estimate of the incidence of foodborne illness in Australia is

internationally comparable, the approach first used by Mead et al.5 and subsequently by

Scallan et al.6 in the USA has been adopted. This approach entails determining the total

amount of gastroenteritis in the country and secondly the proportion of gastroenteritis that is

foodborne. The product of these two estimates gives the total number of cases of foodborne

gastroenteritis. However, a slightly different list of pathogens to those used in the study by

Scallan et al.6 has been selected, including pathogens and illnesses of particular relevance to

Australia. Additionally, this report estimates the incidence of sequelae occurring as a delayed

reaction to episodes of acute gastroenteritis, such as Reactive arthritis (ReA), irritable bowel

syndrome (IBS), haemolytic uraemic syndrome (HUS) and Guillain-Barré syndrome (GBS).

There are many challenges associated with quantifying the true incidence of foodborne

illness in the community. A wide variety of different pathogens produce symptoms of

gastroenteritis. Often clinical cases of gastroenteritis are assessed as ‘presumed infectious’

Foodborne illness in Australia circa 2010Page 1

and a pathogen is never identified, either because a stool test is not performed, the stool test

fails to identify a known pathogen, or because the pathogen is totally unknown. Scallan et al.7

estimated that in the USA, 38.4 million (90% CrI: 19.8–61.3 million) episodes, or 80% of

domestically acquired foodborne illness were caused by unspecified agents. Only a few

decades ago, pathogens such as Campylobacter spp., Shiga toxin-producing Escherichia coli

(STEC), and norovirus were completely unknown.3

One of the most uncertain areas is estimating what proportion of illnesses is due to

contaminated food. Most of the pathogens included in this study may be transmitted to

humans through contaminated foods or water, as well as from other infected persons, from

the environment or from animals. To account for this, expert elicitation was used to estimate

the proportion of illness thought to result from contaminated food.

Most cases of gastroenteritis are mild and self-limiting and most people do not visit a

doctor or submit a specimen for testing. In Australia, state and territory health departments

forward reports of gastrointestinal diseases such as campylobacteriosis, cryptosporidiosis,

salmonellosis, STEC and shigellosis to the National Notifiable Diseases Surveillance System

(NNDSS). However, if people suffering gastroenteritis do not visit a doctor or submit a

specimen for testing their illness will not be recorded in either NNDSS or in state

surveillance systems.

Not all foodborne illnesses are notifiable to health departments, for example

campylobacteriosis is not notifiable in New South Wales (NSW), norovirus, which is one of

the most common causes of gastroenteritis in the developed world, is not notifiable to the

NNDSS and poisoning by ciguatera is only recorded in Queensland. Surveillance data

represents only a small fraction of the total incidence and multipliers are devised to adjust for

underreporting and incomplete population coverage in the Australian surveillance system as

well as for infections acquired overseas. The uncertainty in estimates must be taken into

consideration. The approach used in the current study was to quantify each component of

uncertainty with a plausible probability distribution. Simulations of these distributions were

then used to generate an interval that contains the credible estimates of the number of

foodborne cases of gastroenteritis.

In this report, improvements in the understanding of foodborne illness and the agents

responsible since circa 2000 are taken into account. Specifically, these estimates circa 2010

are based upon new data from:

a review of the literature to determine the best methodological approaches to estimating foodborne illness;


a nationally representative gastroenteritis survey conducted in 2008–9;

updated estimates of underdiagnosis and underreporting of infectious foodborne illnesses to surveillance;

an expert consultation to determine which pathogens and potentially foodborne toxins are currently considered to be of most concern;

an expert elicitation to estimate the proportion of several key pathogens that are transmitted by contaminated food;

NNDSS, state surveillance and the OzFoodNet outbreak register;

hospitalisation separations from each state and territory in Australia; and

mortality statistics for all of Australia from the Australian Bureau of Statistics (ABS).

Although there are many challenges in estimating foodborne illness, the estimates are

valuable for development of public health policy. These estimates provide big-picture

information on the safety of food in Australia and help to evaluate national intervention and

control strategies. In addition, the updated foodborne illness estimates provide a rational basis

for undertaking additional costing studies. This report provides an updated picture on the

incidence of foodborne illness in Australia circa 2010.

AIMSThe aims of this study were to:

1. estimate the incidence, hospitalisations and deaths due to domestically acquired foodborne illness in Australia circa 2010, and

2. examine whether there have been changes to the incidence of foodborne illness over time.


COMPARING FOODBORNE ILLNESS INTERNATIONALLYEstimation of the incidence of foodborne illness has proven important for public health

policy, both in Australia and internationally.3,8,9 Studies to estimate incidence require complex

methods due to the need to synthesize information from various sources, on many different

pathogens and agents, and the need to consider many outcomes of illness. The literature was

examined to review national studies in other countries that estimated the incidence of

foodborne illness due to different pathogens. This included the previous Australian study

estimating foodborne illness incidence circa 2000.10 The report reviews methods used to

estimate the incidence of foodborne illnesses from nine key papers, which were the:

USA assessment of 28 different pathogens or agents in 19995

USA assessment of 30 different pathogens or agents in 20116,7

France assessment of 24 pathogens or agents published in 200511

Australia circa 2000 assessment of 25 pathogens, agents or sequelae, published in 20053,10

United Kingdom (UK) assessment of 24 pathogens or agents, published in 200212

Greece assessment of 15 pathogens or agents, published in 201113

Netherlands assessment of 18 pathogens, agents or sequelae, published in 2012 14

New Zealand assessment of 10 pathogens or agents, published in 200015

Jordan assessment of four pathogens or agents, published in 2009.16

Recent national burden of foodborne illness estimates are shown in Table 1 for

Greece,13 UK,12 France,11 Australia,3 New Zealand,15 USA,6,7 Canada,17 and the Netherlands.14

It should be noted when directly compare these figures that the definition of ‘foodborne

illness’ varies considerably across studies. Excluding the Australian figures, the incidence

rates per million inhabitants per year ranged widely across countries from 4,500 cases in

France to 369,305 cases in Greece. Hospitalisation rates per million inhabitants per year

ranged from 173 in France to 905 in Greece, and deaths per million inhabitants per year

ranged from three in Greece to 12 in France. When the incidence rates are compared across

all countries with estimates, Greece has the highest incidence, followed by Australia, USA,

New Zealand, the Netherlands, UK, and lastly France. In these studies, a similar pattern was

seen for the rate of hospitalisations across countries, and deaths ranged from two to 12

foodborne deaths per million inhabitants (Table 1).


While these studies have commonalities in the overarching approach to estimating the

burden of foodborne illness, some countries assess ‘all types’ of foodborne illness, while

others are more limited in scope. For example, France and New Zealand did not include

unknown pathogens. Not taking unknown pathogens into account results in an underestimate

of the total burden on society. Similarly, not all studies included sequelae or non-

gastrointestinal illnesses in their estimates. In addition, there are significant variations in the

study designs used, the sampling methodologies and case definitions. The authors of these

studies relied on surveillance data, hospital statistics, survey data, previously published

studies, population based cohort studies, expert elicitation studies, reports of outbreak

investigations, and community incidence data from cross-sectional surveys. It is important to

recognise that differences in the sources of data and methods used impacts on results making

it difficult to interpret comparisons between countries and across time.

A landmark paper regarding national foodborne illness estimation is the study from the

USA by Mead et al.5 This paper has been cited an extraordinary number of times and the

methodology replicated by other countries. The method used by Mead and colleagues to

estimate the burden of infectious illnesses transmitted by food was to estimate the total

burden of illnesses due to infections that could potentially be transmitted by food, and then

the proportion of these thought to be foodborne. The burden of illness in the Mead study

included total incidence, hospitalisations and deaths.5

The Mead methodology5 involves:

a) estimating the burden of all gastroenteritis potentially due to foodborne infection;

b) estimating the burden of gastroenteritis due to specific pathogens potentially transmitted by food;

c) estimating the burden of gastroenteritis due to unknown pathogens potentially transmitted by food;

d) estimating the burden of non-gastroenteritis illnesses due to specific pathogens potentially transmitted by food;

e) estimating the proportion of illness burden that is due to transmission of pathogens by food; and

f) accounting for uncertainty.


Country -Author

Years of study

Scope of Study All illnesses(estimate per million

population)

Hospitalisations (estimate per

million population)

Deaths (estimate

per million popn)

Disability-adjust life

years

Unknown pathogens included

Greece -Gkogka13

1996–2006 Total foodborne illness, includes sequelae

369,305(95% CrI: 68,283–910,608)

905(499–1,340)

3(2.0–4.8)

896(470–1,461)

Yes

UK - Adak12 1992–2000 Acute gastroenteritis and listeriosis

26,161 406 9 NA Yes

France -Vaillant11

1997–2000 Total foodborne illness 4,500 173–304 4–12 NA No

Australia - Abelson4

1996–2000 Non-gastroenteritis including sequelae

2,449 156 2 NA No

Australia -Hall3

1996–2000 Acute gastroenteritis 281,250(95% CrI: 208,333–359,375)

766(594–922)

4(2–6) NA Yes

New Zealand - Lake15

2000–2005 Acute gastroenteritis & sequelae

128,421(95% CrI: 34,801–330,075)

NA NA 632(344–1,066)

No

USA - Scallan7

2000–2008 Acute gastroenteritis: unknown pathogens

128,404(90% CrI: 66,318–204,670)

240(33–526)

6(1–11)

NA Yes

USA -Scallan6

2000–2008 Acute gastroenteritis and non-gastroenteritis: known pathogens

31,438(90% CrI: 22,074–42,475)

187(132–253)

5(2–8)

NA No

Canada - Thomas17

2000–2010 Total foodborne illness 126,500(90% CrI: 97,953–158,066)

NA NA NA Yes

Netherlands – Havelaar14

2008–2009 Total foodborne illness, includes sequelae

41,000 NA 5 NA Yes

Table 1: National level estimates of the annual burden of foodborne illness in different countries, 2000–2012.* Adapted from Gkogka et al.13


*The different definitions of foodborne illness, study designs, and sampling methodologies used in these studies have not been taken into account in this comparison. NA – Not available


Since 1999, there have been some significant modifications and improvements in each

of the steps used by Mead et al.5 More recently, disability-adjusted life years (DALYS),

general practitioner (GP) visits, medication, and costings have also been used to estimate the

burden of total gastroenteritis, as well as foodborne illness incidence, hospitalisations and

deaths. Also, new studies have been published, providing more information on incidence of

certain pathogens. Cohort and cross-sectional studies are being used to inform estimates of

the level of underreporting of illness to surveillance systems and large cohort studies

estimating the incidence of total gastroenteritis have become invaluable to this work. In order

to estimate the proportion of illness that is due to contaminated food, recent studies have used

more formal studies of ‘expert opinion’ that have become increasingly rigorous in approach

over time, and outbreak studies have become more available. Simulation techniques are now

used to model uncertainty with varying levels of sophistication.

In the last estimation of the incidence of foodborne illness in Australia, circa 2000,3,10,18

the rate of foodborne acute gastroenteritis was 281,250 (95% CrI: 208,333–359,375) per

million inhabitants per year. The number of illnesses resulting from non-gastroenteritis

foodborne illnesses (invasive listeriosis, toxoplasmosis and hepatitis A), and from their

sequelae (HUS, IBS, GBS and ReA) was estimated to be 2,449 cases per million population

annually.4

More recently, the second study of infectious intestinal disease in the community (IID2

study) estimated that around 25% of people in the UK suffer from an episode of Infectious

Intestinal Disease (IID) each year and that norovirus, sapovirus, rotavirus and Campylobacter

spp. were the most commonly identified pathogens.1 The IID2 study will form an important

basis for future estimates of foodborne illness burden in England and Wales.

Scallan et al.6,7 recently updated estimates of foodborne illness incidence for the USA.

The authors estimated the number of foodborne illnesses, hospitalisations and deaths caused

by 31 domestically acquired pathogens, using surveillance data for the years 2000–2008, and

used data from surveys, hospital records and death certificates to estimate domestically

acquired foodborne illnesses, hospitalisations and deaths caused by unspecified agents.

There were an estimated 31,438 (90% CrI: 22,074–42,475) cases of acute gastroenteritis and

acute non-gastroenteritis caused by known pathogens, 187 (90% CrI: 132–253)

hospitalisations and five deaths (90% CrI: 2–8), per million inhabitants.6 Eighty per cent of

foodborne illness was due to unspecified agents,7 which was similar to the 73% estimated by

Hall et al.3 for Australia circa 2000. In the USA, norovirus was found to account for the most

illnesses (58%) and other important pathogens were non-typhoidal Salmonella spp. (11%),


Clostridium perfringens (10%) and Campylobacter spp. (9%) circa 2000. Pathogens resulting

in the most hospitalisations were non-typhoidal Salmonella spp. (35%), norovirus (26%),

Campylobacter spp. (15%) and Toxoplasma gondii (8%). Non-typhoidal Salmonella spp.

were also responsible for the most deaths (28%), closely followed by T. gondii (24%),

Listeria monocytogenes (19%) and norovirus (11%). The incidence rate for unspecified

agents was estimated at 128,404 (90% CrI: 66,318–204,670) cases, 240 (90% CrI: 33–526)

hospitalisations, and six (90% CrI: 1–11) deaths per million per year in the USA.

The main types of outcome measures of disease burden used in the estimation of

foodborne illnesses internationally have been:

Incidence of new illnesses, hospitalisations, and deaths in a certain time period and population. For some longer lasting illnesses, prevalence may be relevant.

DALYs.

General practitioner (GP) visits, medication use, days of lost productivity.

The direct and indirect costs due to a disease in a set time period. Most of the outcomes above are required to conduct an economic assessment and are combined with raw costing information.

Incidence, hospitalisations, sequelae and deaths are all useful outcome measures and

some studies have shown that DALYs also give another useful perspective. However DALYs

do not take into account differences in socioeconomic and cultural circumstances between

individuals and require subjective value judgements on how to weight or discount for age of

onset, disability weights, and future losses.19 Questions remain about the ability of different

countries, including Australia, to provide high quality, detailed data and to compare disability

weights across countries. Refer to Technical Appendix 1 for further comparisons.


METHODS This report provides estimates of community incidence, hospitalisations, and deaths for

23 pathogens and four sequelae following illness with these pathogens for Australia circa

2010. These pathogens (listed in Table 2) were chosen after review of key papers, and

through consultations with OzFoodNet staff, communicable disease experts, microbiologists

and food safety specialists around Australia. Pathogens that are considered rare in Australia,

or are mostly acquired overseas were excluded.

The fundamental approach to this study builds on methods that were used in Australia

circa 2000, and that have also been used in international estimation efforts in the USA and

the Netherlands.3,6,14 This approach uses estimates of incidence, hospitalisations or deaths to

generate estimates of the yearly incidence of foodborne illness that are domestically acquired

and allows for variability in yearly estimates as well as other sources of uncertainty as

described below.

PATHOGENS INCLUDED

In February 2012, foodborne disease epidemiologists and food safety specialists were

consulted about what pathogens to include in an estimation of foodborne illness in Australia

circa 2010. There was consensus amongst experts to include similar pathogens to the

previous estimate circa 2000. Experts recommended that the ‘Assessment of foodborne

illness, Australia circa 2010’ include 19 pathogens, two chemical agents and four sequelae of

foodborne illness (Table 2). In addition to these recommended pathogens, three viruses

(adenovirus, astrovirus and sapovirus) were included as they had been associated with a

reasonable proportion of all gastroenteritis episodes in recent international studies.1 Finally, a

category of gastroenteritis due to unspecified agents was included.


Table 2: Pathogens, agents and conditions recommended by experts for inclusion in Australian estimation of foodborne illness circa 2010*

Viruses Bacteria Protozoa Chemicals SequelaeRotavirusNorovirusHepatitis A

Bacillus cereusShigellaStaphylococcus aureusVibro parahaemolyticusShiga-toxin producing E. coli

Other pathogenic E. coliCampylobacter spp.Clostridium perfringensListeria monocytogenesSalmonella, non-typhoidalSalmonella TyphiYersinia enterocolitica

Toxoplasma gondiiCryptosporidium spp.Giardia lamblia

ScombrotoxicosisCiguatera

Guillain-Barré syndromeIrritable bowel syndromeReactive arthritisHaemolytic uraemic syndrome

*In addition to these recommended pathogens, the circa 2010 study included three viruses (adenovirus, astrovirus and sapovirus) and a category of gastroenteritis due to unspecified agents.

DATA SOURCES

The main sources of data used in this study are listed in Box 1. Estimates of incidence

relied on notifiable surveillance data at the national and state level, other surveillance data

available through the OzFoodNet outbreak register, cross-sectional data from the National

Gastroenteritis Survey II (NGSII) and cohort study data from the water quality study

(WQS).2,20 Estimates of severe illness were made using hospitalisation and mortality data.

Further details of the use of data sets for this study are provided in Technical

Appendices 3 and 7.


Box 1: Australian datasets used to assess foodborne Illness, hospitalisations and deaths circa 2010

State and territory surveillance: In all Australian jurisdictions, doctors and pathology laboratories are required by legislation to report patients with diagnoses of certain infections to a state or territory health department. Basic information is then stored on each case-patient. The specific illnesses that are reportable vary slightly between each jurisdiction; however data elements and case definitions for many illnesses are largely standardised across Australia. The numbers of infections or foodborne syndromes where illnesses were not nationally notifiable were used from state or territory surveillance, as well as the proportion of infections acquired from overseas travel from infections where illnesses were not nationally notifiable.

National Notifiable Diseases Surveillance System (NNDSS): NNDSS was established in 1990 and is managed by the Australian Government Department of Health through the Communicable Diseases Network Australia. State and territory health departments collect notifications of communicable disease under their respective public health legislation. Under the National Health Security Agreement, states and territories forward de-identified notification data on the nationally agreed set of 65 communicable disease to the Australian Government Department of Health for the purposes of national communicable disease surveillance, although not all 65 diseases are notifiable in each jurisdiction. The national numbers of infections or syndromes from NNDSS from 2006–2010, which are nationally notifiable and potentially due to food, were used to assess the incidence of foodborne illness.

OzFoodNet outbreak register: In 2000, the Australian Government established OzFoodNet—a national network of epidemiologists in state and territory health departments to enhance national surveillance of foodborne illness. OzFoodNet epidemiologists collect enhanced surveillance data on a variety of foodborne illnesses. The OzFoodNet outbreak register aggregates data on outbreaks of disease into a national dataset. A cleaned version of the outbreak register for the years 2006 to 2009 was used for those pathogens that occur in outbreaks, especially those that are not nationally notified.

National Gastroenteritis Surveys I & II (NGSI & NGSII): OzFoodNet and the National Centre for Epidemiology and Population Health (NCEPH), funded by the Australian Government Department of Health and New South Wales Food Authority, conducted national cross-sectional surveys to estimate the national incidence of gastroenteritis meeting a specific case definition. The surveys used computer assisted telephone interviews. NGSI was conducted in 2001–2, while NGSII was conducted in 2008–9.

Water Quality Study (WQS): The WQS was conducted by the Cooperative Research Centre for Water Quality and Treatment between 1997 and 1999 in the suburbs of Melbourne. This randomised-controlled trial included 600 families and examined the relationship between water quality and gastroenteritis. For further information on the WQS (refer to <http://www.med.monash.edu.au/epidemiology/infdis/waterqstudy.html>. After applying an age adjustment, the Water Quality Study was used to determine the proportion of infectious agents causing diarrhoea in community gastroenteritis.

Hospitalisation data: Principal and additional hospital diagnoses for the pathogens of interest were obtained for all states and territories. Data from Tasmania, Australian Capital Territory, Western Australia, South Australia, and Northern Territory was for 2006–2010, data from New South Wales was for 2009 and 2010, and data from Victoria was for the 2009–2010 financial year.


Box 1: Australian datasets used to assess foodborne Illness, hospitalisations and deaths circa 2010 (continued)

Australian Bureau of Statistics mortality data: The ABS supplied aggregated data on deaths registered in Australia from 2001 to 2010, by age group and sex, where the underlying or multiple cause of death was from an illness that could potentially be foodborne.

Other: The assessment of foodborne illness also relied on data from expert elicitations as well as published reports in the literature. The expert elicitations took place in 2005 and 2009 and included eleven participants comprising of two public health physicians, two microbiologists, one food safety officer, two public health veterinarians, three foodborne disease epidemiologists and one research scientist.

INCIDENCE

TOTAL INCIDENCE OF GASTROENTERITISTo estimate the annual incidence of gastroenteritis in Australia, OzFoodNet, in

conjunction with the National Centre for Epidemiology and Population Health (NCEPH), and

funded by the Australian Government Department of Health and New South Wales Food

Authority, conducted the NGSII over a 12-month period between February 2008 and January

2009. This was a cross sectional study using a nationally-representative telephone survey to

improve the estimates of the incidence of gastroenteritis and assess whether there were

significant changes in incidence from the previous National Gastroenteritis Survey (NGSI)

conducted in 2001–2. Gastroenteritis was defined as experiencing ≥3 loose stools and/or ≥2

vomits in a 24 hour period, or if the person had concomitant respiratory symptoms,

gastroenteritis was defined as ≥4 loose stools and/or ≥3 vomits in a 24 hour period. NGSII

provided an incidence estimate of 0.74 (95%CI 0.64–0.84) episodes of gastroenteritis per

person per year, or 15.9 million cases each year in Australia (Technical Appendix 2).

INCIDENCE BY PATHOGENThree main approaches were used to calculate the incidence of illness due to each

pathogen or agent:

1. Notifiable surveillance approach using data from NNDSS or state-based notification systems;

2. Pathogen fraction approach using data from the NGSII together with cohort studies such as the WQS2,20 conducted in Melbourne between 1997 and 1999;

3. Other surveillance approach using data from the OzFoodNet outbreak register, or from hospitalisations.

Approaches 1 and 3 both consist of scaling up from surveillance data to estimates in the

community by use of underreporting and outbreak multipliers. Approach 2 can be thought of


as a top-down approach; it attributes a fraction of all gastrointestinal illness to the given

pathogen.

These three approaches were considered to form a hierarchy, with the notifiable

surveillance approach used by preference, followed by the pathogen fraction approach, and

the other surveillance approach where other data sources were not available. Where possible,

more than one approach was used for each pathogen to provide a secondary estimate used as

a tool to validate the methods.

In the final calculations, the notifiable surveillance approach was used for 11

pathogens, the pathogen fraction approach for 6, and the other surveillance approach for five

pathogens or agents (Technical Appendix 4 and Technical Appendix 11). An additional

approach (Technical Appendix 6) based on USA seroprevalence data was applied to

toxoplasmosis,21 owing to a lack of Australian data.

MULTIPLIERS AND UNCERTAINTYWithin each approach, multipliers were applied to adjust for factors such as the age

profile of cohort studies, and the proportion of the population covered by surveillance. The

study by Hall et al.22 was used as a basis for estimating underreporting multipliers for

moderate illnesses and bloody diarrhoea, and the underreporting factor for serious illnesses

was estimated as one reported illness for every two illnesses that occur in the community, as

in Mead et al.5 In addition, an outbreak multiplier was applied to further adjust for estimates

made using other surveillance data. More details of the underreporting multipliers are

provided in Technical Appendix 5.

Two other key multipliers were the ‘domestically acquired multiplier’ and the

‘foodborne multiplier’. The domestically acquired multiplier adjusted total incidence data to

exclude infections acquired overseas. For many pathogens, this multiplier was estimated

using NNDSS data by state and territory. Other pathogens with short duration of illness were

assumed to be 100% domestically acquired.

Foodborne multipliers were estimated for nine pathogens using expert elicitation data

using a Delphi approach from 2009 and for a further nine pathogens using a Delphi approach

from 2005. All illness due to the seafood toxins was assumed to be caused by food. More

details on the estimation of the proportion of illness that is foodborne are given in Technical

Appendix 5.

Uncertainty in estimates of incidence and in the value of key multipliers was included

by means of distributions. The main distributions used were empirical distributions and


program evaluation review technique (PERT) distributions. Empirical distributions were used

for counts of cases detected by year through notifiable or other surveillance systems. PERT

distributions are based on the beta distribution, and are commonly used for expert elicitation

and risk assessment studies. This distribution was used widely in the analysis as it allows for

asymmetric distributions, and can be easily produced from many data sources including

expert elicitation data.

INCIDENCE OF SEQUELAEThis report estimates the incidence of sequelae illnesses for GBS, HUS, IBS and ReA

using data from Australian and international literature, as well as from NNDSS. Where

available, data from large cohort studies, Australian surveillance studies, and meta-analyses

were used in conjunction with adjusted notification data from preceding foodborne

gastrointestinal infections. More details on the literature and methods used to estimate the

number of cases of foodborne illness sequelae can be found in Technical Appendix 8.

HOSPITALISATIONS AND DEATHS FROM FOODBORNE ILLNESSYearly hospitalisations were estimated from state and territory hospitalisation data

(ICD-10-AM codes) and deaths from the ABS mortality data (ICD-10 codes) for the 24

potentially foodborne illnesses in Table 2, as well as unknown gastroenteritis, listed as either

principal or additional diagnosis. Hospitalisation data were provided over the period 2006 to

2010 for most states and territories, and this was used to calculate national estimates for

2006–2010. A large number of hospitalisations and deaths due to gastrointestinal illness do

not identify a causal pathogen, and these were categorised as ‘gastroenteritis of unknown

aetiology’ (refer to Technical Appendix 7 for ICD-10-AM codes used).

Monte Carlo simulations were used to adjust for travel associated cases, and to estimate

the proportion of hospitalisations and deaths that were foodborne. Similar multipliers were

used to adjust raw hospitalisation and mortality data as for incidence data, as described in

Technical Appendix 7. More details on methods to estimate sequelae hospitalisations and

deaths can be found in Technical Appendix 9.

CHANGES IN METHODS TO CALCULATE INCIDENCE FROM 2000

There were several changes in the methods used to calculate incidence and hospitalisations

since the estimates circa 2000, including updated underreporting multipliers22 and a more

rigorous expert elicitation.23 These changes make direct comparison of the circa 2010


findings with those of the circa 2000 estimates potentially misleading (Box 2). To estimate

changes over time, circa 2000 estimates were recalculated with original data but using

identical methods to those used to calculate the circa 2010 estimates. The circa 2000

incidence rate for foodborne gastroenteritis and key pathogens was then calculated using

population statistics from the ABS from 1996 to 2000.24 Only pathogens for which there were

surveillance data from both time periods were included in this analysis. More details on the

new methods to calculate incidence from 2000 can be found in Technical Appendix 10.

Box 2: Change in methods to estimate foodborne illness in Australia since circa 2000

Domestically acquired multiplier: In 2000, this was estimated using Victorian data only. In 2010, data from all jurisdictions were obtained to inform the estimate of the proportion of cases that were acquired in Australia. Pathogen-specific changes are provided in Table T5.1 of Technical Appendix 5.

Under-reporting multiplier: A study published by Hall et al. (2008) produced more accurate, data-driven estimates of the under-reporting multipliers for Salmonella, Campylobacter, and STEC. These changes have a considerable impact on the circa 2010 estimates – for instance the multiplier for Salmonella decreased from 15 to seven following the use of this study.

Foodborne multiplier: An additional expert elicitation process allowed us to update this multiplier to include assessments of uncertainty from experts. Table T5.2 of Technical Appendix 5 provides all of the foodborne multipliers by pathogen and their data source. Changes in the foodborne multipliers have a considerable effect on the overall proportion of gastroenteritis that is attributed to food.

Outbreak multiplier: Given that illnesses with short duration are less likely to be confirmed, the outbreak multiplier was estimated based on the total number of ill patients in confirmed outbreaks, using Salmonella as the reference pathogen. This multiplier was estimated using NNDSS data and the OzFoodNet outbreak register, which are more nationally representative than the Victorian state outbreak and surveillance data used in 2000.

Choice of approach: As in 2000, there were three main approaches to estimate incidence of pathogens using notifiable surveillance data, cohort data, and outbreak data. However, this time two methods for each pathogen were used to test the sensitivity of the results to the choice of method, and to help identify the most appropriate method. Although only the results of the primary method are reported, the secondary approach is listed in the pathogen sheets in Technical Appendix 11. In some cases, such as Cryptosporidium spp., the choice of primary method changed from 2000 to 2010.

Water Quality Study: In 2000, the rates from the WQS were used but in 2010 rates were adjusted to account for the sample selection of families with children. This is an improvement as the age structure of the sample for the WQS does not accurately reflect the age structure of the Australian population. As gastroenteritis is age related, adjustment gives a better estimate for the whole population. For rotavirus, WQS data were adjusted for changes over time, based on published information.

Hospitalisation data: In this assessment the numbers of hospitalisations for both principal and additional diagnoses was acquired, rather than only principal diagnosis as in 2000. The 2000 estimate included a multiplier to estimate total hospitalisations while circa 2010 data included all hospitalisations. A multiplier was applied in both 2000 and 2010 to account for underreporting.


RESULTS

INCIDENCE OF GASTROENTERITIS DUE TO FOOD

Based on data collected from 2006–2010, each year, an estimated 4.1 million (90% CrI:

2.3–6.4 million) cases of foodborne gastroenteritis occur in Australia, equating to 0.19 cases

per person per year. Foodborne gastroenteritis was considered to include any infectious

gastroenteritis caused by eating food, including food contaminated just before eating.3 An

estimated 25% (90% CrI: 13–42%) of 18 known gastrointestinal pathogens were transmitted

by contaminated food (Table 3). Bacterial pathogens had the highest foodborne proportion

with 36% estimated to be transmitted by contaminated food, compared to 16% for viruses,

and 11% for parasites.

Of the total 4.1 million cases of foodborne gastroenteritis annually, about 0.8 million

(20%) were estimated to be due to one of the 18 known pathogens. The remaining 3.3

million cases (80%) were due to unknown or unidentified pathogens. The key known

pathogens in terms of incidence were: pathogenic E. coli, norovirus, Campylobacter spp., and

non-typhoidal Salmonella spp. Together these four pathogens make up 93% of foodborne

episodes due to the 18 known pathogens.

INCIDENCE OF OTHER FOODBORNE ILLNESS

In addition to foodborne gastroenteritis, circa 2010, contaminated food was estimated to

cause around 5,100 cases of non-gastrointestinal illness each year in Australia (Table 4). T.

gondii was the most common cause of non-gastrointestinal illness due to food. The

percentage of these illnesses that were estimated to be transmitted by contaminated food

ranged from 12% for hepatitis A up to 100% for the fish-associated diseases

scombrotoxicosis and ciguatera.

INCIDENCE OF SEQUELAE FOLLOWING FOODBORNE GASTROENTERITIS

Circa 2010, contaminated food was estimated to cause around 35,840 episodes of

sequelae following acute gastroenteritis each year in Australia (Table 5). This represents a

rate of 1,620 sequel illnesses per million population resulting from foodborne illness. IBS

was the most common sequelae, causing an estimated 19,500 episodes each year; followed by

ReA resulting in 16,200 episodes each year.


PathogenMedian number of domestically acquired

episodes of gastroenteritis (90% CrI)Median percentage

foodborne(90% CrI)

Median number of domestically acquired episodes of gastroenteritis (90%

CrI)Bacillus cereus 3,350 (900–10,100) 100% (98–100) 3,350 (900–10,100)Campylobacter spp. 234,000 (147,000–374,000) 77% (62–89) 179,000 (108,500–290,000)Clostridium perfringens 16,500 (2,600–53,400) 98% (86–100) 16,100 (2,550–50,600)STEC 4,300 (2,050–9,500) 56% (32–83) 2,350 (950–5,850)Other pathogenic E. coli 1,100,000 (833,000–1,450,000) 23% (8–55) 255,000 (85,800–632,000)

Salmonella, non-typhoidal

56,200 (31,900–101,000) 72% (53–86) 39,600 (21,200–73,400)

Salmonella Typhi 20 (8–45) 75% (2–97) 15 (5–30)Shigella 3,000 (1,650–5,400) 12% (5–23) 350 (150–850)Staphylococcus aureus 1,300 (200–7,050) 100% (95–100) 1,300 (200–7,000)Vibrio parahaemolyticus 60 (15–170) 75% (5–96) 40 (10–120)Yersinia enterocolitica 1,500 (900–2,500) 84% (28–94) 1,150 (650–1,950)Adenovirus 88,400 (28,800–205,000) 2% (1–3) 1,650 (500–4,650)Astrovirus 67,100 (20,900–155,000) 2% (1–3) 1,300 (350–3,400)Norovirus 1,550,000 (1,220,000–1,940,000) 18% (5–35) 276,000 (78,100–563,000)Rotavirus 44,800 (18,500–90,800) 2% (1–3) 850 (300–2,000)Sapovirus 81,600 (63,400–102,000) 18% (5–35) 15,000 (7,450–24,300)Cryptosporidium spp. 17,900 (8,150–39,800) 10% (1–27) 1,700 (150–6,100)Giardia lamblia 32,800 (19,800–56,400) 6% (1–50) 3,700 (800–10,600)Subtotal 3,090,000 (2,810,000–3,900,000) 25% (13–42) 798,000 (528,000–1,310,000)Unknown aetiology 12,800,000 (10,500,000–14,500,000) 25% (13–42) 3,310,000 (1,800,000–

5,152,000)Total 15,900,000 (13,700,000–18,000,000) 25% (13–42) 4,110,000 (2,330,000–6,390,000)

Table 3: Estimated annual number of episodes of domestically acquired foodborne gastroenteritis caused by pathogens in Australia, circa 2010


Table 4: Estimated annual number of episodes of acute foodborne illness due to pathogens or syndromes that do not result in gastroenteritis in a typical year in Australia, circa 2010

Pathogen/acute illness Median percentage foodborne(90% CrI)

Median number of domestically acquired foodborne illnesses

(90% CrI)

Hepatitis A 12% (5–24) 40 (10–100)Listeria monocytogenes 98% (90–100) 150 (50–200)Toxoplasma gondii 31% (4–74) 3,750 (1,400–7,150)Ciguatera 100% (100–100) 150 (40–300)Scombrotoxicosis 100% (100–100) 1,050 (0–2,450)Total 40% (25–59) 5,140 (3,530–7,980)

Sequelae Foodborne agents

Median percentage foodborne(90% CrI)

Median number of domestically acquired foodborne illnesses

(90% CrI)

Guillain-Barré syndrome

Campylobacter spp. 25% (10–43)

70 (30–150)

Haemolytic uraemic syndrome

Shiga-toxin producing E. coli

33% (17–53)

70 (25–200)

Irritable bowel syndrome

Campylobacter spp., non-typhoidal Salmonella spp., Shigella

13% (8–20)

19,500 (12,500–30,700)

Reactive arthritis Campylobacter spp., non-typhoidal Salmonella spp., Shigella, Yersinia enterocolitica

48% (36–62)

16,200 (8,750–30,400)

Total 35,840 (25,000–54,000)

Table 5: Estimated annual incidence of sequelae following foodborne gastroenteritis in a typical year in Australia, circa 2010

CHANGES IN INCIDENCE FROM 2000

A direct comparison of incidence estimates with those circa 2000 is complicated by

changes in methods between the two studies. Points of difference and method-related issues

are described in Box 2 in the Methods section.


FOODBORNE GASTROENTERITISThe proportion of gastroenteritis that was estimated to be foodborne circa 2010 was

25%, compared to 32% in 2000. These changes are largely driven by changes in the estimates

of the foodborne multiplier for other (non-STEC) pathogenic Escherichia coli and norovirus,

as ‘other pathogenic E. coli’ is assumed to be 23% foodborne compared to 50% foodborne in

2000 and norovirus is assumed to be 18% foodborne compared to 25% foodborne in 2000.

These lower proportions represent better informed estimates by experts rather than true

changes in the proportion of disease transmitted by food. By applying the 2010 estimate of

25% foodborne to the results of the NGSI study that was conducted in 2001, there are an

estimated 4.3 million (90% CrI: 2.2–7.3 million) episodes of foodborne gastroenteritis circa

2000. When adjusted for changes in population size over this period, this represents a 17%

decrease in the rate of foodborne gastroenteritis between 2000 and 2010 (rate ratio 0.83, 90%

CrI: 0.4–1.8), although credible intervals overlap (Table 6).

SALMONELLA AND CAMPYLOBACTERComparing estimates of non-typhoidal Salmonella spp. and Campylobacter spp. with

those circa 2000 requires further adjustment for considerable changes in methods, such as the

halving of the underreporting multiplier for non-typhoidal Salmonella spp. Recalculation of

the 2000 estimates using these new methods gives a revised estimate of 28,000 (90% CrI:

15,000–50,000) yearly cases of foodborne gastroenteritis due to non-typhoidal Salmonella

spp. and 139,000 (90% CrI: 82,500–227,000) yearly cases of foodborne gastroenteritis due to

Campylobacter spp. That is, the 2010 point estimates represent around 11,000 more cases of

foodborne non-typhoidal Salmonella spp. and 40,000 more cases of foodborne

Campylobacter spp.; a rate increase of 24% for non-typhoidal Salmonella spp. and 13% for

Campylobacter spp. when taking into account population changes (Table 6). Credible

intervals for all incidence rate ratios include uncertainty derived from incidence multipliers,

and were considerably wider than intervals for ratios derived from raw surveillance data.


Table 6: Comparison of annual incidence estimates and rates from revised circa 2000 and circa 2010 of foodborne gastroenteritis, key pathogens, and two sequelae using current assumptions, Australia

Pathogen/Illness Incidence circa 2000(90% CrI)

Rate per million population circa 2000(90% CrI)

Incidence circa 2010

(90% CrI)

Rate per million population circa 2010(90% CrI)

Rate ratio

(90% CrI)Foodborne

gastroenteritis4.3 million

(2.2–7.3 million)

224,000(116,000–374,000)

4.1 million(2.3–6.4 million)

186,000(105,000–289,000)

0.83(0.4–1.8)

Campylobacter spp. 139,000(82,500–227,000)

7,400(4500–12,200)

179,000(108,500–290,000)

8,400(5,050–13,650)

1.13(0.5–2.3)

Salmonella spp., non-typhoidal

28,000(15,000–50,000)

1,500(800–2,700)

39,600(21,200–73,400)

1,850(1,000–3,350)

1.24(0.5–2.8)

Salmonella Typhi 9(3–21)

0.5(0–1)

15(5–30)

0.6(0–1)

1.2(0.5–2.6)

Shigella 515(175–1,300)

28(9–70)

350(150–850)

16(6–40)

0.57(0.15–2.25)

Hepatitis A 245(65–725)

13(3–40)

40(10–100)

2(1–5)

0.15(0.06–0.4)

Listeria monocytogenes

125(70–185)

7(4–10)

150(50–200)

7(3–10)

1(0.4–1.9)

Giardia lamblia 2,600(565–7,400)

140(30–405)

3,700(800–10,600)

175(35–490)

1.25(0.2–7.5)

Guillain-Barré syndrome

50(25–100)

2.8(1–6)

70(30–150)

3.1(2–6)

1.13(0.5–3.6)


14,800(9,500–23,500)

850(550–1350)

19,500(12,500–30,700)

915(570–1,440)

1.07(0.5–2)


OTHER PATHOGENSComparing estimates of other key pathogens shows that the number of hepatitis A and

Shigella cases has decreased since 2000, with rate decreases of 85% and 43% respectively

(Table 6). The rate of L. monocytogenes has remained the same, and the rates of Salmonella

Typhi and Giardia lamblia increased 20% and 25% respectively (Table 6). Comparing

estimates of two sequelae shows an increase in the rate of both GBS and IBS from 2000 to

2010 (Table 6). While estimates of Staphylococcus aureus circa 2010 are lower than those

estimated circa 2000, this may be a consequence of changes in methods for outbreak data, or

the considerable variation in yearly cases associated with outbreaks. The estimates for S.

aureus and Cryptosporidium for circa 2000 could not be recalculated using new methods as

source data has changed, and the new source of data was not available circa 2000. No new

sources of data were available for ‘other pathogenic E. coli’, and the other viral pathogens, so

that meaningful comparison of the two time periods is not possible.

SEVERE FOODBORNE ILLNESS

HOSPITALISATIONSEach year, there were an estimated 30,600 hospitalisations with foodborne

gastrointestinal illness as the principal or additional diagnosis (Table 7). Of these, 5,900

were due to known pathogens each year, with Campylobacter spp. (3,200 hospitalisations)

and non-typhoidal Salmonella spp. (2,100 hospitalisations) the main causes. The remaining

24,700 hospitalisations were due to foodborne gastroenteritis of unknown aetiology.

Further to this, there were an estimated 240 hospitalisations for acute foodborne illness

that were not gastrointestinal, with L. monocytogenes as a key cause. Sequelae resulting from

gastrointestinal illness were estimated to cause 1,080 hospitalisations each year, as a

secondary outcome from illness due to foodborne pathogens.

DEATHSThere were an estimated 60 (90% CrI: 45–75) deaths each year due to foodborne

gastroenteritis (Table 7). Of these, non-typhoidal Salmonella spp. was the most commonly

identified pathogen, causing an estimated 15 foodborne deaths each year. Gastroenteritis of

unknown aetiology was a principal or additional cause of death in an estimated 39 people per

year.


Non-gastrointestinal foodborne illness caused an additional 16 deaths each year, with

most of these attributed to L. monocytogenes. Finally, 10 deaths each year were due to

sequelae following foodborne illness with pathogens such as STEC, non-typhoidal

Salmonella spp., Campylobacter spp., Shigella, and Yersinia enterocolitica. The leading

cause of death from foodborne sequelae was GBS following campylobacteriosis, which

resulted in an estimated six deaths each year.

CHANGES IN SEVERE ILLNESS FROM CIRCA 2000

There were relatively few methodological changes associated with estimates of

hospitalisations and deaths due to foodborne pathogens allowing for direct comparison

between 2000 and 2010 for many pathogens (refer to Hall et al3,10 for circa 2000 estimates).

Estimates of hospitalisations for both non-typhoidal Salmonella spp. and

Campylobacter spp. increased from 2000 to 2010, with a rise in around 1,000 hospitalisations

for each pathogen. In particular, the estimated hospitalisations for non-typhoidal Salmonella

spp. approximately doubled in the ten year period from an estimated 1,060 hospitalisations

circa 2000 to an estimated 2,100 hospitalisations in circa 2010.

As was noted for incidence, estimates of hospitalisations for Staphylococcus aureus, as

well as for Vibrio parahaemolyticus, declined since 2000, although numbers for each are

small and diagnosis in hospital settings is likely to be poor. The decline in hospitalisations for

rotavirus was less marked than that for incidence, but does show a decrease from an

estimated 70 foodborne hospitalisations per year in 2000 to 50 each year in 2010. The

increase in estimated hospitalisations for foodborne norovirus from four per year in 2000 to

150 per year in 2010 reflects increased testing and diagnosis of this pathogen.


Table 7: Estimated annual number of hospitalisations and deaths caused by domestically acquired foodborne pathogens in Australia, circa 2010Pathogen/Illness ICD-10-AM code Median number of

hospitalisations(90% CrI)

Median number of deaths (90% CrI)

Gastrointestinal foodborne illnessBacillus cereus A05.4 25 (4–45) 0 (0–0)Campylobacter spp. A04.5 3,200 (2,100–4,500) 3 (2–4)Clostridium perfringens A05.2 0 (0–2) 1 (0–1)STEC A04.3 7 (2–15) 0 (0–0)Other pathogenic E. coli A04.0, A04.1, A04.4 20 (6–50) 0 (0–1)Salmonella, non-typhoidal A02.0–A02.9 2,100 (1,300–3,000) 15 (8–20)Salmonella Typhi A01.0 15 (6–35) 0 (0–0)Shigella A03 25 (9–50) 0 (0–0)Staphylococcus aureus A05.0 10 (7–20) 0 (0–0)Vibrio parahaemolyticus A05.3 1 (0–1) 0 (0–0)Yersinia enterocolitica A04.6 35 (10–65) 1 (0–1)Adenovirus A08.2 15 (8–25) 0 (0–0)Astrovirus n/a not applicable not applicableNorovirus A08.1 150 (35–350) 1 (0–2)Rotavirus A08.0 50 (30–100) 0 (0–0)Sapovirus n/a not applicable not applicableCryptosporidium spp. A07.2 40 (6–100) 0 (0–0)Giardia lamblia A07.1 100 (25–300) 0 (0–0)Subtotal 5,900 (4,700–7,500) 21 (14–26)Unknown aetiology A08.4, A09, A09.0, A09.9 24,700 (22,600–27,800) 39 (27–54)Total (foodborne gastroenteritis)

30,600 (28,000–34,000) 60 (45–75)

Non-gastrointestinal foodborne illnessHepatitis A B15.9 20 (6–50) 0 (0–2)Listeria monocytogenes A32 150 (100–250) 15 (9–20)Toxoplasma gondii B58 30 (10–60) 1 (0–2)Ciguatera T61.0 25 (10–40) 0 (0–0)Scombrotoxicosis T61.1 8 (5–10) 0 (0–0)Total (non-gastrointestinal) 240 (180–350) 16 (10–21)Sequelae resulting from foodborne illnessGuillain-Barré syndrome G61.0 70 (30–150) 6 (2–10)Haemolytic uraemic syndrome D59.3 70 (25–200) 2 (1–3)Irritable bowel syndrome K58.0, K58.9 915 (550–1,400) 2 (1–2)Reactive arthritis M02.1, M02.3, M02.8,

M03.225 (20–40) 0 (0–0)

Total (sequelae) 1,080 (700–1,600) 10 (5–14)

DISCUSSION


There are an estimated 4.1 million (90% CrI: 2.3–6.4 million) cases of gastroenteritis

attributed to contaminated food in Australia each year. This equates to each Australian

experiencing an episode of foodborne gastroenteritis approximately every five years. While

foodborne gastroenteritis is often not serious, it results in considerable costs to society

through medical costs and days of work lost. Approximately one in five people with

gastroenteritis seek medical attention, which could mean that foodborne illness results in as

many as a million visits to a doctor annually.18

Acute non-gastroenteritis illnesses and sequelae due to contaminated food are also a

significant contributor to the incidence of foodborne illness in Australia. While there are

notably fewer cases from these illnesses than from foodborne gastroenteritis, they cause more

serious symptoms and are of longer duration. In this study, sequelae responsible for the

highest rates of illness included IBS and ReA, which are not fatal, but cause disabling

symptoms.25,26 It is important to recognise that the current study estimates yearly incident

cases only, and does not measure the long-term burden of these sequelae.

It was concerning to identify that foodborne Campylobacter spp. were responsible for

80% of the sequelae cases estimated in this study, as well as approximately 179,000 cases of

gastroenteritis every year. The rate of infection due to Campylobacter spp. has increased over

the last 10 years and is higher than many other developed nations. For example, the

Australian Campylobacter spp. rate is approximately 10 times higher than the USA,27 double

that of the Netherlands14 and slightly higher than that of the UK.1 In New Zealand, source

attribution and policy regulation led to interventions that effectively lowered Campylobacter

spp. rates and sequelae rates as well. Interventions introduced in 2006 in New Zealand,

including Campylobacter spp. performance targets at primary processing and the promotion

of freezing all fresh poultry meat, which by 2008 resulted in the rate of Campylobacter spp.

notifications decreasing by 54% and a fall in GBS hospitalisations by 13%.28,29 If Australia

were able to introduce successful interventions and experience the same decreases in

infection as New Zealand, the rate of foodborne Campylobacter spp. cases in the Australian

community would be expected to drop from approximately 8,400 to 3,864 cases per million,

leading to a decline in the incidence of sequelae from 1,620 to 870 cases per million per year.

In addition to Campylobacter spp., norovirus, ‘other pathogenic E. coli’, and non-

typhoidal Salmonella spp. were the main contributors to the incidence of foodborne illness.

Reducing the number of cases of these illnesses in the community would reduce the societal

burden of foodborne gastroenteritis. Meats, particularly chicken meat, and eggs have been

shown to be important food sources for infections due to Campylobacter spp.,non-typhoidal


Salmonella spp.,30-33 and in Australia, fish is the source of seafood toxins responsible for

ciguatera and scombrotoxicosis.34 Generating robust attribution data to identify food sources

for specific pathogens, as well as contributing factors throughout the production and supply

chain may help target food safety policies and interventions.

Similar foodborne illness estimation studies have been conducted in the USA,5,6 UK,12

Canada,17 and the Netherlands.14 The proportion of gastroenteritis estimated to be due to

foodborne transmission in the current study (25%) is remarkably similar to the UK estimate

(26.2%) and the most recent USA estimate (25.8%), but lower than that of the Netherlands

(39%). Although the Canadian study does not report an overall proportion foodborne,

analysis of their results put it around 20%. In addition, the proportion of foodborne

gastroenteritis due to unknown aetiology that was estimated in the USA (80%) is the same

proportion that was estimated in this study. The overall estimates of the proportion that is

foodborne depend on the selection of pathogens for inclusion, the incidence of common

pathogens, and the estimated proportion that are foodborne.

While this study builds on an earlier Australian study estimating the incidence of

foodborne illness circa 2000, there have been a number of methodological improvements

since then. Data from the OzFoodNet outbreak register were used for pathogens that are not

included in the national surveillance system. National data from NNDSS were used to

determine the proportion of cases that were associated with travel. Hospitalisation data from

all jurisdictions and death data from the ABS were used to estimate the incidence of severe

illness. Use of national data takes account of variation in foodborne illness patterns by state

and territory, and provides more representative national estimates. Use of both principal and

additional diagnosis data from hospitals more accurately captures hospitalisation patterns by

pathogen.

A new expert elicitation undertaken in 2009 for this estimation effort was incorporated,

further improving data quality.23,35 The proportion of foodborne transmission was generally

estimated to be lower and uncertainty bounds wider, compared to the estimates found in the

Delphi process in 2005. This may be a reflection of a general perception that environmental

sources of infection have been somewhat neglected and that health departments have a

primary focus on ‘foodborne disease’.23,35 These lower foodborne proportions translated into

fewer illnesses, hospitalisations, and deaths being attributed to food, compared to the circa

2000 study.

In estimating community incidence, underreporting multipliers were used to adjust for

the proportion of people who were infected with a pathogen or agent but did not seek


treatment or submit specimens for testing. Following a 2008 study by Hall et al.,22 new

multipliers were used for non-typhoidal Salmonella spp., Campylobacter spp., and STEC.

The non-typhoidal Salmonella spp. underreporting multiplier of 7 (95% CI: 4–14) was

extrapolated to all other moderate illnesses, besides Campylobacter spp. These new

underreporting multipliers are smaller than the multiplier of 15 (CrI: 5–25) that was used in

2005.3 This is one of the main reasons that direct comparison between the estimates circa

2000 and circa 2010 can be misleading. However, application of these new multipliers to

data circa 2000 has validated that there have been increases in the incidence of illness due to

Campylobacter spp. and non-typhoidal Salmonella spp. between 2000 and 2010.

An underreporting multiplier for serious illnesses and the underdiagnosis multiplier for

hospitalisations and deaths of two cases for every case reported was used,3,10 which is

consistent with studies by Mead et al.5 and Scallan et al.6 The use of this multiplier for

hospitalisations and deaths was supported by comparing data from the OzFoodNet outbreak

register to hospital and ABS deaths data, which suggested that a multiplier of at least two was

necessary to account for underdiagnosis. There are limited data on pathogen-specific

underdiagnosis and further studies are required to thoroughly validate this multiplier and

assess whether there are pathogen-specific differences in underdiagnosis of severe illness.

Another improvement in this study was the higher level of detail and research that went

into the estimates for sequelae, including extensive literature searches and the addition of a

bacterial multiplier in estimates of the foodborne proportion. Many of the international

estimation efforts have not undertaken the task of estimating foodborne sequelae.5-7,12 The

current estimation work revealed the additional burden from foodborne pathogens, such as

Campylobacter spp., non-typhoidal Salmonella spp., and STEC, that is not seen when just

looking at the incidence of acute illness. For instance, while foodborne Campylobacter spp.

infection is directly responsible for three (90% CrI: 2–4) deaths each year, it also causes a

further six (90% CrI: 2–10) deaths from GBS and is potentially responsible for deaths from

other sequelae. Overall, Campylobacter spp. was responsible for around 80% of sequelae

estimated in this study, highlighting the importance of control measures to reduce both acute

and chronic illness from contaminated food.

To validate the estimates of incidence, a primary and additional estimation method

were used depending on the available data for each pathogen. When the additional approach

estimates were similar to the primary approach, this validated the estimate and gave more

confidence in the data. When the two approaches produced different estimates, this allowed

for a closer examination of the data to determine which source was more reliable and should


be used as the primary method. Using these two approaches gives us confidence that

estimates were created using the best available data for each pathogen. One consequence of

this was a reduction in estimates of incidence for Cryptosporidium spp. and G. lamblia in

2010 when compared with 2000, which reflect a methodological shift from the use of the

1997–1999 WQS data for these pathogens in 2000, to the use of notification data for these

pathogens in 2010. Estimates circa 2010 used the Australian NNDSS for Cryptosporidium

spp., and state data for Victoria for G. lamblia.

By applying the current estimation methods to the data from 1996–2000 (which were

used for the circa 2000 report), there was an increase in estimates of foodborne

Campylobacter spp. and non-typhoidal Salmonella spp. annual incidence by about 40,000

and 11,000 cases respectively from circa 2000 to circa 2010. This estimated increase is

supported by the fact that notifications for both pathogens have been increasing over the 10

year period and higher numbers of hospitalisations were observed in 2010 when compared

with 2000. In addition, the circa 2010 estimates for GBS and other sequelae increased over

this time. These increases are concerning and highlight the importance of government and

industry partnerships to reduce the incidence of specific foodborne pathogens,28,29 as well as

the promotion of food safety education for consumers.

There was a clear decline in estimates for rotavirus in 2010 compared to 2000 reflecting

the success of the immunisation program, which began in 2006, in reducing illness in young

children.36 Similarly, estimated foodborne illness due to hepatitis A declined from 245

foodborne cases per year circa 2000 to 40 foodborne cases per year in 2010, reflecting

improved public health control of this pathogen, such as vaccination programs in Indigenous

communities and other interventions.37 While these vaccination programs were not targeting

foodborne transmission of disease, these declines show the systemic effects that public health

interventions can have and the benefits for food safety due to reduced infection pressure.

Overall foodborne gastroenteritis in Australia from the NGSII decreased when

compared to the NGSI, which consequently reduced the estimate of overall foodborne

gastroenteritis. The reduction in incidence of gastroenteritis may be due to a range of factors,

including biases in the study methodology as the response rate decreased over time, or a true

decrease in incidence of pathogens causing gastroenteritis, such as rotavirus.36

STUDY LIMITATIONS


In a complex study of this type, there are many gaps in data availability and each data

source had strengths and weaknesses. While NNDSS and the OzFoodNet outbreak register

were nationally representative, there may have been differences in the way that jurisdictions

reported or coded their data, and some outbreaks may not always be captured in the

OzFoodNet outbreak register, such as those due to illnesses like hepatitis A and listeriosis,

unless a source is identified.

In contrast, there may have been outbreaks or unusual increases in disease that occurred

during the study period. In 2009, there was a large outbreak of hepatitis A associated with

semi-dried tomatoes across Australia.38 Similarly, outbreaks of salmonellosis and other agents

were routine occurrences.39 It was decided that data would not be adjusted in the estimation

to account for outbreaks or investigated clusters, as these events represent the real situation of

foodborne illness in Australia and form an important part of the burden of disease.

Data from the WQS2,20 were used for pathogens that were not nationally notifiable or

had limited outbreak data (Technical Appendix 3). This study is the best of its kind in

Australia, but is now over 15 years old. The study centred on families with children in the

Melbourne area, and as such may not represent all foodborne illness in Australia. Where

WQS data were used, changes over time and the differing age structure to the general

population were adjusted for. Participation and stool submission in the study may also be an

issue: only about one third of gastroenteritis cases submitted a stool sample, not all ‘known’

pathogens were tested for, and only 17% of stools that were examined had a pathogen

identified.2,20

One aspect of using WQS data is that the estimate of the incidence of ‘other pathogenic

E. coli’ is very high. The WQS reported a high recovery rate from participants’ stool samples

of enteropathogenic E. coli, at least some of which had atypical virulence characteristics.40

Apart from notifiable disease data on Shiga toxin-producing E. coli, there were no other data

available on pathogenic E. coli from other sources. The public health significance of

enteropathogenic E. coli is uncertain, as little is known about sources of infection in the

community. Consequently, the range of uncertainty around the estimates of incidence of

‘other pathogenic E. coli’ is very wide. Given these concerns – and the lack of a second

dataset for comparison – these estimates should be viewed with caution. It is also worth

noting that hospitalisation rates for ‘other pathogenic E. coli’ were low, as this hospitalisation

code is a generic category and laboratories are unlikely to test for pathogenic E. coli as a

cause of diarrhoea.


While the study by Hall et al.22 provided a more appropriate underreporting multiplier

for specific pathogens than was previously available, both the underreporting multiplier and

the outbreak multiplier in the circa 2010 estimates were largely based on non-typhoidal

Salmonella spp. Although pathogens responsible for moderate illness may act similarly to

non-typhoidal Salmonella spp., in reality, multipliers are likely to be illness specific. Ideally,

underreporting and outbreak multipliers would be specific to each pathogen.

Another limitation was the amount of research and data available for the proportion of

diarrhoea that was due to astrovirus, adenovirus, and sapovirus. While they may cause

gastroenteritis commonly, they are rarely tested for and laboratory tests are probably

unreliable. In addition, sapovirus is a newly identified virus and hospitalisations and deaths

could not be estimated as there was no ICD-10-AM code for the illness. The total envelope of

disease resulting from these three pathogens estimated circa 2010 is similar to that estimated

for adenovirus and astrovirus circa 2000, however the data to inform these estimates were

sparse.

This study attempted to account for many weaknesses in the data through the estimation

of uncertainty. In this study, the uncertainty bounds were expanded from 95% credible

intervals to 90% credible intervals. The main reason for doing this was to ensure consistency

with international efforts.6 Wider credible intervals reflected greater uncertainty. When these

intervals were examined, it was identified that over half the uncertainty arose from the

distribution for the foodborne multiplier, with distributions for the underreporting and

pathogen fraction multipliers the next most important sources of uncertainty.

The implication of this is that a massive change in the incidence of foodborne

pathogens would need to be observed in order to see a significant difference in the rate ratio

over time. When comparing rates from 2010 to 2000, only the incidence of hepatitis A was

significantly lower, representing a seven-fold decrease. Studies targeted at high incidence

pathogens—norovirus and ‘other pathogenic E. coli’ in particular—would help to reduce this

uncertainty and improve understanding for disease prevention. Given the significant

uncertainty in estimation methods, there is a need to improve the ability to detect less-

dramatic changes in disease.

POLICY IMPLICATIONS AND FURTHER RESEARCH


Foodborne illness estimates are an important tool for public health policy as they

elucidate the societal burden of foodborne illness, and allow economic assessment to

determine the direct costs of illness from contaminated food. A costing report of the circa

2000 estimates suggested that foodborne illness in Australia costs AUD$125 billion dollars

annually.4 Updating this figure for the circa 2010 estimates is important for understanding the

direct costs of foodborne illness. In addition to the direct costs incurred, foodborne illness is

also a significant public health issue due to the potential costs that food contamination and

food recalls may have on industry. Outbreaks and sporadic illness occur from a variety of

food commodities and many different industries and even whole economies can be affected.

A recent USA study found that most foodborne illnesses from outbreaks can be

attributed to food commodities that constitute a major portion of the diet in the USA.41 As

these types of commodities are frequently consumed, food contamination often leads to

massive recalls, along with industry wide damage. These high costs demonstrate the public

health importance of reducing foodborne illness and provide an opportunity for regulatory

policy to reduce these common illnesses. The study by Painter et al.41 attributes outbreaks to

food commodities in the USA. Similar studies in Australia to identify the food commodities

responsible for the greatest amount of illness, as well as for contributing factors through the

production and supply chain, could complement these estimates of incidence, hospitalisation

and death and provide increased support for targeted interventions in the food industry.

There are several areas where further research is needed to strengthen the estimates and

improve public health policy. Foodborne illness estimate and burden studies can assist with

identifying vulnerable populations. For this, it is important to consider specific sub-studies to

examine the specific foodborne burden affecting sub-populations, such as the young or the

elderly. For these groups, certain pathogens have higher incidence and the ameliorable risk

factors will be different. Another area that is needed to improve the understanding of

foodborne illness in Australia is better validation of hospitalisations admission records and

death certificates for potentially foodborne agents.

Importantly, this report highlights the benefits of a large prospective community cohort

study that follows participants for reports of symptoms of diarrhoea and/or vomiting, similar

to the IID2 study in the UK1 to better identify rates of infectious intestinal disease rates in the

community due to specific pathogens or agents. For pathogens that are not covered by the

national surveillance system and not commonly seen in outbreaks (such as G. lamblia and

‘other pathogenic E. coli’), this report had to rely on the Melbourne WQS2 which is now 15

years old, and has other limitations mentioned previously. A large prospective cohort study


would improve current estimates for infectious gastroenteritis and aid in determining the

burden and causes of specific pathogens in the Australian community, along with

identification of preventable risk factors.

Finally, it would be important to repeat this estimation exercise circa 2020 to examine

changes in disease incidence over time. Ideally, there would be a well-articulated plan to

improve the information base for this assessment.

CONCLUSIONThere are an estimated 4.1 million (90% CrI: 2.3–6.4 million) cases of foodborne

gastroenteritis occurring each year, along with 5,140 (90% CrI: 3,530–7,980) cases of non-

gastrointestinal foodborne illness and 35,840 (90% CrI: 25,000–54,000) cases of sequelae.

The majority of foodborne illness occurs as gastroenteritis, but non-gastrointestinal illness

and sequelae are also important due to the fact that they result in hospitalisations and

occasional deaths. Over time, the incidence of foodborne gastroenteritis decreased slightly,

while salmonellosis and campylobacteriosis had increased. There is a need to improve the

evidence-base in Australia to improve the understanding of foodborne illness by conducting

research into specific pathogens and the overall causes of gastroenteritis through a large-scale

cohort study. The results of this study should assist policy makers to advocate for improved

regulation and control of foodborne illness.


TECHNICAL APPENDIX 1: FURTHER COMPARING FOODBORNE ILLNESS INTERNATIONALLY

INCIDENCE OF INFECTIOUS GASTROENTERITIS

The estimate of total infectious gastroenteritis incidence in the community may be

made using either a prospective cohort study or a retrospective cross sectional study with

recall of episodes of gastroenteritis in the previous weeks. These studies are usually run over

one year to capture any seasonal effects. The prospective cohort study design requires

participants to record symptoms relevant to gastroenteritis in a diary during the study period.

The retrospective cross sectional study design usually involves surveying people by

telephone with regard to their symptoms of gastroenteritis in the recent past.

It is important to note that the two different study designs may not give the same

estimates of population incidence of disease. Retrospective cross sectional studies have

tended to provide higher estimates of gastroenteritis rates than do prospective studies. Cross

sectional studies gave incidence close to one case per person per year in the USA, Canada

and Australia,18,42,43 while prospective cohort studies have resulted in incidence considerably

less than one case per person per year at 0.19 and 0.27 episodes per person-year in two UK

studies,1,44 and 0.28 in the Netherlands.45

In the second UK study, the IID2 study in 2008, both a population based cohort study

and a population based cross sectional study were run deliberately to compare incidence from

the two study designs using the same case definition.1 In the IID2 study, using 7-day recall in

the retrospective telephone study, the incidence was 1,530 cases per 1000 person-years,

which was five times higher than the rate estimated in the prospective cohort study (274 cases

per 1000 person-years). When they used 28-day recall in the telephone study, the incidence

was estimated at 533 cases per 1000 person-years, which was twice as high as the rate

estimated in the prospective cohort study. It has been proposed that retrospective studies tend

to result in higher estimates of morbidity due to ‘recall errors’.44,46 If the recall period is short

and the illness severe, then the event is more likely to be memorable and over-reported;

whereas if the recall period is long, the illness mild and less memorable, it is more likely to

be underreported or even forgotten.

While prospective studies do not tend to have the same issue with recall errors, they

can, however, suffer from the effects of reporting fatigue, which occurs when participants

lose interest in reporting their symptoms, possibly leading to an underestimation of the


disease rate.2,44,47,48 A decline in the reported rate of highly credible gastroenteritis was

observed in two Australian studies as the study progressed.2,48 Hellard et al.2 suggest that one

way to avoid this phenomenon may be to use a shorter observation period with a larger

number of participants.

Although more expensive and longer than retrospective cross sectional studies, cohort

studies generally include specimen collection that allows for direct estimation of the

incidence of some specific pathogens. Hellard et al.2 used a randomised clinical trial with a

cohort of study participants to give incidence rates to 16 known enteric pathogens.

Besides study design, the case definition of gastroenteritis can greatly influence

incidence estimates, as there is no standardised international case definition. When four

different case definitions were applied to a given country, the result was four different

incidence estimates for that country.49 This strongly suggests that valid comparisons cannot

be made between studies that use different case definitions. Incidence of illness and cost

estimates will also be influenced by different case definitions because they rely on estimates

of incidence or number of cases. In the study by Majowicz et al.,49 the different case

definitions were also shown to affect gender specific values and age distributions, the

observed duration of illness (by 20%), and the proportion of cases seeking medical care and

submitting stool samples for testing. The use of a standard symptom-based case definition by

investigators would enable much better inter-country comparisons and global estimates of

gastroenteritis incidence.

A further complication in defining an appropriate case definition for gastroenteritis

arises with regard to respiratory symptoms. When a person presents with concurrent

respiratory symptoms and symptoms of diarrhoea or vomiting, they could be due to

respiratory infections, gastrointestinal infections, or both; however this distinction is rarely

made.50 Analysis of population based studies in Australia, Canada and the USA has shown

that respiratory symptoms occur frequently in persons with acute gastrointestinal symptoms

(diarrhoea, vomiting or both), specifically 29% in Australia, 42% in Canada and 47% in

USA.50 Hence, if the case definition of acute gastroenteritis is adjusted to remove cases with

respiratory symptoms, there is a substantial decrease in the estimates of incidence of acute

gastroenteritis. Therefore, when estimating the incidence of foodborne illness from

population studies, it is important to give consideration to cases with gastrointestinal

symptoms possibly arising from respiratory infections and to make appropriate adjustments.

To date, no international consensus has been reached on whether all or some of such cases

should be excluded from analyses.


In the previous estimation of the incidence of gastroenteritis in Australia, Hall et al.3

used a case definition which referred to moderate-to-severe illness, with ≥3 loose stools or ≥2

episodes of vomiting in a 24 hour period in the previous four weeks. Patients with

concomitant respiratory symptoms were excluded unless they had more severe symptoms of

diarrhoea or vomiting, specifically, ≥4 loose stools or ≥3 episodes of vomiting, in a 24 hour

period in the previous four weeks. In the Mead study5 respiratory symptoms were not

excluded, but an adjustment was made to reduce the number of cases from a population based

cross sectional study by 25% based on data from other studies to account for this.51,52 The

case definition in the more recent study by Scallan et al.6 excluded persons with respiratory

symptoms, which effectively halved the incidence rate of gastroenteritis.

PATHOGENS ASSESSED IN AUSTRALIA CIRCA 2000

There are in excess of 200 different pathogens and agents that may be transmitted by

contaminated food and cause gastroenteritis or other syndromes and conditions, such as

hepatitis and septicaemia.5 Foodborne pathogens include viruses, such as norovirus and

hepatitis A; bacteria, such as non-typhoidal Salmonella spp., Campylobacter spp. and

toxigenic E. coli; along with toxins, such as ciguatoxins and histamines.6,8 The Australian

study by Hall et al.10 included 16 infectious agents causing gastroenteritis, five pathogens or

agents causing acute non-gastrointestinal illness, and four syndromes representing sequelae

of acute foodborne infections. The key rationale for covering different pathogens in this study

included: high incidence, severe outcomes of infection or exposure, availability of data,

strong relationships with specific food vehicles, and expert opinion on importance to food

safety.

ESTIMATING INCIDENCE OF SPECIFIC PATHOGENS

The variation in methods used to estimate incidence of illness due to specific pathogens

is largely data driven. Different methods have been used in different countries and across

different pathogens, depending on what data are available. In many cases there are no

‘definitive’ data that give an absolute value of incidence, and varying adjustments and

assumptions are usually required to estimate the incidence in the population. Data sources

include cohort studies and surveys with laboratory testing of stool or other relevant

specimens to identify pathogens, surveys with collection of human sera to identify antibodies

to particular agents, surveillance systems of individual cases of illness due to various

pathogens, or surveillance of outbreaks and data on hospitalisations and deaths. If there is


more than one data source available then this allows for validation across different methods

of estimation.

Population based prospective cohort studies are used to assess the incidence of all

gastroenteritis and they also allow for stool specimens to be tested from those participants

exhibiting relevant symptoms. This means the pathogens can be identified allowing

estimation of the incidence of specific pathogens. This has given very useful results that have

been used in burden studies in the Netherlands,14,45 in the first study of infectious intestinal

disease (IID1) and in the IID2 study in the UK.1,44 The identification of known pathogens in

specimens from population based cohort studies is often not complete, generally being found

in around 30–50% of specimens. Numbers of cases also tend to be small and confidence

intervals may be wide.

Active or mandatory laboratory based surveillance data maintained by health

departments or other organisations are often used in the absence of cohort data and,

depending on quality, are generally preferentially used above some other data sources. In

countries like the USA and Australia, there may be more than one surveillance system

operating, where there is surveillance to a national notifiable diseases system, and an

enhanced surveillance system focussed on foodborne illness where reporting of cases is

actively followed up, such as FoodNet.5,6,53 The number of cases reported to laboratory based

surveillance is usually less than the number of cases in the community and multiplicative

factors are used to adjust the numbers of reported cases to reach a more realistic estimate of

incidence in the community. Multiplicative factors include adjustment factors to account for

coverage of the population under surveillance (necessary when only a proportion of the

population may be under enhanced surveillance), to adjust for the fact that only a proportion

of cases go to the doctor and of these, only a proportion have a stool test ordered. Of these

not all will have a positive test result depending on the sensitivity of the laboratory test, and

further adjustment may be needed to account for incomplete reporting of positive tests to the

surveillance system. Where only outbreak surveillance data are available, a further

multiplicative factor is required to adjust from the number of cases reported in outbreaks to

the number of individual cases that would have been reported if the pathogen were under

individual surveillance.

ACCOUNTING FOR UNDERREPORTING AND UNDERDIAGNOSIS

The underreporting of illnesses to surveillance is widely recognized and strategies to

deal with this range from simply noting the issue to deriving quantitative factors using


varying levels of data and sophistication. The calculation of community incidence from

laboratory based surveillance data requires multipliers that account for the fact that not all

cases in the community are reported. For this to happen, a person must have symptoms, visit

a doctor, have a stool sample taken, the laboratory test has to be positive and the result

reported to surveillance. Each of these steps can be assigned a probability based on the

proportion of people or tests that proceed to the next step. The overall proportion of

individuals recorded by surveillance is the product of all these component probabilities and

the multiplier is the inverse of that proportion. The sequence of steps is commonly known as

the reporting pyramid. The corresponding multiplicative factors comprise the components

‘population factor’, ‘doctor visit factor’, ‘stool test factor’, ‘laboratory sensitivity factor’,

‘reporting to surveillance factor’ and ‘outbreak factor’. In different studies these factors may

have been estimated from ‘opinion’ or on more substantial evidence from various data

sources.

In recent studies, more attention has been paid to trying to estimate these multipliers in

a more pathogen-specific and evidence-based way and including measures of uncertainty. In

the study by Mead et al.,5 multipliers were based on limited evidence and some multipliers

were simply based on the authors’ opinions. In the more recent USA study,6,7 considerable

effort was put into improving multipliers through the use of data to derive estimates and the

modelling of uncertainty. An underreporting multiplier, comprised of a ‘population factor’,

‘reporting to surveillance factor’ and ‘outbreak factor’; and an underdiagnosis factor,

comprised ‘doctor visit factor’, ‘stool test factor’, and ‘laboratory sensitivity factor’ were

used to account for underreporting. The data for the multiplier that adjusts for passive to

active surveillance were taken from pathogens that have both passive and active surveillance,

including Cryptosporidium spp., STEC, L. monocytogenes and non-typhoidal Salmonella

spp. To estimate the underdiagnosis factor, three population surveys of gastroenteritis were

used to estimate the proportion of people who had bloody diarrhoea and sought medical care

(35%), the proportion who then submitted a stool sample (36%), and the proportion of people

who had non bloody diarrhoea and sought medical advice and then submitted a stool sample

(18% and 19% respectively). For severe invasive illness it was assumed that 100% of cases

would seek medical care and 80–100% would have a specimen test. A factor to account for

the percentage of laboratories that tested for a particular pathogen was derived from FoodNet

and other laboratory surveys and gave pathogen-specific factors of 25% to 100% and factors

to account for sensitivity of the test of 28% to 100%. For the five pathogens with only

outbreak data, the underdiagnosis factor was assumed to be the same as for non-typhoidal


Salmonella spp. The outbreak factor to adjust for outbreak data to active surveillance was

derived from ratios from pathogens with both active surveillance and outbreak surveillance.

In Hall et al.,3 most of the pathogen-specific incidence rates were based on methods

using either direct estimation from cohort data from a study in Melbourne, Victoria,2,20 or

surveillance data adjusted for underreporting using multipliers. The multipliers used to adjust

for underreporting cases in the community to individual surveillance were: 15 (CrI: 5–25) for

moderate illnesses, nine (CrI: 1–17) for bloody diarrhoea, and two (CrI: 1–3) for serious

illnesses. The outbreak factor was estimated from cases infected with non-typhoidal

Salmonella spp. in Victorian data that were reported to outbreak surveillance compared with

the number of cases reported to individual surveillance, a factor of about 14. A method

similar to that used by Voetsch et al.54 was used to calculate the underreporting factor of 15

for moderate illness using data from an unpublished case control study of non-tyhoidal

Salmonella spp. cases in the Hunter region and surveillance data (1997–2000). Comparison

of the incidence of cases of non-tyhoidal Salmonella spp. notified to the estimated incidence

of non-tyhoidal Salmonella spp. from the Melbourne water quality study (WQS)2 also

informed this factor of 15.10 The ‘serious illness factor’ of two was based on the precedent set

by the Mead et al.5 study.

In 2008, new underreporting factors for Australia were estimated together with

uncertainty for non-tyhoidal Salmonella spp. at seven (95% CrI: 4–16), for Campylobacter

spp. at 10 (95% CrI: 7–22), and for STEC at eight (95% CrI: 3–75).22 The multipliers resulted

in estimates of community incidence of these pathogens that were lower than the estimates

used in the last Australian foodborne illness incidence study.3 It is notable that the recent

Australian underreporting factor estimates are similar to the recent UK estimates from the

IID2 cohort study where underascertainment ratios for non-tyhoidal Salmonella spp. and

STEC were nine and seven respectively.1 The underascertianment ratio of five for

Campylobacter spp. in the IID2 study was lower than the Australian underreporting estimate

of 10.

ESTIMATING HOSPITALISATIONS AND DEATHS

Although the incidence of hospitalisations and deaths are important indicators of the

severity of foodborne gastroenteritis, it is not easy to estimate them accurately. This is in

large part because most routine surveillance systems underreport mortality from infectious

diseases. In addition, very little information is collected on illness outcome by pathogen-

specific surveillance.


National data on outpatient visits resulting in hospitalisation, hospital discharges and

death certificates generally underestimate the pathogen-specific incidence. This is because

pathogen-specific diagnoses need to be recorded, health care providers need to order

appropriate diagnostic tests, and coding must be accurate. Scallan et al.6 used surveillance

data to determine the proportion of persons that were hospitalised and the proportion that

died, and then applied these proportions to the estimated number of laboratory confirmed

illnesses. The number of hospitalisations and deaths was then doubled to account for

underdiagnosis of those with unconfirmed illnesses.

Mead et al.5 also accounted for underreporting by doubling the number of

hospitalisations and deaths among reported cases. Multiple cause of death data from the

National Vital Statistics System where acute gastroenteritis was listed as the underlying or a

contributing cause, were used to estimate the number of deaths. Acute gastroenteritis was

defined by the following ICD-10 diagnostic codes: A00.9–A08.5 (infectious gastroenteritis of

known cause), A09 (diarrhoea and gastroenteritis of presumed infectious origin), and K52.9

(non-infectious gastroenteritis and colitis, unspecified). A04.7 (enterocolitis due to

Clostridium difficile) and A05.1 (botulism) were excluded. Mead et al.5 also excluded

diagnoses of infectious enteritis associated with respiratory symptoms or a diagnosis of

influenza (ICD-9-CM code 487).

In Australia, Hall et al.3 used data from the National Hospital Morbidity Database to

estimate the total number of hospitalisations and deaths for each of the pathogens under

investigation. In the Netherlands, de Wit et al.45 used linear regression to combine laboratory

data and hospital registration data to obtain estimates for incidence of hospitalisations for

rotavirus infection, only taking into account those with microbiologically confirmed rotavirus

infection. In Canada, Ruzante et al.55 estimated the number of hospitalised cases and the

fatalities from four years of data from the Canadian Institute for Health Information Hospital

Morbidity Database (HMDB) and the Canadian Vital Statistics-Death Database for five

enteric pathogens and three sequelae. Vaillant et al.,11 Mead et al.5 and Adak et al.12 applied

pathogen-specific case fatality ratios, from selected outbreaks and studies, to the estimated

number of cases. Their estimates do not relate to all directly coded deaths, which may have

led to overestimation of the number of deaths attributable to foodborne infection.

ESTIMATING PROPORTION OF ILLNESSES THAT ARE FOODBORNE

A key component of all studies examining the incidence and burden of foodborne

illness is the estimation of the proportion of disease due to food for different pathogens or


agents.56 This is important, as for many agents, disease may be acquired from exposure to a

range of different sources, including contaminated food or water, infected animals or humans

and contact with contaminated environments.57 Many enteric diseases are faecal-oral spread

and may be transmitted via a range of different modes of transmission.8

Estimating the proportion of cases of disease due to different modes of transmission is

difficult due to the paucity of data and is a weak point of studies assessing foodborne illness

incidence.6,57 Public health investigators have used several different methods for assessing the

fraction of disease that may be due to contaminated foods. Source attribution studies, which

specifically examine the route of transmission, may be focused on different points in the food

chain.57 Different approaches to studying source attribution include: microbiological

approaches, epidemiological approaches, intervention studies, and expert elicitation. All

national-level studies examining the incidence of foodborne illness have incorporated some

element of source attribution to arrive at estimates of disease incidence. Investigators have

used a range of these different approaches (Table T1.1).

An estimate of the proportion of transmission of known pathogens that is due to food is

critical for estimation of the incidence of foodborne illnesses of unknown aetiology, which

comprises the majority of the total burden. Even when using the most sensitive diagnostic

screening tests, a recognised pathogen is recovered from only 30–40% of specimens.1 Only

the UK, USA, Greek and Australian studies estimating the incidence of foodborne illness

have taken gastroenteritis of unknown aetiology into account, following the approach adopted

by the Mead study.3,5,7,12,13

Many foodborne illness incidence studies, including the circa 2000 estimates for

Australia, have relied on expert elicitation to attribute illnesses to foodborne spread, due to

the paucity of other sources of data on the mode of transmission. There have been several

methodological improvements over the years to the expert elicitation framework, including

providing experts with prior information from systematic reviews of the literature and

modelling uncertainty in expert opinions using Monte Carlo simulation. This technique is

fairly widespread in risk analysis studies and involves simulating distributions of components

used in the calculations rather than using point estimates. The end product is a final

distribution that provides a central point estimate and a credible interval.

However, there have been expert elicitations where the findings have proven unusual or

contradictory.58 One Canadian study that used cluster analysis to examine opinions from

experts from different backgrounds found that for L. monocytogenes, one group of experts

had a mean of 8% of human infections were foodborne, compared to another cluster


estimating 84% of human infections were foodborne.58 These differences may result from the

type of expert selected, or from the prior information given to participants in these structured

elicitations.

Expert elicitation should be regarded as a structured way to obtain a consensus opinion,

based on evaluation of all available data. It cannot be expected to provide an unbiased

estimate of the relative importance of different transmission routes. An advantage of expert

elicitation is that it allows for information from different sources to be combined in a single

analytical framework.

ESTIMATING UNCERTAINTY

There are many sources of uncertainty in national incidence of disease studies

including: uncertainty in the proportion of illness caused by specified or unknown pathogens,

uncertainty in the proportion of cases that can be attributed to foodborne transmission,

uncertainty in the level of underreporting of cases of illness, inconsistency in the quality of

incidence data, and value choices made in the DALY formula. Furthermore, norovirus and

other viruses are suspected of being responsible for a significant proportion of the ill-defined

intestinal infections and hence have a major impact on overall estimates, but are poorly

captured by routine surveillance.

It is important to try to express uncertainty quantitatively in order that reported

estimates and intervals convey the message that foodborne illness estimations rely on a

number of assumptions, and on data that are often variable, incomplete or minimal. While the

study by Mead et al.5 did not attempt to quantify uncertainty, a number of other more recent

studies have done so. The main method used to account for uncertainty is through Monte

Carlo simulation.

The study that has the most extensive descriptions of the techniques to account for

uncertainty used is the recent USA estimation of foodborne illness by Scallan et al.6,7 In this

study the rationale for the type of distributions selected is given, and the resultant final

estimates are given as a mean and 90% credible interval. This study provides the most

thorough reasoning for the approach used and also sufficient detail in appendices to replicate

their methods.


Table T1.1: Proportion of illnesses attributed to food for different pathogens from national studies. Note that not all pathogens or agents considered in these studies have been included in the comparison table. Adapted from Havelaar et al.59

Characteristics of national studies

Mead5 Adak12 Vaillant11 Hall3 Havelaar59 Scallan6 Gkogka13 Thomas17

Country USA UK France Australia Netherlands USA Greece CanadaPeriod 1990s 1992–2000 1997–2000 2000 2006 2000–2008 1996–2006 2000–2010Data sources* E/O/R O E/O/R/CC/L E/O/R/L E E/O/R/L L E/O/R/LTravel-related cases Included Excluded Included Excluded Included Excluded Included ExcludedCampylobacter spp. 80% 80% 80% 75% 42% 50% 55% 68%Shiga toxin-producing E. coli

50% 65% 51% NA

O157 85% 63% NA NA 40% 68% NA NANon-O157 85% NA NA NA 42% 82% NA NA

Listeria monocytogenes 99% 99% 99% 98% 69% 99% 99% 84%Salmonella spp., non-typhoidal

95% 92% 95% 87% 55% 94% 95% 80%

Bacillus cereus toxin 100% 100% 100% 100% 90% 100% NA 100%Clostridium perfringens 100% 94% 100% 100% 91% 100% NA 100%Staphylococcus aureus 100% 96% 100% 100% 87% 100% NA 100%Hepatitis A virus 50% NA 5% 10% 11% 7% 8% 7%Norovirus 40% 11% 14% 25% 17% 26% NA 31%Rotavirus 1% 3% NA 2% 13% <1% NA 1%Cryptosporidium spp. 10% 6% NA 10% 12% 8% 5.6% 9%Giardia lamblia 10% 10% NA 5% 13% 7% 10% 7%Toxoplasma gondii 50% NA 50% 35% 56% 50% 50% 50%*Data sources: E=Expert opinion; O=Outbreak reports; R=Reported cases; CC=Case control studies; L=literature searchNA – Not applicable


TECHNICAL APPENDIX 2: NATIONAL GASTROENTERITIS SURVEY IIThe National Gastroenteritis Survey II (NGSII) was conducted by the Australian

Government Department of Health, the New South Wales Food Authority and the National

Centre for Epidemiology and Population Health (NCEPH) in 2008–2009. The main aim of

the survey was to estimate the incidence and public health impact of gastroenteritis in

Australia.

For the NGSII, a stratified sample was collected in each state and territory, with a larger

sample collected from New South Wales and the Australian Capital Territory. During the 12

months of the study, the response rate for the telephone-based survey was 49.1%

(7590/15456), which represents the proportion of successful interviews for all eligible

residential households contacted. Overall, 50.9% (7867/15456) of eligible households

refused to participate in the survey. The lowest response rate was 47.3% for Victoria and the

highest was 54.0% for Tasmania. There were a total of 7,590 interviews conducted. Twelve

respondents did not provide an age during the interview and were excluded from analysis, as

data were unable to be weighted. Data were weighted to account for the two-stage sampling

design for jurisdiction and household and data were post-stratified by age category and sex

using 2006 census data to make data more representative of the Australian population.

A total of 555/7578 (7.3%) respondents reported experiencing diarrhoea or vomiting in

the previous four weeks, equating to 25 million episodes of vomiting or diarrhoea each year

in Australia, compared to 29 million cases in the 2001 National Gastroenteritis Survey

(NGSI) (Table T2.1). The primary definition of gastroenteritis was met by 341 cases, which

equates to 15.9 million cases of gastroenteritis in Australia in 2008–09, compared to 18.9

million cases of gastroenteritis in the NGSI. The weighted incidence of gastroenteritis

meeting the case definition in the NGSII was 0.74 (95% confidence interval [CI]: 0.64–0.84)

episodes per person per year.

DEMOGRAPHIC FEATURES AND SEASONALITY

In the NGSII, children under five years old reported the most gastroenteritis and

people over the age of 65 years old were nine times less likely to report gastroenteritis than

people aged less than five years old. Only people aged between 20–29 years old were similar

to children under the age of five years old in regards to their experience of gastroenteritis

(Figure T2.1)

Foodborne Illness in Australia Circa 2010Page 43

Table T2.1: Weighted number of cases of gastroenteritis and incidence meeting various case definitions in Australia in one year, NGSI & NGSII

Definition NGSII-2008 million cases

per year (95% CI)

NGSII-2008 cases per person per year (95%

CI)

NGSI-2001β

million cases per year (95% CI)

NGSI-2001β

cases per person per year (95%

CI)Gastroenteritis meeting case definitionα

15.9 (13.7–18.0) 0.74 (0.64–0.84) 18.9 (16.6–21.1) 0.97 (0.86–1.09)

Any diarrhoea or vomitingγ

24.8 (22.2–27.4) 1.16 (1.04–1.28) 29.0 (26.3–31.7) 1.49 (1.36–1.63)

Diarrhoea onlyδ

14.1 (12.1–16.1) 0.66 (0.57–0.76) 17.2 (15.1–19.3) 0.89 (0.78–1.00)

α–Case definition of ≥3 loose stools and/or ≥2 vomits in 24 hours that was not due to a known non-infectious cause. If respiratory symptoms were also present, a case was required to experience ≥4 loose stools and/or ≥3 vomits in 24 hours and no non-infectious cause. γ–Includes any person reporting symptoms of diarrhoea or vomiting and does not exclude cases due to non-infectious causes.δ–Includes any person reporting ≥3 loose stools in 24 hours.β–Data for NGSI-2001 analysed using same data weighting procedure as NGSII-2008 .

0

0.2

0.4

0.6

0.8

1

1.2

1.4

1.6

1.8

2

0-4 yo 5-9 yo 10-19 yo 20-29 yo 30-39 yo 40-49 yo 50-59 yo 60-64 yo 65+

Age category (years)

Epi

sode

s pe

r per

son

per y

ear

Females Males

Figure T2.1: Weighted annual incidence of gastroenteritis by age group and sex in Australia, NGSII

People living in rural areas or on farms were significantly less likely to experience

gastroenteritis (adjusted odds ratio 0.57; 95% CI 0.35–0.93). There was no significant

difference in the proportion of respondents reporting gastroenteritis by Indigenous status

(Table T2.2). There was also no significant difference in the proportion of respondents


reporting gastroenteritis by sex, state or territory, level of education or income, size of

household, or whether a person had health insurance or not.

Table T2.2 Four week period prevalence of gastroenteritis by demographic features in Australia, NGSII

Characteristic No. surveyed

Prevalence in previous four weeks (No. with gastroenteritis)

Prevalence in previous four weeks (Crude %)

Prevalence in previous four weeks [Weighted % (95%CI)]

SexFemale 4554 199 4.4 5.2 (4.3–6.2)Male 3024 142 4.7 6.2 (5.0–7.5)Age category (years)0–4 249 42 16.9 12.1 (7.5–16.8)5–9 249 16 6.4 6.3 (3.0–9.7)10–19 497 20 4.0 5.6 (3.0–8.2)20–29 507 47 9.3 9.1 (6.1–12)30–39 802 53 6.6 6.1 (4.1–8.1)40–49 1064 54 5.1 6.1 (4.2–8.0)50–59 1374 52 3.8 3.5 (0–8.2)60–64 755 25 3.3 2.7 (0–6.0)65+ 2081 32 1.5 1.2 (0–5.9)Geographical distribution by state and territoryNSW/ACT 2367 113 4.8 6.1 (4.9–7.3)NT 869 53 6.1 7.2 (5.1–9.2)Qld 869 42 4.8 6.5 (4.5–8.5)SA 870 33 3.8 5.1 (3.3–7.0)Tas. 868 38 4.4 5.2 (3.4–7.0)Vic. 867 34 3.9 5.2 (3.3–7.1)WA 868 28 3.2 4.5 (2.6–6.3)Indigenous statusNon-indigenous 7406 330 4.5 5.7 (4.9–6.5)Indigenous 161 10 6.2 5.1 (0.8–9.4)Highest level of education in householdPrimary 106 4 3.8 5.4 (0–13.3)Years 7–10 1714 62 3.6 5.4 (3.7–7.2)Years 11–12 1425 56 3.9 4.9 (3.3–6.6)Apprenticeship 562 29 5.2 6.8 (3.6–10)Diploma/certificate 1298 55 4.2 5.2 (3.4–6.9)University 2222 124 5.6 6.2 (4.8–7.7)Household income≤$25000 1641 46 2.8 4.4 (2.7–6.2)$25,000 to <50,000 1399 62 4.4 6.1 (4.2–8.0)$50,000 to <100,000 2024 108 5.3 6.3 4.9–7.8)≥$100,000 1452 81 5.6 5.8 (4.1–7.5)Unknown income 1062 44 4.1 5.0 (3.2–6.9)Health insurance status


Characteristic No. surveyed

Prevalence in previous four weeks (No. with gastroenteritis)

Prevalence in previous four weeks (Crude %)

Prevalence in previous four weeks [Weighted % (95%CI)]

Health insurance 4307 187 4.3 5.2 (4.2–6.1)No insurance 3183 151 4.7 6.4 (5.1–7.6)Unknown insurance 88 3 3.4 6.0 (0–13.1)Urban or rural locationUrban/town 6411 304 4.7 6.1 (5.2–7.0)Rural/remote 1138 36 3.2 3.1 (1.7–4.5)Number of people in household1 person 1736 47 2.7 3.1 (1.9–4.3)2 people 2768 100 3.6 5.6 (4.1–7.0)3 people 1064 66 6.2 6.4 (4.4–8.4)4 people 1226 80 6.5 6.8 (4.9–8.6)5 people 519 36 6.9 7.7 (4.8–10.5)6 or more people 231 12 5.2 3.2 (0.7–5.7)

SYMPTOMS AND DURATION

The most common symptoms amongst cases were diarrhoea (87%), followed by loss

of appetite (75%), nausea (66%) and stomach cramps (63%) (Table T2.3). Approximately

one in four cases reported upper respiratory symptoms of sore throat, cough or runny nose.

These cases fulfilled the stricter criteria of at least four loose stools or three vomits within a

24 hour period. Blood in stools was reported by 3.5%. The majority of cases had vomiting

and/or diarrhoea for one or two days, and 6.5% of cases experienced symptoms for at least a

weak. The estimated median duration of gastroenteritis was two days (interquartile range 1–3

days) and a mean of 3.5 days (95% CI: 2.9–4.1). Cases who experienced vomiting either in

isolation or in combination with diarrhoea tended to have a shorter duration of symptoms.

HEALTHCARE SEEKING BEHAVIOUR

Twenty-eight per cent of respondents who reported gastroenteritis sought some kind of

health professional advice or treatment for their symptoms (Table T2.4). A total of 94 cases

visited at least one healthcare facility for their illness, which would equate to 4.77 million

(95% CI 3.55–5.99 million) visits each year, with 2.71 million (95% CI: 1.8–3.62 million) of

these visits to a doctor.

Cases with gastroenteritis lasting two or more days were more likely to see a doctor

compared to those who were ill for one to two days. The presence of vomiting and ear ache

were significant predictors of cases going to see a doctor, as well as indigenous status (odds


ratio (OR) 7.02, 95% CI: 1.38–35.8). People with stomach cramps were less likely to seek

medical attention.

Of the cases that had diarrhoea only and saw a doctor, approximately 24% were asked

to submit a stool specimen. Of those who were asked to submit a stool sample, 11 out of 12

cases subsequently submitted a specimen. Duration of illness of five days or more was

associated with higher likelihood of a case submitting a specimen (OR 4.4, 95% CI: 1.96–

9.87).

Around 37% of cases reported taking at least one medication to treat or relieve

symptoms. This equates to about 5.5 million people taking medication for gastroenteritis

each year. Approximately 6% of those taking medications were prescribed antibiotics, which

after weighted analysis equates to an estimated 520,944 (95% CI: 54,699–987,189) courses

of antibiotics that are prescribed in Australia for treatment each year.

Table T2.3: Unweighted and weighted proportion of cases with gastroenteritis reporting different symptoms in Australia, NGSII (n=341)

Symptom No. reporting

(unweighted)

Crude % (unweighted)

Missing (unweighted)

Weighted % (95% CI)

Diarrhoea 298 87.4 0 84.1 (78.7–89.5)Bloody stools 12 3.5 6 2.9 (0.8–4.9)Loss of appetite 256 75.1 2 78.4 (72.7–84.1)Nausea 225 66.0 6 66.7 (59.9–73.5)Stomach cramps 216 63.3 9 64.5 (57.7–71.4)Headache 142 41.6 13 45.3 (38.1–52.6)Fever or chills 151 44.3 6 50.0 (42.9–57.1)Vomiting 161 47.2 0 49.8 (42.7–56.8)Muscle/body aches 135 39.6 16 43.4 (36.3–50.6)Respiratory symptoms 87 25.5 3 29.2 (22.6–35.9)Stiff neck 51 15.0 19 14.7 (9.8–19.6)Ear ache 22 6.5 1 7.0 (3.7–10.3)

Table T2.4: Number of gastroenteritis cases seeking health careα in Australia, NGSII (n=94)

Health facility Number visiting %Emergency/casualty 12 3.5Doctor’s surgery/health clinic 53 15.5Pharmacy 49 14.4Other 9 2.6α Some cases may have sought care from more than one health facility.


MISSED WORK OR ACTIVITIES

Gastroenteritis had a considerable impact on cases’ work, school and recreational

activities in the survey, with 65% reporting that their illness interfered for a median of one

day (range 1–30). Extrapolation from the data indicates that about 1.19 million days were

lost from paid work each month, with 71% of instances where a person missed paid work due

to the person being ill themselves and 29% due to another person having to care for a person

with gastroenteritis.


TECHNICAL APPENDIX 3: DATA SOURCES

DATA SOURCES USED TO CALCULATE COMMUNITY INCIDENCE

Here, the data sources used to calculate community incidence of the 23 pathogens

included in this assessment are described.

DATA SOURCES

There were three main sources of data: notifiable surveillance at the national or state

and territory level, other surveillance through the OzFoodNet outbreak register, and estimates

of incidence using the Australian gastroenteritis study60 together with cohort studies such as

the Water Quality Study (WQS).2,20 The data source and estimation approach used for each

pathogen is explained in Table T3.1.

Table T3.1: Data sources and estimation approach used for each pathogen or syndrome

Pathogen or syndrome Data source Estimation approachCampylobacterSalmonella, non-typhoidalSalmonella TyphiShigellaCryptosporidiumHepatitis AListeria monocytogenes

National Notifiable Diseases Surveillance System (NNDSS)

Notifiable surveillance approach

Giardia lambliaShiga toxin-producing Escherichia coli (STEC)Vibrio parahaemolyticusYersinia enterocolitica

State and territory surveillanceNotifiable surveillance

approach

Other pathogenic E. coliAdenovirusAstrovirusNorovirusRotavirusSapovirus

Water Quality Study 2,20

andNational Gastroenteritis Survey

II 60

Pathogen fraction approach

Bacillus cereusClostridium perfringensStaphylococcus aureusCiguateraScombrotoxicosis

OzFoodNet outbreak registerOther surveillance

approach


Pathogen or syndrome Data source Estimation approachToxoplasma gondii USA seroprevalence study 21 Special calculations

NOTIFIABLE SURVEILLANCE: NNDSS AND STATE AND TERRITORY NOTIFICATIONS

The National Notifiable Diseases Surveillance System (NNDSS) provides national data

for diseases that are notifiable in Australia, such as salmonellosis, shigellosis and

cryptosporidiosis. Some diseases are notifiable in some states and territories, but not in

others; for example, campylobacteriosis is not notifiable in New South Wales, but is

notifiable in all other states and territories. In these cases, notification data were used for the

available states and territories and include a population adjustment multiplier to estimate

national notification rates (Technical appendix 4). In each case, the total number of

confirmed notifications is used for all available years over the period 2006–2010.

Additionally, further data were requested through Communicable Diseases Network

Australia (CDNA), who own the NNDSS data, to determine the proportion of cases that were

acquired in Australia. Details of the use of these data are described below under the section

‘Domestically acquired multiplier’ (Technical Appendix 4).

OTHER SURVEILLANCE: OZFOODNET OUTBREAK REGISTER

The OzFoodNet outbreak register includes all outbreaks of foodborne gastroenteritis

identified from 2000 and this study used data over the period 2006–2008. The register

provides data on the number of individuals ill in each outbreak, the pathogen identified, and

the total number of individuals with laboratory confirmed illness in each outbreak.

NATIONAL GASTROENTERITIS SURVEY II 2008

The NGSII was a nationally representative telephone survey conducted February 2008

to January 2009 to estimate gastroenteritis incidence in Australia. It provides age-specific

rates of gastroenteritis in the community (Technical Appendix 2).

RESEARCH STUDIES

Australian and international cohort studies were used to assess the proportion of

gastroenteritis that was due to specific pathogens. A key study was the 1997 WQS, which

was a double-blinded, randomised, controlled trial of families conducted in Melbourne,

Australia, between September 1997 and February 1999.2,20 Six hundred families were


allocated to receive either real or sham water treatment units installed in their houses and

study participants reported any gastroenteritis symptoms weekly. The study provides testing

data on 795 faecal specimens identifying pathogens causing gastroenteritis, and these data

were used to calculate a pathogen fraction multiplier for included pathogens. As there was no

significant difference in incidence of gastroenteritis in control and experimental families, the

study found that waterborne pathogens do not play a major role in gastroenteritis in

Melbourne.2


TECHNICAL APPENDIX 4: APPROACHES FOR CALCULATING COMMUNITY INCIDENCE

MAIN APPROACHES

Three main approaches were used for calculating the incidence of illness in the

community. Where possible, a primary and secondary approach were applied to each

pathogen to provide a cross-check of the results. The three approaches were based on the

source of the data as:

1. Notifiable surveillance approach using data from NNDSS or state and territory notifications;

2. Pathogen fraction approach using data from the NGSII together with cohort studies, such as the WQS;

3. Other surveillance approach using data from the OzFoodNet outbreak register, or from hospitalisations.

These approaches were considered to form a hierarchy, with the notifiable surveillance

approach used by preference, and outbreak data used only when other sources were not

available. For each approach, the final estimate is produced from a statistical model that

incorporated uncertainty in case numbers and in multipliers using probability distributions.

That is, at each stage of the calculation, the estimate was represented by a probability

distribution, and the final estimates and credible intervals were computed from this

distribution.

In each approach, input data arises from specific data sources (discussed above), or

from multipliers that are listed below. Three main input distribution types were used:

empirical, pert, and lognormal.

EMPIRICAL DISTRIBUTION

Source distributions on the number of cases are typically represented by an empirical or

discrete distribution driven by the data. For example, if the number of cases notified to

NNDSS for the years 2006–2010 were 15416, 16980, 15539, 16075 and 16967, this would be

represented as a discrete distribution with 20% of the probability mass at 15416, 20% of the

probability mass at 16980, and so on. This use of empirical distributions for such data was

used previously by Scallan et al.,6 and avoids assumptions about the expected shape of this

distribution.


PERT DISTRIBUTION

The PERT distribution is widely used for expert elicitation and risk assessment studies.

It is based on the beta distribution, and within @Risk (risk analysis software), can be

specified either using a minimum, maximum and modal value, or by three percentile points,

such as a median value and 95% credible intervals. This distribution is used widely in this

analysis, as it allows for asymmetric distributions, and can be easily produced from many

data sources including expert elicitation data.

LOGNORMAL DISTRIBUTION

When re-calculating underreporting multipliers, the PERT distribution did not

adequately capture the shape of these multiplier data. A lognormal distribution was adopted

instead, as the distribution providing the best fit as measured by @Risk, and demonstrating

an improved fit on the normal distribution used previously.22


TECHNICAL APPENDIX 5: MULTIPLIERS

METHODS USED TO CALCULATE COMMUNITY INCIDENCE

MULTIPLIERS

Approaches to calculating foodborne illness used key multipliers to either scale up

(surveillance approaches) from detected cases to the full community incidence or scale down

(pathogen fraction approach) from the envelope of all gastroenteritis to the proportion that

was due to specific pathogens. Here, the multipliers used in this approach and the data that

were used to derive these multipliers are described.

POPULATION ADJUSTMENT MULTIPLIER

This multiplier was used where notifiable surveillance data were not available for all

states and territories in Australia, and was necessary to scale up the number of infections

according to the proportion of the population covered by surveillance. For example,

campylobacteriosis is notifiable in all states and territories except New South Wales. In this

example, the total number of cases was adjusted for the remaining states and territories by a

population adjustment multiplier of 1.5 to approximate the total number of cases that would

be expected if all states and territories undertook notifiable surveillance of

campylobacteriosis.

DOMESTICALLY ACQUIRED MULTIPLIER

For some pathogens, a proportion of cases acquired their infections overseas. As data

from the WQS used for the pathogen fraction calculations were centred on families, it was

assumed all these incident cases were domestically acquired. For Campylobacter spp.,

Cryptosporidium spp., hepatitis A, L. monocytogenes, non-tyhoidal Salmonella spp.,

Salmonella Typhi, Shigella, and STEC, the domestically acquired multiplier was calculated

from NNDSS data on the proportion of cases that acquired their infection within Australia.

These data contained a number of missing entries, varying by pathogen, state or territory and

year, with the most complete data for Victoria and Western Australia. Four methods were

considered for adjusting for this missing data:

1. Extrapolate travel patterns from Western Australia to the Northern Territory and travel patterns from Victoria to all other states and territories;


2. Extrapolate travel patterns from Western Australia to both the Northern Territory and Queensland, and travel patterns from Victoria to all other states and territories;

3. Discard all missing data and calculate the proportion of cases acquired in Australia for the existing data only;

4. Assume all unidentified cases are domestically acquired.

Method 1 was adopted as the primary approach, and the other methods were used as a

comparison and to identify a range for the multiplier. Specifically, the median estimate was

made using all five years of data combined, while the minimum and maximum value reflects

the largest and smallest proportion estimated by all four methods over each year of 2006–

2010. Table T5.1 presents the resulting parameters for the PERT distribution, including

median value, minimum and maximum, together with the estimations used circa 2000. For

Cryptosporidium spp., non-tyhoidal Salmonella spp., and Shigella, estimates on the full data

over 2006–2010 using methods 1 and 3 were reassuringly similar, while the expanded ranges

reflect the yearly variability and sensitivity to missing data. Larger differences are seen for

hepatitis A, Salmonella Typhi, and STEC. There were very few missing data for hepatitis A

and Salmonella Typhi which raises confidence in these estimates. Only zero to two overseas

acquired cases of STEC were recorded per year, and this is reflected in the higher estimate of

domestically acquired infection for this pathogen. This multiplier was also used for

calculations of hospitalisations and deaths for ‘other pathogenic E. coli’.

Estimates for the domestically acquired multiplier for G. lamblia were made using

Victorian data over 2006–2009,61-64 using the total proportion to derive the median and the

variability over years to give a range. Domestically acquired multipliers for

V. parahaemolyticus and Y. enterocolitica were calculated from Western Australian data in a

similar manner using OzFoodNet Annual Reports from 2006–2010. Given the higher rate of

overseas acquired infections in Western Australia as compared with other jurisdictions, the

proportion from overseas was reduced for other states and territories using a multiplier of

0.72 based on data for non-tyhoidal Salmonella spp. Even with this adjustment, the

multiplier for V. parahaemolyticus is much lower than that used in the USA suggesting a

greater proportion of overseas-acquired cases in Australia;6 more information on the

behaviour of this pathogen in states and territories outside Western Australia would be

valuable to confirm these results.


Finally, it was assumed that all cases of adenovirus, Bacillus cereus, ciguatera,

C. perfringens, L. monocytogenes, norovirus, rotavirus, scombrotoxicosis, Staphylococcus

aureus, and T. gondii were acquired in Australia. Domestically acquired multipliers were not

needed for the remaining pathogens (astrovirus and sapovirus) for which incidence was

calculated using the pathogen fraction approach, and that do not have specific codes to

calculate hospitalisations and deaths.

Table T5.1: Circa 2010 estimates and ranges for the proportion of infections that were domestically acquired compared to the previously published estimates circa 2000, where available

Pathogen Estimate circa 2010 (range)

Estimate circa 2000 65

Adenovirus 100% (100% – 100%)Bacillus cereus 100% (100% – 100%)Campylobacter spp. 97% (91% – 99%) 96%Ciguatera 100% (100% – 100%)Clostridium perfringens 100% (100% – 100%)Cryptosporidium spp. 97% (92% – 99%)Giardia lamblia 85% (84% – 89%)Hepatitis A 58% (42% – 77%)Listeria monocytogenes 100% (100% – 100%)Norovirus 100% (100% – 100%)Other pathogenic E. coli 99% (93% – 100%)Rotavirus 100% (100% – 100%)Salmonella, non-typhoidal spp.

85% (70% – 95%) 92%

Salmonella Typhi 11% (2% – 25%)Scombrotoxicosis 100% (100% – 100%)Shigella 70% (45% – 84%) 60%Staphylococcus aureus 100% (100% – 100%)Shiga toxin-producing E.coli

99% (93% – 100%) 79%

Toxoplasma gondii 100% (100% – 100%)Vibrio parahaemolyticus 18% (0% – 33%)Yersinia enterocolitica 90% (80% – 100%) 98%


UNDERREPORTING MULTIPLIER

Only a fraction of community cases visit a health professional, have a sample taken and

have their illness recorded in surveillance data. Using data from a 2008 paper by Hall et al.,22

underreporting multipliers were estimated based on lognormal distributions of seven (95%

CrI: 4–14) for non-tyhoidal Salmonella spp., 10 (95% CrI: 6.5–18.5) for Campylobacter spp.,

and eight (95% CrI: 3–18.5) for STEC. Where underreporting multipliers were needed for

other pathogens, the non-tyhoidal Salmonella spp. multiplier was applied except in the case

of pathogens leading to very severe illness (hepatitis A, L. monocytogenes, and

Salmonella Typhi) where the underreporting multiplier was assumed to be two (95% CrI: 1–

3). Details of the choice of multiplier for each pathogen are provided in Technical Appendix

11.

FOODBORNE MULTIPLIER

For most pathogens, the proportion of illness that is foodborne was estimated using data

from Delphi based expert elicitations. For nine pathogens, a 2009 elicitation was used, and

for another eight, a similar 2005 elicitation was used.23,35 The 2009 elicitation was informed

by systematic reviews for each pathogen that included scientific literature, reports and

surveillance data. The foodborne multiplier for sapovirus was extrapolated from elicited

norovirus estimates, and used best judgement assumptions for three additional viruses and the

two marine biotoxins. Refer to table T5.2 for a listing of pathogens, multipliers and the data

source for each. A comparison of these estimates with those used in prior studies is provided

elsewhere.23 Expert elicitation data from 2009 included a best estimate and 90% CrI for

Campylobacter spp., C. perfringens, STEC, ‘other pathogenic E. coli’, non-tyhoidal

Salmonella spp., Shigella, norovirus, hepatitis A, and L. monocytogenes. A PERT distribution

was fitted to each expert’s assessment, fitting the best estimate as the median and setting the

90% CrI where possible. In a few cases, a PERT distribution could not be fitted in this way,

and the best estimate had to be adjusted to be the mode of the distribution (if the median point

was too close to an upper bound or lower bound), or an interval bound had to be adjusted to

be a min or max if the PERT distribution led to values outside the interval 0 to 1. A

combined empirical distribution was calculated by computing the point-wise mean value of

individual uncertainty distribution for each expert. The median, 5th and 95th percentiles of


this empirical distribution were then used to describe a final PERT distribution that was input

into the relevant @Risk spreadsheet.

The 2005 questionnaire provided a best estimate from participants. To include uncertainty in

this estimate, a 90% credible interval was generated about each estimate, assuming an upper

bound 10 percentage points higher and a lower bound 10 percentage points lower. For

example, an estimate of 30% foodborne was modelled as a PERT distribution with median as

0.3, 95% bound 0.4, and 5% bound 0.2. The exception to this was where estimates were too

close to zero (or one) for this method. In these cases, symmetric estimates half the distance

from zero (or one) were assumed. That is, an estimate of 5% foodborne was modelled as a

PERT with median as 0.05, 5% bound as 0.025 and 95% bound as 0.075. The combined

distribution was calculated as for the expert elicitation data. The 2005 expert elicitation did

not achieve consensus for some pathogens; in particular, best estimates ranged from 2%–95%

for S. Typhi, 5%–100% for V. parahaemolyticus, and 33%–90% for Y. enterocolitica. Given

the variability arising from these expert data, the sensitivity of the results to the choice of

distribution was tested by simulating the full empirical distribution of the foodborne

multiplier for each of these pathogens, and comparing estimates of foodborne illness with

those using the PERT distribution. In general, median estimates were little changed, but

credible intervals were a little wider under the empirical distribution. The largest change was

for Y. enterocolitica, where the estimate of domestically acquired foodborne illness was

1,150 (650–1,950) using a PERT distribution and 1,100 (350–2,050) using empirical data.

OUTBREAK MULTIPLIER

For pathogens not captured by notifiable surveillance or by cohort studies, data from

outbreaks were used in the other surveillance approach. Only a fraction of cases are

associated with outbreaks. The outbreak multiplier adjusted for this to estimate the total

number of cases that would be captured if notifiable surveillance was in place for that

pathogen. Many of the pathogens for which this method was used have a short duration of

illness, and thus low rates of laboratory confirmation. To adjust for this, the multiplier was

calculated based on total number of ill (but not necessarily lab confirmed) cases associated

with a confirmed outbreak (where laboratory confirmation of at least one case or of a food

source has been occurred). Non-typhoidal Salmonella was chosen as the reference pathogen

for the outbreak multiplier as it has the most complete data. The outbreak multiplier was

calculated as the ratio of the number of ill cases in outbreaks of non-tyhoidal Salmonella spp.


to the total number of laboratory confirmed domestically acquired cases of non-tyhoidal

Salmonella spp. in the NNDSS for the same year. For example, in 2008 there were 8, 316

laboratory confirmed cases of non-tyhoidal Salmonella spp. in NNDSS, of which 85%

(range: 70–90) were assumed to be acquired in Australia. The total number of ill cases

associated with non-tyhoidal Salmonella spp. outbreaks in 2008 was 524, giving an outbreak

multiplier of around 13.5 for this year. Extending this approach to calculate multipliers for

each year from 2006–2008, and for data for all years combined, an outbreak multiplier of 14,

with range 5–20 is estimated.

Table T5.2: Estimates of the foodborne multiplier with 90% credible interval using PERT distributions for each of the 23 pathogens used in the circa 2010 study*Pathogen/Illness Foodborne multiplier with

90% CrIData source 23,35

Adenovirus 0.02 (0.01–0.03) Assumption

Astrovirus 0.02 (0.01–0.03) Assumption

Bacillus cereus 1.00 (0.98–1.00) 2005 EE as PERT

Campylobacter spp. 0.77 (0.62–0.89) 2009 EE as PERT

Ciguatera 1.00 (1.00–1.00) Assumption

Clostridium perfringens 0.98 (0.86–1.0) 2009 EE as PERT

Cryptosporidium spp. 0.10 (0.01–0.27) 2005 EE as PERT

Other pathogenic E. coli 0.23 (0.08–0.55) 2009 EE as PERT

Giardia lamblia 0.06 (0.01–0.50) 2005 EE as PERT

Hepatitis A 0.12 (0.05–0.24) 2009 EE as PERT

Listeria monocytogenes 0.98 (0.90–1.00) 2009 EE as PERT

Norovirus 0.18 (0.05– 0.35) 2009 EE as PERT

Rotavirus 0.02 (0.01–0.03) Assumption

Salmonella, non-typhoidal 0.72 (0.53–0.86) 2009 EE as PERT

Salmonella Typhi 0.75 (0.02–0.97) 2005 EE as PERT

Sapovirus 0.18 (0.05–0.35) Norovirus multiplier

Scombrotoxicosis 1.00 (1.00–1.00) Assumption

Shigella 0.12 (0.05–0.23) 2009 EE as PERT


Pathogen/Illness Foodborne multiplier with 90% CrI

Data source 23,35

Staphylococcus aureus 1.00 (0.95–1.00) 2005 EE as PERT

Shiga toxin-producing E. coli 0.56 (0.32–0.83) 2009 EE as PERT

Toxoplasma gondii 0.31 (0.04–0.74) 2005 EE as PERT

Vibrio parahaemolyticus 0.75 (0.05–0.96) 2005 EE as PERT

Yersinia enterocolitica 0.84 (0.28–0.94) 2005 EE as PERT

*EE = Expert elicitation

GASTROENTERITIS MULTIPLIER

For pathogens captured by cohort studies such as the WQS,2,20 a proportion of all

gastroenteritis cases was attributed to that pathogen using the pathogen fraction approach.

The first step of this approach was to determine the total incidence of gastroenteritis. To do

this, the NGSII study was used to estimate the total number of gastroenteritis episodes per

person per year, weighted by the Australian population. This estimate served to provide a

gastroenteritis multiplier, which was then multiplied by the total Australian population for the

years 2006–2010 to give the estimated number of cases of gastroenteritis for each year. The

gastroenteritis multiplier was modelled as an alternative PERT distribution with median 0.74

and 95% confidence interval (0.64–0.84), based on the estimates and uncertainty intervals

estimated by the NGSII study.

PATHOGEN FRACTION MULTIPLIER

The pathogen fraction multiplier attributed a proportion of the total number of

gastroenteritis episodes to particular pathogens. The primary data source for this was the

WQS.2,20 While data from the UK infectious intestinal disease (IID2) study1 was also used as

a comparator, the WQS gave the most reliable picture of the incidence of illness due to

different pathogens in Australia. The data from the study were age-adjusted (using age

ranges 0–4, 5–14, 15+ years) to the Australian population (circa 2010) take account of the

higher numbers of children in the WQS. For example, the raw data for adenovirus in the

WQS was nine positive samples from a total of 713 samples taken from participants with a

highly credible episode of gastroenteritis. However, eight of those positives were from

participants aged 0–4 years old, an age group over sampled in the study. Using data on the


incidence of gastroenteritis by age from the NSGII study, and the Australian population as a

reference, age-adjusted estimates were calculated for each pathogen based on the WQS data.

For example, for adenovirus, an estimate of four samples positive for adenovirus from 713

gastroenteritis episodes was derived. This resulted in a pathogen fraction multiplier of 0.0056

(95% CI: 0.0015–0.0143), which was then modelled in @Risk using an alternative PERT

distribution. Note that the pathogen sheets provided in Technical Appendix 11 provide the

age adjusted estimates for each pathogen, so will differ slightly from studies reporting

findings of the WQS.

Finally, no Australian cohort study that gave estimates of prevalence of astrovirus or

sapovirus for all age groups could be found. Instead, the pathogen fraction multiplier from

the WQS for adenovirus and norovirus, together with cohort data from children66,67 was used

to calculate multipliers relating astrovirus to adenovirus, and sapovirus to norovirus.

Although the use of children only in this approach is not ideal, it allowed the use of

Australian data. An alternative approach using data from the UK IID2 study1 was also

considered, but was found to lead to unexpectedly high estimates for astrovirus and sapovirus

that were not consistent with estimates for other viral pathogens estimated used data from the

WQS. These differences perhaps arise from differences in the gastroenteritis case definitions

in the UK IID2 study compared to the NGSII study conducted in Australia. The incidence of

gastroenteritis in the NGSII study was almost fourfold higher than the IID2, which may

indicate large differences in case definitions.

TIME TREND MULTIPLIER

The WQS was undertaken before the addition of a rotavirus vaccine to the

nationalvaccination schedule in 2007. In calculating rotavirus incidence circa 2010, a time

trend multiplier was included to adjust for the reduction in rotavirus in 2010 compared with

pre-vaccination levels. In calculating this multiplier, data from a study of rotavirus

hospitalisations by age before and after the introduction of the national vaccination program

was used.36 By comparing age-specific hospitalisation rates in 2010 with that prior to

vaccination, a time-trend multiplier of 0.34 (95% CI: 0.32–0.36) was estimated in order to

adjust for the decline in rotavirus notifications following vaccination.


TECHNICAL APPENDIX 6: OTHER METHODS TO ESTIMATE INCIDENCE

METHODS TO CALCULATE COMMUNITY INCIDENCE

TOXOPLASMOSIS – SPECIAL CALCULATIONS

The calculations for toxoplasmosis differed from all other methods, as seroprevalence

studies were used to estimate yearly incident cases assuming a constant force of infection

with age.21 While there is an Australian study of toxoplasmosis,68 the sample size was too

small to rely on for this estimate. In adopting this USA study rather than European studies

(refer to Pappas et al.69 for a systematic review), comparability is ensured with the circa 2000

estimates, and a conservative approach is taken to estimating Australian incidence of

toxoplasmosis. This incidence estimate was then adjusted by a “proportion symptomatic”

multiplier of 15% (90% CrI:11–21) in line with the approach by Hall et al. circa 200010 and

that of Scallan et al.6

UNKNOWN PATHOGENS

The NGSII survey of gastroenteritis conducted in 2008–2009 was used to estimate the

total envelope of domestically acquired gastrointestinal illness, and so calculate the incidence

of unknown pathogens by subtracting the incidence of domestically acquired gastrointestinal

pathogens from that of the survey. Credible intervals for unknown pathogens were estimated

using @Risk, assuming all cases in the NGSII were domestically acquired. The foodborne

multiplier for all known pathogens of 25% (90% CrI: 15–39) was calculated as a weighted

average of the foodborne multiplier for each pathogen, weighted by the number of

domestically acquired cases of each pathogen. Although this value is remarkably similar to

that estimated by Scallan et al., it is worth noting that it is based entirely on Australian expert

elicitation data, together with incidence calculations using Australian data, and so is entirely

independent of that study. Examination of the two studies will identify differences in many

components of the calculations. The foodborne multiplier was applied to unknown pathogens

to estimate the total number of domestically acquired foodborne illness due to unknown

pathogens, again using @Risk for credible intervals.


TECHNICAL APPENDIX 7: HOSPITALISATIONS AND DEATHS METHODS

DATA SOURCES

Hospitalisation data from all Australian states and territories for 2006–2010 (where

available), and deaths data from the ABS were obtained, using ICD-10 codes for deaths and

ICD-10-AM codes for hospitalisations as in Table T7.1. Both astrovirus and sapovirus were

excluded from this analysis as lacking appropriate codes. Patients were included as a

hospitalisation if the appropriate code was included as the principal or an additional

diagnosis. Table T7.2 shows the percentage of all hospital diagnoses that were listed as the

principal diagnosis for each pathogen for 2010 (the year with most complete data). In the

circa 2000 estimates,3,10 only data on principal diagnoses were used, with a multiplier of two

(credible interval 1–3) for all pathogens to model both principal and additional diagnoses. It

is clear form Table T7.2 that diagnosis patterns vary considerably by pathogen, so that use of

both principal and additional diagnosis provides a more complete picture of hospitalisations.

Since only one year of hospitalisation data for Victoria and two years for New South

Wales were available, it was necessary to extrapolate from these data to the remaining years

to derive a distribution of the number of hospitalisations across all states and territories,

which was modelled as an empirical distribution. In most cases, it was assumed the same

number of hospitalisations each year, but some pathogens required further adjustment due to

evident outbreak trends. For example, an outbreak of hepatitis A associated with sundried

tomatoes coincided with the one year of hospitalisation data for Victoria. A ratio of

hospitalisations in South Australia to Victoria was used to estimate Victorian hospitalisations

for the missing years. As vaccination against rotavirus resulted in a decrease in incidence,

hospitalisations, and deaths, post universal vaccination data, from 2008–2010 only, were used

to estimate hospitalisations circa 2010.

ESTIMATION

To calculate estimates of hospitalisations and deaths, a statistical model was used that

incorporated uncertainty in case numbers and in multipliers using probability distributions.

That is, at each stage of the calculation, the estimate was represented by a probability

distribution, and the final estimates and credible intervals were computed from this

distribution. Input data was obtained from specific data sources (discussed above) or from


multipliers that are described below. A fuller description of these probability distributions is

provided in the methods section for incidence.

MULTIPLIERS

UNDERDIAGNOSIS MULTIPLIER

Recorded hospitalisations and deaths associated with each pathogen reflect only those

individuals that have been tested and confirmed for the pathogen. Following previous

studies, this was adjusted for using an underdiagnosis multiplier of two,5 including a

distribution for the multiplier with range 1–3 as in Hall et al. and Scallan et al.3,6 The

appropriateness of the multiplier for hospitalisations was confirmed as follows. Firstly, the

OzFoodNet outbreak register was used to calculate the proportion of all ill cases associated

with an outbreak that were hospitalised. This proportion was then compared to the ratio of

incidence to hospitalisations both with and without the underdiagnosis multiplier. Although

there was some variability by pathogen, overall, 3% of ill cases from 14 pathogens or

illnesses in the OzFoodNet outbreak register were hospitalised. In contrast, the ratio of all

incident cases to all hospitalised cases was around 0.01 when the underdiagnosis multiplier

was included (and 0.005 otherwise). Although outbreak cases may be more severe than all

incident cases (on average), and under–ascertainment of cases or under-recording of

hospitalizations may have biased the validation of the multiplier, these results suggest that an

underdiagnosis multiplier was appropriate. Further work would assist in better quantifying

this multiplier.


This multiplier adjusted for the proportion of cases that acquired infection in Australia,

and was adopted from the method for incidence. More details of the data and methods behind

this multiplier are provided in the methods section for calculating incidence.


This multiplier adjusted for the proportion of illness that is foodborne using expert

elicitation data, and was used for incidence, hospitalisations and deaths. Details are provided

in the methods section for incidence.


Table T7.1: Mortality and hospitalisation codes for each pathogen

Pathogen or illness Mortality ICD-10 Code and description

ICD-10-AM

Adenovirus A08.2: Adenoviral enteritis A08.2: Adenoviral enteritisBacillus cereus A05.4: Foodborne Bacillus cereus

intoxicationA05.4: Foodborne Bacillus cereus intoxication

Campylobacter spp. A04.5: Campylobacter enteritis A04.5: Campylobacter enteritisCiguatera T61.0: Ciguatera fish poisoning T61.0: Ciguatera fish poisoningClostridium perfringens

A05.2: Foodborne Clostridium perfringens intoxication

A05.2: Foodborne Clostridium perfringens [Clostridium welchii] intoxication

Cryptosporidium spp. A07.2: Cryptosporidiosis A07.2: CryptosporidiosisGuillain-Barré syndrome

G61.0: Guillain-Barré syndrome G61.0: Guillain-Barré syndrome

Giardia lamblia A07.1: Giardiasis [lambliasis] A07.1: Giardiasis [lambliasis]Hepatitis A* B15: Acute hepatitis A B15.9: Hepatitis A without hepatic

comaHaemolytic uraemic syndrome

D59.3: Haemolytic uraemic syndrome D59.3: Haemolytic uraemic syndrome


K58: Irritable bowel syndrome K58.0: Irritable bowel with diarrhoeaK58.9: Irritable bowel without diarrhoea

Listeria monocytogenes

A32: Listeriosis A32.0–A32.9: Listeriosis

Norovirus A08.1: Acute gastroenteropathy due to Norwalk agent

A08.1: Acute gastroenteropathy due to Norwalk agent

Other pathogenic E. coli

A04.0: Enteropathogenic Escherichia coli infectionA04.1: Enterotoxigenic Escherichia coli infectionA04.2: Enteroinvasive Escherichia coli infectionA04.4: Other intestinal Escherichia coli infection

A04.0: Enteropathogenic Escherichia coli infectionA04.1: Enterotoxigenic Escherichia coli infectionA04.2: Enteroinvasive Escherichia coli infectionA04.4: Other intestinal Escherichia coli infections

Reactive arthritis* M02.1: Postdysenteric arthropathyM02.8: Other reactive arthropathies

M02.1: Postdysenteric arthropathy, multiple sitesM02.3: Reiter’s disease, multiple sitesM02.8: Other reactive arthropathies, multiple sitesM03.2: Other postinfectious arthropathies in diseases classified


Pathogen or illness Mortality ICD-10 Code and description

ICD-10-AM

elsewhere, multiple sitesRotavirus A08.0: Rotaviral enteritis A08.0: Rotaviral enteritisSalmonella, non-typhoidal*

A02: Other Salmonella infections A02.0–A02.9: Salmonellosis

Salmonella Typhi A01: Typhoid and paratyphoid fevers A01: Typhoid feverScombrotoxicosis T61.1: Scombroid fish poisoning T61.6: Scombroid fish poisoningShigella A03: Shigellosis A03.0–A03.9: ShigellosisStaphylococcus aureus

A5.0: Foodborne staphylococcal intoxication

A05.0: Foodborne staphylococcal intoxication

Shiga toxin-producing E. coli

A04.3: Enterohaemorrhagic Escherichia coli infection

A04.3: Enterohaemorrhagic Escherichia coli infection

Toxoplasma gondii B58: Toxoplasmosis B58.0–B58.9: ToxoplasmosisVibrio parahaemolyticus

A05.3: Foodborne Vibrio parahaemolyticus intoxication

A05.3: Foodborne Vibrio parahaemolyticus intoxication

Yersinia enterocolitica

A04.6: Enteritis due to Yersinia enterocolitica

A04.6: Enteritis due to Yersinia enterocolitica

Other* A04.8: Other specified bacterial intestinal infectionA04.9: Bacterial intestinal infection unspecifiedA05.8: Other specified bacterial foodborne intoxicationsA05.9: Bacterial foodborne intoxication unspecifiedA07.8: Other specified protozoal intestinal diseasesA07.9: Protozoal intestinal disease, unspecifiedA08.3: Other viral enteritisA08.4: Viral intestinal infection, unspecifiedA09: Diarrhoea and gastroenteritis of presumed infectious originT61.2 Other fish and shellfish poisoningT61.8 Toxic effect of other seafoodT61.9 Toxic effect of unspecified seafoodT62: Toxic effect of other noxious substances eaten as foodT64: Toxic effect of aflatoxin and other mycotoxin food contaminants

A08.4: Viral intestinal infection, unspecifiedA09: Diarrhoea and gastroenteritis of presumed infectious originA09.0: Other gastroenteritis and colitis of infectious originA09.9: Other gastroenteritis and colitis of unspecified origin


*Pathogens are starred if there are differences between the mortality codes (ICD-10) and the hospitalisation codes (ICD-10-AM)

Table T7.2: The percentage of all hospital diagnoses that were listed as principal for each pathogen, based on 2010 data for all states and territories

Pathogen or illness Percentage of all diagnoses listed as principal

Adenovirus 82%

Bacillus cereus 75%

Campylobacter spp. 79%

Ciguatera 83%

Clostridium perfringens 100%

Cryptosporidium spp. 59%

Other pathogenic E. coli 59%

Giardia lamblia 34%

Guillain-Barré syndrome 71%

Irritable bowel syndrome 69%

Haemolytic uraemic syndrome 30%

Hepatitis A 77%

Listeria monocytogenes 48%

Norovirus 37%

Reactive arthritis 50%

Rotavirus 77%

Salmonella, non-typhoidal 77%

Salmonella Typhi 93%

Scombrotoxicosis 100%

Shigella 76%

Staphylococcus aureus 100%

Shiga toxin-producing E. coli 59%

Toxoplasma gondii 39%


Pathogen or illness Percentage of all diagnoses listed as principal

Vibrio parahaemolyticus 50%

Yersinia enterocolitica 64%

HOSPITALISATIONS AND DEATHS DUE TO UNKNOWN PATHOGENS

A large proportion of hospitalisations and deaths did not identify the source of infection

(refer to “other” codes in Table T7.1). These data were adjusted and reported as follows for

hospitalisations, with a similar approach used for deaths. Firstly, the total number of

hospitalisations due to unknown pathogens was calculated from the appropriate codes.

Hospitalisations that were attributed to known pathogens according to the underdiagnosis

multiplier described above were then subtracted from this number. That is, where total

numbers of known gastrointestinal pathogens were increased to adjust for underdiagnosis,

this increase was subtracted from the total unknown gastrointestinal pathogens. A

domestically acquired multiplier of one was assumed for unknown pathogens, but adjusted

for the foodborne multiplier using an average of known pathogens, weighted by the number

of hospitalisations for each pathogen. For hospitalisation data, this gave a foodborne

multiplier of 44% (90% CrI: 38–50), and for death data, a foodborne multiplier of 51% (90%

CrI: 36–71). Although Scallan et al.6 do not report their weighted foodborne multipliers for

hospitalisations and deaths, analysis of their tables suggest their values are 24% for

hospitalisations and 52% for deaths. As noted in Technical Appendix 2, the calculations here

are entirely independent; the Australian hospitalisation estimate is considerably higher

although the estimate for deaths shows good agreement.


TECHNICAL APPENDIX 8: METHODS FOR SEQUELAE INCIDENCE

To calculate community incidence of the four sequelae included in this assessment,

namely Guillain-Barré syndrome (GBS), haemolytic uraemic syndrome (HUS), irritable

bowel syndrome (IBS), and reactive arthritis (ReA), data were used from notifiable

surveillance to estimate incidence of the causal pathogens, and adjusted using a sequelae

multiplier modelled as a PERT distribution.

The approach to estimating sequelae-specific multiplier was to conduct a detailed

search of studies in the literature, and base the circa 2010 estimates on these studies, using

Australian studies and systematic reviews where possible. In literature searches, attention

was confined to peer-reviewed studies published in English between 1995 and 2012. Specific

keywords were used for each sequel to search PubMed. All types of studies were included;

titles and abstracts were reviewed, and full text articles were obtained for all relevant studies.

GUILLAIN-BARRÉ SYNDROME

Community incidence of Campylobacter spp. was estimated using NNDSS data,

adjusted for non-notification in New South Wales as described in Technical Appendix 5.

Incidence of GBS was estimated by further adjusting the distributions for Campylobacter

spp. by a sequelae multiplier estimated from values in published studies. Search terms for

GBS included: “Guillain-Barré syndrome”, “incidence”, “Campylobacter” and “Australia”,

with each selected paper including at least one search term. Four papers were identified that

estimated the proportion of Campylobacter spp. cases that led to GBS, with the one by Baker

et al.29 being excluded because it was only a study of hospitalised cases (Table T8.1). These

studies were used to estimate a range for the sequelae multiplier of 0.000192 to 0.000945,

with a midpoint value of 0.000304.

Table T8.1: References used to estimate the sequelae multiplier for Guillain-Barré syndrome – that is the proportion of Campylobacter spp. cases that result in GBS

Reference Study years Country # GBS cases /campylobacteriosis patients

Baker et al.29 1995–2008 New Zealand 35 / 8,448 (0.4%)Tam et al.70 1991–2001 UK 3 / 15,587 (0.019%)McCarthy & Giesecke71 1987–1995 Sweden 9 / 29,563 (0.03%)Allos72 1964–1996* Global/USA 1/1058 cases (0.0945%)* Years of reviewed studies


HAEMOLYTIC URAEMIC SYNDROME

Community incidence of STEC infection was estimated using South Australian

notifications, with multipliers to extend to the national population as described in Technical

Appendix 5. Incidence of HUS was estimated by applying a ‘sequelae multiplier’ estimated

from published studies to STEC. Search terms for HUS included: “Haemolytic uraemic

syndrome”, “HUS”, “incidence”, “STEC” and “E. coli”, with each selected paper including at

least one search term. Several published studies reported that 3–7% of sporadic STEC

infections resulted in HUS.73-76 Australian studies supported this range. Vally et al.77

examined South Australian surveillance data for a 13 year period where among the 460 cases

of STEC notified, there were 14 reported cases of HUS, resulting in an estimate of 3.0%

(95% CrI: 1.7–5.1) of STEC cases developing into HUS.

IRRITABLE BOWEL SYNDROME

Community incidence of Campylobacter spp., non-tyhoidal Salmonella spp., and

Shigella infections were estimated using NNDSS data as described in Technical Appendices

4 and 5. Incidence of IBS was estimated by further multiplying each incidence distribution

by a ‘sequelae multiplier’ estimated from published studies. Search terms for IBS included:

“Irritable bowel syndrome”, “IBS”, “Post-infection irritable bowel syndrome”, “PI-IBS”,

“incidence”, “Australia’, “Campylobacter”, “Salmonella” and “Shigella” with each selected

paper including at least one search term. The sequelae multiplier for IBS was based on a

meta-analysis conducted in the Netherlands by Haagsma et al.,78 who estimated that

approximately 8.8% (90% CI: 7.2–10.4%) of bacterial cases of gastroenteritis go on to

develop IBS 10 to 12 months after infection.

REACTIVE ARTHRITIS

Community incidence of Campylobacter spp., non-tyhoidal Salmonella spp., Shigella,

and Y. enterocolitica infections were estimated using NNDSS data (Technical Appendices 4

and 5). Incidence of ReA was estimated by multiplying incidence distributions for each of

these by sequelae multipliers for each of these four pathogens. Search terms for ReA included

“reactive arthritis”, “ReA”, “incidence”, “Australia”, “Campylobacter”, “Salmonella”,

“Shigella”, and “Yersinia”. Table T8.2 lists studies used to estimate the sequelae multipliers

following infection with each pathogen. From these studies sequelae multipliers of 0.07

(range: 0.028–0.16) for Campylobacter spp., 0.085 (range: 0–0.26) for non-tyhoidal


Salmonella spp., 0.097 (range: 0.012–0.098) for Shigella, and 0.12 (range: 0–0.231) for

Y. enterocolitica were adopted.

Table T8.2: References used to estimate the sequelae multiplier for reactive arthritis (ReA) for each of four pathogens: Campylobacter spp., non-tyhoidal Salmonella spp., Shigella, and Yersinia enterocoliticaReference Study years Country ReA cases /Gastroenteritis casesReA cases/Campylobacter casesSchonberg-Norio et al.79 2002 Finland 8/201 (4.0%)Doorduyn et al.80 2005 The Netherlands 20/434 (4.6%)Townes et al.81 2002–2004 USA 302/2384 (12.7%)Schiellerup et al.82 2002–2003 Denmark 131/1003 (13.1%)Pope et al.83 1966–2006 Europe 1 – 5%Hannu et al.84 2000 Finland 9/350 (2.6%)Rees et al.85 1998–1999 USA 9/324 (2.8%)Hannu et al.86 1997–1998 Finland 45/609 (7.4%)Locht & Krogfelt87 1997–1999 Denmark 27/173 (15.6%)ReA cases / non-tyhoidal Salmonella casesArnedo-Pena et al.88 2005 Spain 16/155 (10.3%)Doorduyn et al.80 2005 The Netherlands 8/181 (4.4%)Townes et al.81 2002–2004 USA 204/1356 (15.0%)Schiellerup et al.82 2002–2003 Denmark 104/619 (16.8%)Lee et al.89 1999 Australia 38/261 (14.6%)Rees et al.85 1998–1999 USA 2/100 (2.0%)Buxton et al.90 1999–2000 Canada 17/66 (25.7%)Hannu et al.91 1999 Finland 5/63 (7.9%)Locht et al.92 1999 Denmark 17/91 (18.7%)Rudwaleit et al.93 1998 Germany 0/286 (0%) (Children only)Dworkin et al.94 1994 USA 63/217 (29.0%)McColl et al.95 1997 Australia 13/312 (4.2%) & 6/112 (5.3%)Urfer et al.96 1993 Switzerland 1/156 (0.6%)Mattila et al.97 1994 Finland 22/191 (11.5%)Thomson et al.98 1984–1989 Canada 27/423 (6.4%)ReA cases /Shigella casesTownes et al.81 2002–2004 USA 29/298 (9.7%)Schiellerup et al.82 2002–2003 Denmark 10/102 (9.8%)Rees et al.85 1998–1999 USA 1/81 (1.2%)


Reference Study years Country ReA cases /Gastroenteritis casesReA cases /Yersiniosis casesHuovinen et al.99 2006 Finland 6/61 (9.8%)Townes et al.81 2002–2004 USA 5/35 (14.3%)Schiellerup et al.82 2002–2003 Denmark 21/91 (23.1%)Hannu et al.100 1998 Finland 4/33 (12.1%)


TECHNICAL APPENDIX 9: METHODS TO ESTIMATE HOSPITALISATIONS AND DEATHS DUE TO SEQUELAE

Hospitalisation data for 2006–2010 from all Australian states and territories for IBS and

ReA, and death data for all four sequelae from the ABS was obtained, using ICD-10 codes

for deaths and ICD-10-AM codes for hospitalisations as described in Table T7.1, and with

details of the percentage of hospitalisations that were the principal diagnosis in Table T7.2

(Technical Appendix 7). Additional and adjusted multipliers are described below. All

estimated incident foodborne GBS and HUS cases were considered hospitalised, so were not

modelled. Multipliers were still needed for GBS and HUS to estimate deaths.

MULTIPLIERS


This multiplier adjusts for proportion of illnesses that were acquired from food, and

was calculated to estimate hospitalisations and deaths due to sequelae, as it was not necessary

to estimate sequelae incidence since antecedent bacterial gastroenteritis cases were already

adjusted by a foodborne multiplier. Sequelae can arise from a source other than a bacterial

pathogen, from a bacterial pathogen that was not foodborne, or from a foodborne pathogen.

Only this latter category was considered a foodborne source. The proportion foodborne

multiplier is the simulated product of the bacterial multiplier and the weighted foodborne

multiplier and is shown for each sequel in Table T9.1.The approach calculating the

proportion foodborne multiplier for each sequel is described as follows:

Table T9.1: Foodborne multipliers for sequelae hospitalisations and deaths estimation

Sequelae Foodborne multiplierGuillain-Barré syndrome 0.25 (90% CrI: 0.1–0.43)Haemolytic uraemic syndrome 0.33 (90% CrI: 0.17–0.53)Irritable bowel syndrome 0.13 (90% CrI: 0.08–0.20)Reactive arthritis 0.48 (90% CrI: 0.36–0.62)


As not all GBS hospitalisations and deaths are from Campylobacter spp., a bacterial

multiplier was determined from published studies and multiplied with a foodborne proportion


to find the proportion of hospitalisations and deaths from foodborne Campylobacter spp. A

systematic review by Poropatich et al.,101 which found that 31% (range 4.8%–72%) of GBS

cases arise from Campylobacter spp. for the bacterial multiplier, was used as this was the

most recent systematic review. Another systematic review by McGrogan et al.102 reviewed 63

published papers between 1980 and 2009, and estimated that between 6% and 26% of cases

of GBS were due to a prior gastrointestinal infection. However the authors did not look at

Campylobacter spp. specifically. Applying the bacterial multiplier from Poropatich et al.101

together with the Campylobacter spp. foodborne multiplier described in Technical Appendix

5 gave a foodborne multiplier for GBS of 0.25 (90% CrI: 0.11–0.43).


A bacterial multiplier for HUS was first identified from published studies. Table T9.2

presents the percentage of cases of HUS that arise from STEC for four different papers,

including a global systematic review. From this, it was assumed that 61% (range 30%–85%)

of cases of HUS arise from STEC, modelled as a PERT distribution. Multiplying this

bacterial multiplier with the STEC foodborne multiplier described in Technical Appendix 5

led to a foodborne multiplier for HUS of 0.33 (90% CrI: 0.18–0.54).

Table T9.2: The percentage of haemolytic uraemic syndrome (HUS) that arise from STEC

Reference Study years Country Study type Percentage of HUS attributable to STEC

Walker et al.103 1980–2011 Global Systematic review 61% (range 30 – 85%)Askar et al.104 2011 Germany Surveillance 58%Elliot et al.105 1994–1998 Australia Surveillance 51%van de Kar et al.106 1989–1993 The Netherlands Case control 78%

IRRITABLE BOWEL SYNDROME:

The proportion of cases of IBS that arise from one of Campylobacter spp., non-tyhoidal

Salmonella spp., or Shigella spp. was estimated based on the proportion of IBS considered to

be post-infections in the literature. Table T9.3 presents four papers that estimate this

proportion. From this, it was assumed 17% of IBS to be triggered by a gastrointestinal

infection, with a range of 7–33%. This bacterial multiplier was modelled as a PERT

distribution. A foodborne multiplier for the combined three pathogens of 73% (90% CrI:

64%–82%) was calculated as a weighted average of the foodborne multipliers for each


pathogen, weighted by the total number of IBS cases for each pathogen. Multiplied together,

this gave a foodborne multiplier for irritable bowel syndrome of 13% (90% CrI: 8%–20%).

Table T9.3: The percentage of irritable bowel syndrome (IBS) that arises from Campylobacter spp., non-tyhoidal Salmonella spp., or Shigella

Reference Publication year

Country Study type Number of post infectious IBS

cases/IBS cases

% of IBS that is post-

infectiousChaudhary & Truelove107

1962 UK Epidemiological report

34/130 26.2%

Spiller & Garsed108

2009 Global Review - 6%–17%

Haagsma et al.78

2010 The Netherlands

Meta-analysis and estimation

- 17%

Schwille-Kiuntke et al.109

2013 Global Review - 7%–33%

REACTIVE ARTHRITIS

In a review of ReA, Hannu110 compiled population-based studies on the annual

incidence of ReA – both from enteric and urogenital infection. Using this compilation, this

report calculated the proportion of ReA due to enteric infection by dividing the enteric

incidence by the total incidence found in each study (Table T9.4). The midpoint and range of

the proportions from these studies was used for the bacterial multiplier. Therefore, it was

assumed a median of 66.7% of ReA is due to an enteric infection, with a range of 50%–

94.7%. The proportion foodborne was adjusted using a weighted average of the foodborne

multipliers for Campylobacter spp., non-tyhoidal Salmonella spp., Shigella spp., and

Y. enterocolitica, weighted by the total number of ReA cases for each pathogen. This gave a

foodborne multiplier of 72% (90% CrI: 60–82%). Multiplied by the above alternate PERT

distribution of median 66.7% (range 50%–94.7%), gave a foodborne multiplier for reactive

arthritis of 48% (90% CrI: 36%–61%).


Table T9.4: Proportion of reactive arthritis (ReA) attributable to enteric infection, adapted from the table of annual incidence of reactive arthritis based on population studies in Hannu110

Reference Country Publication year

Enteric incidence

per 100,000

population

Urogenital incidence

per 100,000

population

Total incidence

per 100,000

population

Proportion of ReA due to

enteric infection[enteric/total

incidence (%)]Isomaki et al.111

Finland 1978 14 13 27 14/27 (51.9%)

Kvien et al.112

Norway 1994 5 5 10 5/10 (50%)

Savolainen et al.113

Finland 2003 7 3 10 7/10 (70%)

Soderlin et al.114

Sweden 2003 18 1 19 18/19 (94.7%)

Townes115 USA 2010 0.6–3.1 NA NA NAHanova et al.116

Czech Republic

2010 6 3 ~9 6/9 (66.7%)

NA – Not applicable


This multiplier adjusts for the proportion of cases that were acquired infection in

Australia. The choice of domestically acquire multiplier for each sequel is shown in Table

T9.5, with the approach for selecting the multiplier described for each sequelae illness below.


The domestically acquired multiplier for Campylobacter spp. of 0.97 (range: 0.91–

0.99), as described elsewhere was adopted.


Given the relatively small numbers of notified cases of HUS, the domestically acquired

multiplier for STEC of 0.99 (range: 0.93–1.0) was adopted. A comparison of available travel

data for HUS was in agreement with this assumption.


IRRITABLE BOWEL SYNDROME

A combined domestically acquired multiplier for Campylobacter spp., non-tyhoidal

Salmonella spp., and Shigella of 91% (90% CrI: 88%–94%) was calculated as a weighted

average of the domestically acquired multipliers for each pathogen, weighted by the total

number of IBS cases for each pathogen.

REACTIVE ARTHRITIS

A combined domestically acquired multiplier for Campylobacter spp., non-tyhoidal

Salmonella spp., Shigella, and Y. enterocolitica of 91% (90% CrI: 86%–95%) was calculated

as a weighted average of the domestically acquired multipliers for each pathogen, weighted

by the total number of ReA cases for each pathogen.

Table T9.5: Domestically acquired multipliers for sequelae hospitalisations and deaths estimation

Sequelae Domestically acquired multiplierGuillain-Barré syndrome 0.97 (range: 0.91–0.99)Haemolytic uraemic syndrome 0.99 (range: 0.93–1.0)Irritable bowel syndrome 0.91 (90% CrI: 0.88–0.94)Reactive arthritis 0.91 (90% CrI: 0.86–0.95)


TECHNICAL APPENDIX 10: COMPARISON WITH ESTIMATES CIRCA 2000

COMPARISON OF INCIDENCE FOR FOODBORNE GASTROENTERITIS AND KEY PATHOGENS

As methods and data sources have changed since the circa 2000 estimation effort,

incidence estimates were recalculated for total foodborne gastroenteritis and for key

pathogens and two sequelae, using the same methods and data sources as were used for the

calculation of the circa 2010 estimates. The envelope of foodborne gastroenteritis for 2000

was estimated using the NGSI and the 2010 foodborne proportion of 25%. NNDSS data

from 1996 to 2000 and latest pathogen-specific foodborne proportions were used to

recalculate estimates for Campylobacter spp., non-typhoidal Salmonella spp., Salmonella

Typhi, hepatitis A, L. monocytogenes, GBS, IBS, and Victorian state surveillance data from

1996 to 2000 for the G. lamblia estimate. Incidence rates for 2000 were determined by using

the recalculated circa 2000 estimates and population numbers from the ABS from 1996 to

2000.24


TECHNICAL APPENDIX 11: PATHOGEN AND ILLNESS SHEETS

ADENOVIRUS

Primary data: Water Quality Study Alternate data: IID2

Model input, source & comments Distribution Data for model inputReported illness:Gastroenteritis multiplier – based on the NGSII

Pathogen fraction multiplier – based on age adjusted WQS of an estimated four positive isolates per 713 specimens (Hellard et al.2)

Alternate PERT

Alternate PERT

2.5%, median, 97.5% values:0.64, 0.74, 0.84

2.5%, median, 97.5% values:0.0015, 0.0056, 0.0143

Population adjustment:Australian resident population 2006–2010 June quarter - ABS(http://www.abs.gov.au/AUSSTATS/[email protected]/DetailsPage/3101.0Dec%202011?OpenDocument) (Accessed on 16/8/12)

Empirical By year (2006–2010):20697880, 21015936, 21384427, 21778845, 22065317

Domestically acquired multiplier:All isolations in the WQS were domestically acquired

N/A

Time trend multiplier:No time trend

N/A

Underreporting:WQS is community surveillance

N/A

Total illness:Population at risk*gastroenteritis multiplier*pathogen fraction multiplier*time trend multiplier

Outcome 5%, medial, 95% values:28800, 88400, 205000

Rate of total illness per million:circa 2010

Outcome 5%, median, 95% values:1300, 4150, 9675

Foodborne multiplier:Assumed to be the same as rotavirus

Alternate PERT

5%, median, 95% values:0.01, 0.02, 0.03

Total foodborne illness:Total illness*Foodborne multiplier


Rate of foodborne illness per million: circa 2010



http://www.abs.gov.au/AUSSTATS/[email protected]/DetailsPage/3101.0Dec%202011?OpenDocument

ASTROVIRUS

Primary data: Water Quality Study Alternate data:


Pathogen fraction multiplier – based on age adjusted WQS of an estimated four positive isolates of adenovirus per 713 specimens (Hellard et al.2)

Pathogen comparison multiplier – Kirkwood multiplier67 comparing adenovirus to astrovirus

Alternate PERT

Alternate PERT

Constant

2.5%, median, 97.5% values:0.64, 0.74, 0.84

2.5%, median, 97.5% values:0.0015, 0.0056, 0.0143

0.76




N/A


N/A


N/A





Foodborne multiplier:Assumed to be the same as rotavirus

Alternate PERT

5%, median, 95% values:0.01, 0.02, 0.03





Model input, source & comments Distribution Data for model inputRate of foodborne illness per million: circa 2010


BACILLUS CEREUS

Primary data: Outbreak Alternate data:

Model input, source & comments Distribution Data for model input

Reported illness:The number of B. cereus outbreak-associated illnesses reported to OzFoodNet 2006–2008.

Empirical By year (2006–2008):14, 35, 75

Population adjustment:

Australian resident population 2006–2010 June quarter - ABS(http://www.abs.gov.au/AUSSTATS/[email protected]/DetailsPage/3101.0Dec%202011?OpenDocument) (Accessed on 16/8/12)


Domestically acquired multiplier:Assumed to be 100% domestically acquired due to the short incubation period

PERT Minimum, modal, maximum values:1, 1 ,1

Underreporting:

Outbreak multiplier used to adjust from outbreak to surveillance (O-S)

Multiplier used to adjust for underreporting from surveillance to community (S-C)Non-typhoidal Salmonella multiplier adapted from Hall et al.22

PERT

Log normal

Minimum, modal, maximum values:5, 14, 20

Mean, standard deviation:7.44, 2.38

Total illness:Outbreak cases*Underreporting(O-S)(S-C)*Proportion travel-related




Foodborne multiplier:Based on 2005 expert elicitation

PERT Minimum, modal, maximum values:0.98, 1, 1








CAMPYLOBACTER SPP.

Primary data: NNDSS Alternate data: Water Quality Study

Model input, source & comments Distribution Data for model inputReported illness:NNDSS dataAvailable from: (http://www9.health.gov.au/cda/source/rpt_4.cfm)(Accessed on 12/11/13)

Empirical By year(1996–2000):12169, 11984, 12647, 12373, 13676(2006–2010):15416, 16980, 15539, 16075, 16967


Correction factor: Hall et al.22

NSW Population - June quarterBy year (1996–2000 & 2006–2010):6176461, 6246267, 6305799, 6375103, 6446558 & 6816087, 6885204, 6975891, 7069707, 71449281/(1-NSW Pop/AUS Pop)

Empirical

Constant

By year(1996–2000):18310714, 18517564, 18711271, 18925855, 19153380(2006–2010):20697880, 21015936, 21384427, 21778845, 220653171.5

Domestically acquired multiplier:NNDSS travel data

PERT Minimum, modal, maximum values:0.91, 0.97, 0.99

Underreporting:

Multiplier used to adjust for underreporting from surveillance to community (S-C)Campylobacter spp. multiplier adapted from Hall et al.22

Log normal Mean, standard deviation:10.45, 2.98

Total illness circa 2010:Reported cases (NNDSS)*correction factor*travel adjustment* underreporting(S-C)




Foodborne multiplier: Alternate 5%, median, 95% values:



http://www9.health.gov.au/cda/source/rpt_4.cfm

Model input, source & comments Distribution Data for model inputExpert elicitation study 2009 PERT 0.62, 0.77, 0.89Total foodborne illness:Total illness*Foodborne multiplier

Outcome 5%, median, 95% values:108500, 179000, 290000 (circa 2010)82500, 139000, 227000 (circa 2000)

Rate of foodborne illness per million: circa 2010 andcirca 2000


CIGUATERA

Primary data: Queensland notifications Alternate data: Outbreak


Reported illness:The number of ciguatera notifications reported in Queensland in OzFoodNet Queensland Annual Reports 2006–2010



Correction factor:Qld Population and NT populationAustralian population/Qld and NT population

Empirical

Empirical

By year (2006–2010):20697880, 21015936, 21384427, 21778845, 22065317

1.05

Domestically acquired multiplier:Assumed to be 100% domestically acquired


Underreporting:


Log Normal Mean, standard deviation:7.44, 2.38

Total illness:Reported cases (Qld notifications)*Population adjustment*Underreporting(O-S)(S-C)*Proportion travel-related







Foodborne multiplier:Assumed to be 100% foodborne

PERT Minimum, modal, maximum values:1, 1, 1





CLOSTRIDIUM PERFRINGENS

Primary Data: Outbreak Alternate Data: Water Quality Study

Model input, source & comments Distribution Data for model inputReported illness:The number of C. perfringens outbreak-associated illnesses reported to OzFoodNet 2006–2008.







Underreporting:


Multiplier used to adjust for underreporting from surveillance to community (S-C)Non-typhoidal Salmonella multiplier adapted from Hall et al. 22

PERT

Log normal





Model input, source & comments Distribution Data for model inputTotal illness:Outbreak cases*Underreporting(O-S)(S-C)*Proportion travel-related




Foodborne Multiplier:Expert elicitation study 2009

PERT Minimum, modal, maximum values:0.86, 0.98, 1





CRYPTOSPORIDIUM SPP.



Reported illness:NNDSS dataAvailable from: (http://www9.health.gov.au/cda/source/rpt_4.cfm)(Accessed on 28/06/12)






Underreporting:







Total illness:Reported cases (NNDSS)*travel adjustment*underreporting(S-C)




Foodborne multiplier:Based on 2005 Delphi survey

Alternate PERT

5%, median, 95% values:0.01, 0.1, 0.27





GIARDIA LAMBLIA

Primary Data: Victoria notifications Alternate Data: Water Quality Study

Model input, source & comments Distribution Data for model inputReported illness:Victorian state notifications from: O’Grady and Tallis;117 Brown et al.61-64

Giardiasis became a non-notifiable disease in Victoria in 2010

Empirical By year(1996–2000):1085, 1060, 999, 921, 866(2006–2009):1192, 1382, 1434, 1433


Correction factor:Victorian population - June quarterBy year (1996–2000 & 2006–2009):4534984, 4569297, 4606970, 4652462, 4704065& 5126540, 5204607, 5293088, 5395137Australian population/Victorian population

Empirical

Constant

By year(1996–2000):18310714, 18517564, 18711271, 18925855, 19153380(2006–2009):20697880, 21015936, 21384427, 21778845

4.03

Domestically acquired multiplier:Victorian notification data




Model input, source & comments Distribution Data for model inputUnderreporting:


Log Normal Mean, standard deviation:7.44, 2.38

Total illness:Reported cases (Victorian notifications)*correction factor*travel adjustment*underreporting(S-C)




Foodborne multiplier:Based on Delphi 2005






GUILLAIN-BARRÉ SYNDROME (GBS)

Primary data: NNDSS and literature Alternate Data: Hospitalisations and literature



Reported illness:As GBS is not notified, sequelae multiplier, or the proportion of foodborne Campylobacter spp. cases that develop GBS, was used to determine the number of foodborne GBS cases. This proportion was a midpoint between estimates from the literature reported in McCarthy and Giesecke,71 Tam et al.,70 and Allos.72 Refer to Technical Appendix 8 for further explanation.

Antecedent bacterial gastroenteritis cases: estimated number of foodborne Campylobacter spp. cases.

Sequelae multiplier

Outcome

PERT

5%, median, 95% values:108500, 179000, 290000 (circa 2010)82500, 13900, 227000 (circa 2000)

Minimum, modal, maximum values:0.000192, 0.00034, 0.000945

Total foodborne illness:Foodborne Campylobacter spp. cases*Proportion of cases that develop GBS


Rate of foodborne illness from Campylobacter spp. per million:circa 2010 and circa 2000

Outcome 5%, median, 95% values:2, 3.1, 6 (circa 2010)1, 2.8, 6 (circa 2000)

HAEMOLYTIC URAEMIC SYNDROME (HUS)

Primary data: South Australian STEC surveillance and literature Alternate data: Hospitalisations


Model input, source & comments Distribution Data for model inputReported illness:As HUS is a sequel to STEC, the proportion of foodborne STEC cases that develop HUS was used to determine the number of foodborne HUS cases. This proportion (sequelae multiplier is from Vally et al.77 Refer to Technical Appendix 8 for further explanation.

Antecedent bacterial gastroenteritis cases: estimated number of foodborne STEC cases

Sequelae multiplier

Outcome

PERT



Total foodborne illness:Foodborne STEC cases*Proportion of cases that develop HUS


Rate of foodborne illness from STEC per million:circa 2010

Outcome 5%, median, 95% values:1, 3.3, 9


HEPATITIS A

Primary data: NNDSS Alternate data:


Reported illness:NNDSS dataAvailable from: (http://www9.health.gov.au/cda/source/rpt_4.cfm) (Accessed on 15/05/12)

Empirical By year(1996–2000):2058, 3032, 2466, 1551, 809(2006–2010):281, 166, 277, 564, 267


Empirical By year(1996–2000):18310714, 18517564, 18711271, 18925855, 19153380

(2006–2010):20697880, 21015936, 21384427, 21778845, 22065317



Underreporting:

Multiplier used to adjust for underreporting from surveillance to community (S-C)

Alternate PERT

2.5%, median, 97.5% values:1, 2, 3





Foodborne multiplier:Expert elicitation study 2009

Alternate PERT

5%, median, 95% values:0.05, 0.12, 0.24







Rate of foodborne illness per million: 2010 andcirca 2000


IRRITABLE BOWEL SYNDROME (IBS)

Primary data: NNDSS and literature Alternate data: Literature


Reported illness:As IBS is not notified, the proportion of foodborne Campylobacter spp., Salmonella spp., and Shigella cases that develop IBS was calculated to determine the number of foodborne IBS cases. Refer to Technical Appendix 8 for further explanation

Antecedent bacterial gastroenteritis cases: Estimated number of foodborne Campylobacter spp. cases

Estimated number of foodborne non-typhoidal Salmonella spp. cases

Estimated number of foodborne Shigella spp. cases

Sequelae multiplier

Outcome

Outcome

Outcome

Alternate PERT

5%, median, 95% values108500, 179000, 290000 (circa 2010)82500, 139000, 227000 (circa 2000)



2.5%, median, 97.5% values:0.072, 0.088, 0.104

Total foodborne illness:Foodborne bacterial gastroenteritis cases from 3 pathogens*Proportion of these cases that develop IBS

Outcome 2.5%, median, 97.5% values:12500, 19500, 30700 (circa 2010)9500, 14800, 23500 (circa 2000)



Rate of foodborne illness per million: circa 2010 Outcome 2.5%, median, 97.5% values:570, 915, 1440 (circa 2010)550, 850, 1350 (circa 2000)

LISTERIA MONOCYTOGENES

Primary data: NNDSS Alternate data: Outbreak


Reported illness:NNDSS dataAvailable from: (http://www9.health.gov.au/cda/source/rpt_4.cfm)(Accessed on 12/11/13)


(2006–2010):61, 50, 68, 92, 71



(2006–2010):20697880, 21015936, 21384427, 21778845, 22065317

Domestically acquired multiplier:Assumed to be 100% as the majority of travellers are not at high risk


Underreporting:

Multiplier used to adjust for underreporting from surveillance to community (S-C)

Alternate PERT

2.5%, medial, 97.5% values:1, 2, 3















NOROVIRUS

Primary data: Water Quality Study Alternate data: Outbreak


Pathogen fraction multiplier – based on age adjusted WQS of an estimated 69 positive isolates per 703 specimens (Sinclair et al.20)

Alternate PERT

Alternate PERT

2.5%, median, 97.5% values:0.64, 0.74, 0.84

2.5%, median, 97.5% values:0.0772, 0.0982, 0.1226




N/A


N/A


N/A



Model input, source & comments Distribution Data for model inputTotal illness:Population at risk*gastroenteritis multiplier*pathogen fraction multiplier*time trend multiplier





Alternate PERT

5%, median, 95% values:0.05, 0.18, 0.35





OTHER PATHOGENIC ESCHERICHIA COLI



Pathogen fraction multiplier – based on age adjusted WQS of an estimated 50 positive isolates per 713 specimens (Hellard et al.2)

Alternate PERT

Alternate PERT

2.5%, median, 97.5% values:0.64, 0.74, 0.84

2.5%, median, 97.5% values:0.0525, 0.074, 0.0914





N/A


N/A


N/A



Model input, source & comments Distribution Data for model inputTotal illness:Population at risk*gastroenteritis multiplier*pathogen fraction multiplier*time trend multiplier





Alternate PERT

5%, median, 95% values:0.08, 0.23, 0.55





REACTIVE ARTHRITIS (REA)

Primary data: NNDSS and literature Alternate data:


Reported illness:As ReA is not notified, the proportion of foodborne Campylobacter spp., Salmonella spp., Shigella, and Y. enterocolitica cases that develop ReA was calculated to determine the number of foodborne ReA cases. The proportion for each of the four pathogens was calculated from the literature. Refer to Technical Appendix 8 for further explanation

Antecedent bacterial gastroenteritis cases:Estimated number of foodborne Campylobacter spp. cases

Estimated number of foodborne non-typhoidal Salmonella spp. cases

Estimated number of foodborne Shigella spp. cases

Estimated number of foodborne Yersinia enterocolitica cases

Outcome

Outcome

Outcome

Outcome







Sequelae multiplier for Campylobacter spp.

Sequelae multiplier for non-typhoidal Salmonella spp.

Sequelae multiplier for Shigella spp.

Sequelae multiplier for Yersinia enterocolitica

Alternate PERT

Alternate PERT

PERT

Alternate PERT

Minimum, medial, maximum values:0.028, 0.07, 0.16

Minimum, medial, maximum values:0, 0.085, 0.26


Minimum, medial, maximum values:0, 0.12, 0.231

Total foodborne illness:Foodborne bacterial gastroenteritis cases from pathogens*Proportion of cases that develop ReA


Rate of foodborne illness per million:circa 2010


ROTAVIRUS



Pathogen fraction multiplier – based on age adjusted WQS of an estimated 6 positive isolates per 713 specimens (Hellard et al.2)

Alternate PERT

Alternate PERT

2.5%, median, 97.5% values:0.64, 0.74, 0.84

2.5%, median, 97.5% values:0.0031, 0.0084, 0.0182




N/A

Time trend multiplier:Based on Dey et al.36

Alternate PERT

2.5%, median, 97.5% values:0.318, 0.338, 0.359



Model input, source & comments Distribution Data for model inputUnderreporting:WQS is community surveillance

N/A






Alternate PERT

5%, median, 95% values:0.01, 0.02, 0.03





SALMONELLA SPP. , NON-TYPHOIDAL


Model input, source & comments Distribution Data for model inputReported illness:NNDSS dataAvailable from: (http://www9.health.gov.au/cda/source/rpt_4.cfm) (Accessed on 11/12/13)

Empirical By year(1996–2000)5744, 6955, 7513, 7008, 6187

(2006–2010)8241, 9502, 8316, 9524, 11928



(2006–2010):20697880, 21015936, 21384427, 21778845, 22065317






Model input, source & comments Distribution Data for model inputUnderreporting:



Total illness circa 2010:Reported cases (NNDSS)*travel adjustment*underreporting(S-C)





Alternate PERT

5%, median, 95% values:0.53, 0.72, 0.86



Rate of foodborne illness per million: circa 2010 and circa 2000


SALMONELLA TYPHI


Model input, source & comments Distribution Data for model inputReported illness:NNDSS dataAvailable from: (http://www9.health.gov.au/cda/source/rpt_4.cfm)(Accessed on 28/06/12)


(2006–2010):77, 90, 105, 115, 95



(2006–2010):20697880, 21015936, 21384427, 21778845, 22065317




Model input, source & comments Distribution Data for model inputDomestically acquired multiplier:NNDSS travel data


Underreporting:

Multiplier used to adjust for underreporting from surveillance to community (S-C)Multiplier of two for serious illnesses

Alternate PERT

2.5%, median, 97.5% values:1, 2, 3










Outcome 5%, median, 95% values:0, 0.6, 1 (circa 2010)0, 0.5, 1 (circa 2000)

SAPOVIRUS



Pathogen fraction multiplier – based on age adjusted WQS findings for norovirus of an estimated 69 isolates per 703 specimens (Sinclair et al.20

Pathogen comparison multiplier – Kirkwood multiplier67 comparing norovirus to sapovirus

Alternate PERT

Alternate PERT

Constant

2.5%, median, 97.5% values:0.64, 0.74, 0.84

2.5%, median, 97.5% values:0.0772, 0.0982, 0.1226

0.5


Model input, source & comments Distribution Data for model inputPopulation adjustment:Australian resident population 2006–2010 June quarter - ABS(http://www.abs.gov.au/AUSSTATS/[email protected]/DetailsPage/3101.0Dec%202011?OpenDocument) (Accessed on 16/8/12)



N/A


N/A


N/A

Total illness:Population at risk*gastroenteritis multiplier*pathogen fraction multiplier*pathogen comparison multiplier*time trend multiplier




Foodborne multiplier:Assumed to be the same as norovirus






SCOMBROTOXICOSIS


Model input, source & comments Distribution Data for model inputReported illness:The number of scombrotoxicosis outbreak-associated illnesses reported to OzFoodNet 2006–2008.

Empirical By year (2006–2008)12, 17, 0



Model input, source & comments Distribution Data for model inputPopulation adjustment:





Underreporting:



PERT

Log Normal



Total illness:Outbreak cases*Underreporting (O-S)(S-C)*Proportion travel-related




Foodborne multiplier:Assumed to be 100% foodborne






SHIGELLA



Reported illness:NNDSS dataAvailable from: (http://www9.health.gov.au/cda/source/rp


(2006–2010):





t_4.cfm )(Accessed on 12/11/13) 545, 597, 828, 618, 550



(2006–2010):20697880, 21015936, 21384427, 21778845, 22065317

Domestically acquired multiplier: NNDSS travel data


Underreporting:








Alternate PERT

5%, median, 95% values:0.05, 0.12, 0.23





STAPHYLOCOCCUS AUREUS




Model input, source & comments Distribution Data for model inputReported illness:The number of S. aureus outbreak-associated illnesses reported to OzFoodNet 2006–2008.

Empirical By year (2006–2008)3, 14, 50


Australian resident population 2006–2008 June quarter - ABS(http://www.abs.gov.au/AUSSTATS/[email protected]/DetailsPage/3101.0Dec%202011?OpenDocument)(Accessed on 16/8/12)


Domestically acquired multiplier:Assumed to be 100% domestically acquired due to short incubation period


Underreporting:



PERT

Log normal



Total illness:Outbreak cases*Underreporting(O-S)(S-C)*Proportion travel-related





PERT Minimum, modal, maximum values:0.95, 1, 1





SHIGA TOXIN-PRODUCING ESCHERICHIA COLI

Primary data: South Australia surveillance Alternate data: NNDSS



Model input, source & comments Distribution Data for model inputReported illness:South Australian state STEC surveillance from the study by Vally et al.77

Empirical By year (2006–2010)35, 40, 39, 62, 32


Correction factor:SA population - June quarterBy year (2006–2010):1567888, 1582559, 1597343, 1614375, 1629434Australian population/SA population


13.4



Underreporting:

Multiplier used to adjust for underreporting from surveillance to community (S-C)STEC multiplier adapted from Hall et al.22


Total illness:Reported cases (SA surveillance)*correction factor*travel adjustment*underreporting(S-C)





Alternate PERT

5%, median, 95% values:0.32, 0.56, 0.83





TOXOPLASMA GONDII

Primary data: Seroprevalence estimate Alternate data:




Reported illness:USA seroprevalence data21 extrapolated to the Australian population for 2010 by age group

Empirical 0–4: 57095–9: 574910–19: 1074420–29: 1172830–39: 1080940–49: 1037750–59: 890360–69: 652170–79: 371380+: 2342

Total: 76095


Australian resident population 2010 June quarter by age group - ABS(http://www.abs.gov.au/AUSSTATS/[email protected]/DetailsPage/3101.0Dec%202011?OpenDocument) (Accessed on 16/8/12)

Empirical 0–4: 14416795–9: 135221110–19: 285205020–29: 324034730–39: 310822440–49: 310587750–59: 277351160–69: 211415870–79: 125311480+: 824146

Domestically acquired multiplier:Assumed to be 100% domestically acquired


Proportion symptomatic:Scallan et al.6 and Abelson et al.4


Total illness:Estimated yearly cases*travel adjustment*proportion symptomatic










VIBRIO PARAHAEMOLYTICUS

Primary data: Western Australia notifications Alternate data:



Model input, source & comments Distribution Data for model inputReported illness:WA notifications(http://www.public.health.wa.gov.au/cproot/4195/2/12172_DiseaseWAtch.pdf)



Correction factor:WA population - June quarterBy year (2006–2009):2059381, 2113841, 2178577, 2246659, 2296129Australian population/WA Population

Empirical

Constant

By year (2006–2009):20697880, 21015936, 21384427, 21778845

9.61

Domestically acquired multiplier:OzFoodNet WA Annual Reports 2006–2010

PERT Minimum, modal, maximum values:0, 0.18, 0.33

Underreporting:



Total illness:Reported cases (WA surveillance)*correction factor*travel adjustment*underreporting(S-C)










YERSINIA ENTEROCOLITICA

Primary data: State and territory notifications Alternate data:



http://www.public.health.wa.gov.au/cproot/4195/2/12172_DiseaseWAtch.pdf


Reported illness:State notifications from Queensland, SA, WA, and NT extrapolated from state data to the Australian population to determine the expected number of notifications if all states were reporting

Empirical By year (2006–2010)214, 249, 326, 242

Population adjustment:Australian resident population 2006–2010 June quarter

Queensland resident population (June quarter)

SA resident population (June quarter)

WA resident population (June quarter)

NT Population (June quarter) - ABS(http://www.abs.gov.au/AUSSTATS/[email protected]/DetailsPage/3101.0Dec%202011?OpenDocument) (16/8/12)Notifications / estimated Australian population


By year (2006–2010):4090908, 4177089, 4270091, 4365426, 4424158By year (2006–2010):1567888, 1582559, 1597343, 1614375, 1629434By year (2006–2010):2059381, 2113841, 2178577, 2246659, 2296129By year (2006–2010)210627, 215021, 220935, 226841, 230315

Domestically acquired multiplier:OzFoodNet WA Annual Reports 2006-2010


Underreporting:Multiplier used to adjust for underreporting from surveillance to community (S-C)Non-typhoidal Salmonella multiplier adapted from Hall et al.22


Total illness:Reported cases(Extrapolated state notifications)*travel adjustment*underreporting(S-C)








Rate of foodborne illness per million:circa 2010




REFERENCES

1. Tam C, Rodrigues L, Viviani L, et al. Longitudinal study of infectious intestinal disease in the UK (IID2 study): incidence in the community and presenting to general practice. Gut 2012;61:69-77.

2. Hellard ME, Sinclair MI, Forbes A, Fairley CK. A randomized, blinded, controlled trial investigating the gastrointestinal health effects of drinking water quality. Environmental Health Perspectives 2001;109:773-8.

3. Hall G, Kirk M, Becker N, et al. Estimating foodborne gastroenteritis, Australia. Emerg Infect Dis 2005;11:1257-64.

4. Abelson P, Potter Forbes M, Hall G. The annual cost of foodborne illness in Australia. Canberra: Australian Government Department of Health and Ageing; 2006. Report No.: 0642829063.

5. Mead P, Slutsker L, Dietz V, et al. Food-related illness and death in the United States. Emerg Infect Dis 1999;5:607-25.

6. Scallan E, Hoekstra R, Angulo F, et al. Foodborne illness acquired in the United States - major pathogens. Emerg Infect Dis 2011;17:7-15.

7. Scallan E, Griffin PM, Angulo FJ, Tauxe RV, Hoekstra RM. Foodborne illness acquired in the United States—unspecified agents. Emerg Infect Dis 2011;17:16-22.

8. Tauxe RV, Doyle MP, Kuchenmüller T, Schlundt J, Stein CE. Evolving public health approaches to the global challenge of foodborne infections. Int J Food Microbiol 2010;139, Supplement:S16-S28.

9. Morris JG, Jr. How safe is our food? Emerg Infect Dis 2011;17:126-8.

10. Hall G, Kirk M. Foodborne illnesses in Australia : annual incidence circa 2000. Canberra: Australian Government Department of Health and Ageing; 2005. Report No.: 0642825769.

11. Vaillant V, De Valk H, Baron E, et al. Foodborne infections in France. Foodborne Pathogens and Disease 2005;2:221-32.

12. Adak GK, Long SM, O'Brien SJ. Trends in indigenous foodborne disease and deaths, England and Wales: 1992 to 2000. Gut 2002;51:832-41.

13. Gkogka E, Reij MW, Havelaar AH, Zwietering MH, Gorris LGM. Risk-based Estimate of Effect of Foodborne Diseases on Public Health, Greece. Emerg Infect Dis 2011;17:1581-90.

14. Havelaar A, Haagsma J, Mangen M-J, et al. The disease burden of foodborne pathogens in the Netherlands, 2009. Int J Food Microbiol 2012;156:231-8.

15. Lake RJ, Cressey PJ, Campbell DM, Oakley E. Risk ranking for foodborne microbial hazards in New Zealand: burden of disease estimates. Risk Anal 2010;30:743-52.

16. Gargouri N, Walke H, Belbeisi A, et al. Estimated burden of human Salmonella, Shigella, and Brucella infections in Jordan, 2003-2004. Foodborne Pathog Dis 2009;6:481-6.


17. Thomas MK, Murray R, Flockhart L, et al. Estimates of the Burden of Foodborne Illness in Canada for 30 Specified Pathogens and Unspecified Agents, Circa 2006. Foodborne Pathog Dis 2013;10:639-48.

18. Hall G. Results from the National Gastroenteritis Survey 2001-2002. Canberra: National Centre for Epidemiology and Population Health; 2004.

19. Buzby JC, Roberts T. The economics of enteric infections: human foodborne disease costs. Gastroenterology 2009;136:1851-62.

20. Sinclair MI, Hellard ME, Wolfe R, Mitakakis TZ, Leder K, Fairley CK. Pathogens causing community gastroenteritis in Australia. J Gastroenterol Hepatol 2005;20:1685-90.

21. Jones JL, Kruszon-Moran D, Sanders-Lewis K, Wilson M. Toxoplasma gondii infection in the United States, 1999–2004, decline from the prior decade. Am J Trop Med Hyg 2007;77:405-10.

22. Hall G, Yohannes K, Raupach J, Becker N, Kirk M. Estimating community incidence of Salmonella, Campylobacter, and Shiga toxin-producing Escherichia coli infections, Australia. Emerg Infect Dis 2008;14:1601-9.

23. Vally H, Glass K, Ford L, et al. The proportion of illness transmitted by food for nine enteric pathogens in Australia: an expert elicitation. Foodborne Pathog Dis 2014:doi:10.1089/fpd.2014.1746.

24. Australian Bureau of Statistics. Table 4. Estimated Resident Population, States and Territories (Number). Avaliable from: ABS (http://www.abs.gov.au/AUSSTATS/[email protected]/DetailsPage/3101.0Dec%202011?OpenDocument). In. Canberra: ABS; 2012.

25. Thabane M, Kottachchi DT, Marshall JK. Systematic review and meta-analysis: the incidence and prognosis of post-infectious irritable bowel syndrome. Aliment Pharmacol Ther 2007;26:535-44.

26. Havelaar A, De Wit M, Van Koningsveld R, Van Kempen E. Health burden in the Netherlands due to infection with thermophilic Campylobacter spp. Epidemiol Infect 2000;125:505-22.

27. Vally H, Hall G, Scallan E, Kirk MD, Angulo FJ. Higher rate of culture-confirmed Campylobacter infections in Australia than in the USA: is this due to differences in healthcare-seeking behaviour or stool culture frequency? Epidemiol Infect 2009;137:1751-8.

28. Baker MG, Sears A, Wilson N, et al. Keep the focus on contaminated poultry to further curtail New Zealand's campylobacteriosis epidemic. N Z Med J 2011;124:135-9.

29. Baker MG, Kvalsvig A, Zhang J, Lake R, Sears A, Wilson N. Declining Guillain-Barré syndrome after campylobacteriosis control, New Zealand, 1988-2010. Emerg Infect Dis 2012;18:226-33.

30. Hald T, Vose D, Wegener HC, Koupeev T. A Bayesian approach to quantify the contribution of animal-food sources to human salmonellosis. Risk Anal 2004;24:255-69.



31. Davidson VJ, Ravel A, Nguyen TN, Fazil A, Ruzante JM. Food-specific attribution of selected gastrointestinal illnesses: Estimates from a Canadian expert elicitation survey. Foodborne Pathog Dis 2011;8:983-95.

32. Mullner P, Jones G, Noble A, Spencer SEF, Hathaway S, French NP. Source attribution of food-borne zoonoses in New Zealand: a modified Hald model. Risk Anal 2009;29:970-84.

33. Pires SM, Vigre H, Makela P, Hald T. Using outbreak data for source attribution of human salmonellosis and campylobacteriosis in Europe. Foodborne Pathog Dis 2010;7:1351-61.

34. Sumner J, Ross T. A semi-quantitative seafood safety risk assessment. Int J Food Microbiol 2002;77:55-9.

35. Hall G, Vally H. Expert elicitation: what proportion of illness due to nine pathogens is foodborne? Canberra: Funded by the Australian Government Department of Health and Ageing and NSW Food Authority; 2009.

36. Dey A, Wang H, Menzies R, Macartney K. Changes in hospitalisations for acute gastroenteritis in Australia after the national rotavirus vaccination program. Med J Aust 2012;197:453-7.

37. OzFoodNet Working Group. Monitoring the Incidence and Causes of Disease Potentially Transmitted by Food in Australia: Annual Report of the OzFoodNet Network, 2008. Comm Dis Intell 2009;33:389-413.

38. Donnan EJ, Fielding JE, Gregory JE, et al. A multistate outbreak of hepatitis A associated with semidried tomatoes in Australia, 2009. Clin Infect Dis 2012;54:775-81.

39. OzFoodNet Working Group. Monitoring the incidence and causes of diseases potentially transmitted by food in Australia: Annual Report of the OzFoodNet Network, 2010. Comm Dis Intell 2012;36:E213-E41.

40. Robins-Browne RM, Bordun AM, Tauschek M, et al. Escherichia coli and community-acquired gastroenteritis, Melbourne, Australia. Emerg Infect Dis 2004;10:1797-805.

41. Painter JA, Hoekstra RM, Ayers T, et al. Attribution of Foodborne Illnesses, Hospitalizations, and Deaths to Food Commodities using Outbreak Data, United States, 1998-2008. Emerg Infect Dis 2013;19:407-15.

42. Herikstad H, Yang S, Van Gilder TJ, et al. A population-based estimate of the burden of diarrhoeal illness in the United States: FoodNet, 1996-7. Epidemiol Infect 2002;129:9-17.

43. Majowicz S, Dore K, Flint J, et al. Magnitude and distribution of acute, self-reported gastrointestinal illness in a Canadian community. Epidemiol Infect 2004;132:607-17.

44. Wheeler JG, Sethi D, Cowden JM, et al. Study of infectious intestinal disease in England: rates in the community, presenting to general practice, and reported to national surveillance. BMJ 1999;318:1046-50.


45. de Wit MAS, Koopmans MPG, Kortbeek LM, et al. Sensor, a Population-based Cohort Study on Gastroenteritis in the Netherlands: Incidence and Etiology. Am J Epidemiol 2001;154:666-74.

46. Sudman S, Bradburn N. Effects of time and memory factors on response in surveys. J Am Stat Assoc 1973;68:805-15.

47. Roy S, Scallan E, Beach M. The rate of acute gastrointestinal illness in developed countries. J Water Health 2006;04.Suppl 2:31-69.

48. Rodrigo S, Sinclair M, Forbes A, Cunliffe D, Leder K. Drinking rainwater: A double-blinded, randomized controlled study of water treatment filters and gastroenteritis incidence. Am J Public Health 2011;101:842-7.

49. Majowicz S, Hall G, Scallan E, et al. A common, symptom-based case definition for gastroenteritis. Epidemiol Infect 2008;136:886-94.

50. Hall G, McDonald I, Majowicz SE, et al. Respiratory symptoms and the case definition of gastroenteritis: an international analysis of the potential impact on burden estimates. Epidemiol Infect 2010;138:117-24.

51. Dingle JH, Badger GF, Jordan Jr W. Illness in the Home. A Study of 25,000 Illnesses in a Group of Cleveland Families. Cleveland, Ohio: Press of Western Reserve University; 1964.

52. Monto AS, Koopman JS. The Tecumseh study. XI. Occurence of acute enteric illness in the community. Am J Epidemiol 1980;112:323-33.

53. Kirk M, McKay I, Hall G, et al. Foodborne disease in Australia: the OzFoodNet experience. Clin Infect Dis 2008;47:392-400.

54. Voetsch AC, Van Gilder TJ, Angulo FJ, et al. FoodNet estimate of the burden of illness caused by nontyphoidal Salmonella infections in the United States. Clin Infect Dis 2004;38:S127-S34.

55. Ruzante JM, Majowicz SE, Fazil A, Davidson VJ. Hospitalization and deaths for select enteric illnesses and associated sequelae in Canada, 2001-2004. Epidemiol Infect 2011;139:937-45.

56. Havelaar AH, Brul S, de Jong A, de Jonge R, Zwietering MH, Ter Kuile BH. Future challenges to microbial food safety. Int J Food Microbiol 2010;139 Suppl 1:S79-94.

57. Pires SM, Evers EG, Van Pelt W, et al. Attributing the human disease burden of foodborne infections to specific sources. Foodborne Pathog Dis 2009;6:417-24.

58. Ravel A, Davidson VJ, Ruzante JM, Fazil A. Foodborne proportion of gastrointestinal illness: estimates from a Canadian expert elicitation survey. Foodborne Pathog Dis 2010;7:1463-72.

59. Havelaar AH, Galindo AV, Kurowicka D, Cooke RM. Attribution of foodborne pathogens using structured expert elicitation. Foodborne Pathog Dis 2008;5:649-59.

60. Kirk M, McKercher C, Hall G. Gastroenteritis in Australia: Report of the National Gastroenteritis Survey II 2008: OzFoodNet; 2011.


61. Brown L, Fielding J, Gregory J, et al. Surveillance of notifiable infectious diseases in Victoria, 2006. In: Public Health Branch, ed. Melbourne: Communicable Disease Control Unit, Rural & Regional Health and Aged Care Services, Department of Human Services, Victoria; 2008.

62. Brown L, El-Hayek C, Fielding J, et al. Surveillance of notifiable infectious diseases in Victoria, 2007. In: Health Protection Branch ed. Melbourne: The Communicable Disesase Prevention and Control Unit, Wellbeing, Integrated Care and Ageing, Department of Health, Victoria; 2010.

63. Brown L, El-Hayek C, Franklin L, et al. Surveillance of notifiable infectious diseases in Victoria, 2008. In: Health protection Branch, ed. Melbourne: The Communicable Disease Prevention and Control Unit, Wellbeing, Integrated Care and Ageing, Department of Health, Victoria; 2011.

64. Brown L, Donnan E, El-Hayek C, et al. Surveillance of notifiable infectious diseases in Victoria 2009. In: Health Protection Branch, ed. Melbourne: Communicable Disease Control Unit, Rural & Regional Health and Aged Care Services, Department of Human Services, Victoria; 2011.

65. Hall G, The OzFoodNet Working Group. How much gastroenteritis in Australia is due to food? Estimating the incidence of foodborne gastroenteritis in Australia; 2004.

66. McIver CJ, Hansman G, White P, Doultree JC, Catton M, Rawlinson WD. Diagnosis of enteric pathogens in children with gastroenteritis. Pathology 2001;33:353-8.

67. Kirkwood CD, Clark R, Bogdanovic-Sakran N, Bishop RF. A 5-year study of the prevalence and genetic diversity of human caliciviruses associated with sporadic cases of acute gastroenteritis in young children admitted to hospital in Melbourne, Australia (1998-2002). J Med Virol 2005;77:96-101.

68. Karunajeewa H, Siebert D, Hammond R, Garland S, Kelly H. Seroprevalence of varicella zoster virus, parvovirus B19 and Toxoplasma gondii in a Melbourne obstetric population: implications for management. Australian and New Zealand Journal of Obstetrics and Gynaecology 2001;41:23-8.

69. Pappas G, Roussos N, Falagas ME. Toxoplasmosis snapshots: global status of Toxoplasma gondii seroprevalence and implications for pregnancy and congenital toxoplasmosis. Int J Parasitol 2009;39:1385-94.

70. Tam CC, Rodrigues LC, Petersen I, Islam A, Hayward A, O'Brien SJ. Incidence of Guillain-Barre syndrome among patients with Campylobacter infection: a general practice research database study. J Infect Dis 2006;194:95-7.

71. McCarthy N, Giesecke J. Incidence of Guillain-Barre syndrome following infection with Campylobacter jejuni. Am J Epidemiol 2001;153:610-4.

72. Allos BM. Association between Campylobacter infection and Guillain-Barre syndrome. J Infect Dis 1997;176 Suppl 2:S125-8.

73. McPherson M, Kirk MD, Raupach J, Combs B, Butler JR. Economic costs of Shiga toxin-producing Escherichia coli infection in Australia. Foodborne Pathog Dis 2011;8:55-62.

74. Mead PS, Griffin PM. Escherichia coli O157:H7. Lancet 1998;352:1207-12.


75. Beutin L, Zimmermann S, Gleier K. Human infections with Shiga toxin-producing Escherichia coli other than serogroup O157 in Germany. Emerg Infect Dis 1998;4:635-9.

76. McPherson M, Lalor K, Combs B, Raupach J, Stafford R, Kirk MD. Serogroup-specific risk factors for Shiga toxin-producing Escherichia coli infection in Australia. Clin Infect Dis 2009;49:249-56.

77. Vally H, Hall G, Dyda A, et al. Epidemiology of Shiga toxin producing Escherichia coli in Australia, 2000-2010. BMC Public Health 2012;12:63.

78. Haagsma JA, Siersema PD, De Wit NJ, Havelaar AH. Disease burden of post-infectious irritable bowel syndrome in the Netherlands. Epidemiol Infect 2010;138:1650-6.

79. Schonberg-Norio D, Mattila L, Lauhio A, et al. Patient-reported complications associated with Campylobacter jejuni infection. Epidemiol Infect 2010;138:1004-11.

80. Doorduyn Y, Van Pelt W, Siezen CL, et al. Novel insight in the association between salmonellosis or campylobacteriosis and chronic illness, and the role of host genetics in susceptibility to these diseases. Epidemiol Infect 2008;136:1225-34.

81. Townes JM, Deodhar AA, Laine ES, et al. Reactive arthritis following culture-confirmed infections with bacterial enteric pathogens in Minnesota and Oregon: a population-based study. Ann Rheum Dis 2008;67:1689-96.

82. Schiellerup P, Krogfelt KA, Locht H. A comparison of self-reported joint symptoms following infection with different enteric pathogens: effect of HLA-B27. J Rheumatol 2008;35:480-7.

83. Pope JE, Krizova A, Garg AX, Thiessen-Philbrook H, Ouimet JM. Campylobacter reactive arthritis: a systematic review. Semin Arthritis Rheum 2007;37:48-55.

84. Hannu T, Kauppi M, Tuomala M, Laaksonen I, Klemets P, Kuusi M. Reactive arthritis following an outbreak of Campylobacter jejuni infection. J Rheumatol 2004;31:528-30.

85. Rees JR, Pannier MA, McNees A, Shallow S, Angulo FJ, Vugia DJ. Persistent diarrhea, arthritis, and other complications of enteric infections: a pilot survey based on California FoodNet surveillance, 1998-1999. Clin Infect Dis 2004;38 Suppl 3:S311-7.

86. Hannu T, Mattila L, Rautelin H, et al. Campylobacter-triggered reactive arthritis: a population-based study. Rheumatology (Oxford) 2002;41:312-8.

87. Locht H, Krogfelt KA. Comparison of rheumatological and gastrointestinal symptoms after infection with Campylobacter jejuni/coli and enterotoxigenic Escherichia coli. Ann Rheum Dis 2002;61:448-52.

88. Arnedo-Pena A, Beltran-Fabregat J, Vila-Pastor B, et al. Reactive arthritis and other musculoskeletal sequelae following an outbreak of Salmonella hadar in Castellon, Spain. J Rheumatol 2010;37:1735-42.

89. Lee AT, Hall RG, Pile KD. Reactive joint symptoms following an outbreak of Salmonella typhimurium phage type 135a. J Rheumatol 2005;32:524-7.


90. Buxton JA, Fyfe M, Berger S, Cox MB, Northcott KA. Reactive arthritis and other sequelae following sporadic Salmonella typhimurium infection in British Columbia, Canada: a case control study. J Rheumatol 2002;29:2154-8.

91. Hannu T, Mattila L, Siitonen A, Leirisalo-Repo M. Reactive arthritis following an outbreak of Salmonella typhimurium phage type 193 infection. Ann Rheum Dis 2002;61:264-6.

92. Locht H, Molbak K, Krogfelt KA. High frequency of reactive joint symptoms after an outbreak of Salmonella enteritidis. J Rheumatol 2002;29:767-71.

93. Rudwaleit M, Richter S, Braun J, Sieper J. Low incidence of reactive arthritis in children following a salmonella outbreak. Ann Rheum Dis 2001;60:1055-7.

94. Dworkin MS, Shoemaker PC, Goldoft MJ, Kobayashi JM. Reactive arthritis and Reiter's syndrome following an outbreak of gastroenteritis caused by Salmonella enteritidis. Clin Infect Dis 2001;33:1010-4.

95. McColl GJ, Diviney MB, Holdsworth RF, et al. HLA-B27 expression and reactive arthritis susceptibility in two patient cohorts infected with Salmonella typhimurium. Aust N Z J Med 2000;30:28-32.

96. Urfer E, Rossier P, Mean F, et al. Outbreak of Salmonella braenderup gastroenteritis due to contaminated meat pies: clinical and molecular epidemiology. Clin Microbiol Infect 2000;6:536-42.

97. Mattila L, Leirisalo-Repo M, Pelkonen P, Koskimies S, Granfors K, Siitonen A. Reactive arthritis following an outbreak of Salmonella bovismorbificans infection. J Infect 1998;36:289-95.

98. Thomson GT, DeRubeis DA, Hodge MA, Rajanayagam C, Inman RD. Post-Salmonella reactive arthritis: late clinical sequelae in a point source cohort. Am J Med 1995;98:13-21.

99. Huovinen E, Sihvonen LM, Virtanen MJ, Haukka K, Siitonen A, Kuusi M. Symptoms and sources of Yersinia enterocolitica-infection: a case-control study. BMC Infect Dis 2010;10:122.

100. Hannu T, Mattila L, Nuorti JP, et al. Reactive arthritis after an outbreak of Yersinia pseudotuberculosis serotype O:3 infection. Ann Rheum Dis 2003;62:866-9.

101. Poropatich KO, Walker CL, Black RE. Quantifying the association between Campylobacter infection and Guillain-Barre syndrome: a systematic review. J Health Popul Nutr 2010;28:545-52.

102. McGrogan A, Madle GC, Seaman HE, de Vries CS. The epidemiology of Guillain-Barre syndrome worldwide. A systematic literature review. Neuroepidemiology 2009;32:150-63.

103. Walker CL, Applegate JA, Black RE. Haemolytic-uraemic syndrome as a sequela of diarrhoeal disease. J Health Popul Nutr 2012;30:257-61.


104. Askar M, Faber MS, Frank C, et al. Update on the ongoing outbreak of haemolytic uraemic syndrome due to Shiga toxin-producing Escherichia coli (STEC) serotype O104, Germany, May 2011. Euro Surveill 2011;16.

105. Elliott EJ, Robins-Browne RM, O'Loughlin EV, et al. Nationwide study of haemolytic uraemic syndrome: clinical, microbiological, and epidemiological features. Arch Dis Child 2001;85:125-31.

106. van de Kar NC, Roelofs HG, Muytjens HL, et al. Verocytotoxin-producing Escherichia coli infection in hemolytic uremic syndrome in part of western Europe. Eur J Pediatr 1996;155:592-5.

107. Chaudhary N, Truelove S. The irritable colon syndrome: A study of the clinical features, predisposing causes, and prognosis in 130 cases. Q J Med 1962;31:307-22.

108. Spiller R, Garsed K. Postinfectious irritable bowel syndrome. Gastroenterology 2009;136:1979-88.

109. Schwille-Kiuntke J, Frick JS, Zanger P, Enck P. Post-infectious irritable bowel syndrome--a review of the literature. Z Gastroenterol 2011;49:997-1003.

110. Hannu T. Reactive arthritis. Best Pract Res Clin Rheumatol 2011;25:347-57.

111. Isomaki H, Raunio J, von Essen R, Hameenkorpi R. Incidence of inflammatory rheumatic diseases in Finland. Scand J Rheumatol 1978;7:188-92.

112. Kvien TK, Glennas A, Melby K, et al. Reactive arthritis: incidence, triggering agents and clinical presentation. J Rheumatol 1994;21:115-22.

113. Savolainen E, Kaipiainen-Seppanen O, Kroger L, Luosujarvi R. Total incidence and distribution of inflammatory joint diseases in a defined population: results from the Kuopio 2000 arthritis survey. J Rheumatol 2003;30:2460-8.

114. Soderlin MK, Kautiainen H, Puolakkainen M, et al. Infections preceding early arthritis in southern Sweden: a prospective population-based study. J Rheumatol 2003;30:459-64.

115. Townes JM. Reactive arthritis after enteric infections in the United States: the problem of definition. Clin Infect Dis 2010;50:247-54.

116. Hanova P, Pavelka K, Holcatova I, Pikhart H. Incidence and prevalence of psoriatic arthritis, ankylosing spondylitis, and reactive arthritis in the first descriptive population-based study in the Czech Republic. Scand J Rheumatol 2010;39:310-7.

117. O'Grady K, Tallis G, (Eds). Surveillance of Notifiable Infectious Diseases in Victoria 2000 Melbourne: Communicable Diseases Section, Public Health Division, Victorian Department of Human Services; 2001.


Documents

Glossary and Acronyms - Department of HealthFile/wFoodborne-illness-Australi… · Web viewBy applying this proportion of episodes due to food to the ... which is one of the most