Upload
others
View
5
Download
0
Embed Size (px)
Citation preview
An Investigation of Epidemiological Approaches for Syndromic Surveillance
of Cattle Health using Ontario Condemnation Data
by
Gillian Denise Alton
A Thesis
presented to
The University of Guelph
In partial fulfilment of requirements
for the degree of
Doctor of Philosophy
in
Population Medicine
Guelph, Ontario, Canada
© Gillian D. Alton, June, 2014
ABSTRACT
AN INVESTIGATION OF EPIDEMIOLOGICAL APPROACHES FOR
SYNDROMIC SURVEILLANCE OF CATTLE HEALTH USING
ONTARIO CONDEMNATION DATA
Gillian Denise Alton Advisors: Dr. O. Berke & Dr. D.L. Pearl
University of Guelph, 2014
This thesis is an investigation of quantitative methods for food animal syndromic
surveillance utilizing bovine abattoir condemnation data as a case study to illustrate approaches
to using Ontario provincial abattoir data. There has been little investigation into the suitability of
bovine abattoir condemnation data in Ontario for its use in a food animal syndromic surveillance
system and the quantitative methods necessary for this type of system.
Overall, it was found bovine condemnation data from provincially inspected abattoirs to
be useful for food animal syndromic surveillance since they provided a more regionally detailed
picture of emerging diseases in Ontario than data from federal abattoirs. Disease-related and non-
disease factors such as season and sales price were shown to have an impact on condemnation
rates, and accounting for relevant predictable factors considerably affects the results of
quantitative cluster detection methods. This was demonstrated by comparison of various space-
time scan statistics with distinct options to control for covariate information. The results from
this study found that model-adjusted approaches for controlling for covariates in scan statistics
appeared to perform best in terms of ability to include all important covariates and suitability for
use with bovine abattoir condemnation data. Furthermore, the efficiency of syndromic
surveillance was investigated by comparing various sentinel abattoir selection approaches to
reduce the number of sample sites while still maintaining the overall trends in the full dataset.
The most effective sentinel selection approach utilized data from abattoirs in operation all weeks
of the year, and this approach shows promise for the integration of sentinel sites into a sentinel
syndromic surveillance system.
While these findings suggest that bovine abattoir condemnation data would be suitable
for integration into a food animal syndromic surveillance system, there are some limitations
including data quality issues and current methodological approaches. Future research is
recommended to focus on the following before formalizing a food animal syndromic surveillance
system in Ontario: (i) developing an improved meat inspector training program; (ii) finding ways
to harmonize the condemnation process by standardizing definitions for reasons for
condemnation; and (iii) validating methodological findings from this thesis by studying
simulated and/or documented historical outbreak data.
iv
ACKNOWLEDGEMENTS
I am fortunate to have so many supporters who have helped me along my PhD. I would
like to thank my advisory committee Dr. Olaf Berke, Dr. David Pearl, Dr. Bruce McNab and Dr.
Ken Bateman for all of their support and guidance throughout my graduate studies. To my co-
advisors, Dr. Berke and Dr. Pearl, Dr. Berke, thank you for your support and guidance
throughout my graduate studies. I cannot thank you enough for introducing me to the world of
spatial epidemiology, which I have grown to love, and for encouraging me to pursue a PhD even
on the first day of my Masters. Dr. Pearl, thank you for your creative thinking and encouraging
me to try out new ideas. I appreciate all of the support and advice you have both given me, both
academically and professionally as I have started my career in Public Health. Dr. McNab, thank
you for your advice and insight during the course of my program and for connecting me with
helpful contacts at the Ontario Ministry of Agriculture and Food (OMAF). Dr. Bateman, thank
you for assisting me with connecting my statistical results with practical animal health
explanations. I could not have completed my PhD without such a supportive team as an advisory
committee.
I would also like to thank OMAF for funding and data support for this project as well as
funding through the OMAF/University of Guelph Research Program. I would also like to
especially thank OMAF staff including Dr. Ab Rehmtulla, Dr. Alexandra Reid, Dr. Leslie
Woodcock, Jeff Perkins and Amy Rutgers-Kelly for their assistance with data retrieval and
assisting with questions over the course of this project. I would like to acknowledge the Canada
Foundation for Innovation (CFI) and the Ontario Research Fund for their support for
v
infrastructure and computing equipment in Dr. David Peal’s computer lab. In addition, I would
like to thank the Ontario Graduate Scholarship for funding to help support my graduate studies.
Thank you to my family for their constant love and support. It has been a very long road,
especially during the past two years, and I could not have done this without your continual
support and encouragement. I would like to especially thank my parents for encouraging me to
work hard and for supporting all of my dreams, even if my aspirations involved spending 12
years in school!
To my wonderful partner Jayson who has been by my side through all of my
undergraduate and graduate studies. You were always there to cheer me up when things were not
going as planned, and encouraged me to keep working when I wanted to throw my computer out
the window, and most importantly, you were there to celebrate all of my successes. I could not
have completed this thesis without your love and support.
vi
STATEMENT OF WORK
Data for this project were collected by the Ontario Ministry of Agriculture and Food
(OMAF) and provided to the University of Guelph for this thesis. All data cleaning, merging and
assessment were performed by Gillian Alton in consultation with personnel from the OMAF.
OMAF personnel also provided assistance and clarification with respect to data quality and
issues throughout this thesis. All data analyses were performed by Gillian Alton. Writing and
manuscript generation was performed by Gillian Alton. Editing and advice concerning analyses
for the thesis in its entirety was received from Drs. David Pearl, Olaf Berke, Ken Bateman and
Bruce McNab.
vii
TABLE OF CONTENTS ABSTRACT .................................................................................................................................................. ii
ACKNOWLEDGEMENTS ......................................................................................................................... iv
STATEMENT OF WORK .......................................................................................................................... vi
TABLE OF CONTENTS ............................................................................................................................ vii
LIST OF TABLES ........................................................................................................................................ x
LIST OF FIGURES .................................................................................................................................... xii
CHAPTER ONE ........................................................................................................................................... 1
INTRODUCTION, LITERATURE REVIEW, STUDY RATIONALE AND OBJECTIVES .................... 1
INTRODUCTION .................................................................................................................................... 1
LITERATURE REVIEW ......................................................................................................................... 3
Overview of the Ontario provincial abattoir system and abattoir surveillance in Ontario........................ 3
Current and historical uses of abattoir data ............................................................................................... 5
Recent major health events in Canadian animal agriculture ................................................................. 5
Use of abattoirs for surveillance ........................................................................................................... 7
Types of surveillance systems................................................................................................................... 8
Active vs. passive surveillance ............................................................................................................. 8
Syndromic surveillance ......................................................................................................................... 9
Sentinel surveillance ........................................................................................................................... 13
Quantitative methods for disease surveillance ........................................................................................ 15
Tests for space-time interaction .......................................................................................................... 15
Cumulative sum (CUSUM) methods .................................................................................................. 16
Scan statistics ...................................................................................................................................... 16
Model-based approaches ..................................................................................................................... 17
STUDY RATIONALE AND OBJECTIVES ......................................................................................... 19
REFERENCES ....................................................................................................................................... 21
CHAPTER TWO ........................................................................................................................................ 27
FACTORS ASSOCIATED WITH WHOLE CARCASS CONDEMNATION RATES IN
PROVINCIALLY-INSPECTED ABATTOIRS IN ONTARIO 2001 – 2007: IMPLICATIONS FOR
FOOD ANIMAL SYNDROMIC SURVEILLANCE ................................................................................. 27
ABSTRACT ............................................................................................................................................ 27
viii
INTRODUCTION .................................................................................................................................. 29
METHODS ............................................................................................................................................. 32
Data source and variables ................................................................................................................... 32
Statistical analysis ............................................................................................................................... 34
RESULTS ............................................................................................................................................... 35
Descriptive statistics ........................................................................................................................... 35
Statistical models ................................................................................................................................ 37
DISCUSSION ......................................................................................................................................... 38
CONCLUSIONS..................................................................................................................................... 42
REFERENCES .................................................................................................................................... 43
CHAPTER THREE .................................................................................................................................... 59
SUITABILITY OF BOVINE PORTION CONDEMNATIONS AT PROVINCIALLY-INSPECTED
ABATTOIRS IN ONTARIO CANADA FOR FOOD ANIMAL SYNDROMIC SURVEILLANCE ...... 59
ABSTRACT ............................................................................................................................................ 59
BACKGROUND .................................................................................................................................... 60
METHODS ............................................................................................................................................. 63
Data source and variables ................................................................................................................... 63
Statistical analysis ............................................................................................................................... 65
RESULTS ............................................................................................................................................... 67
Descriptive Statistics ........................................................................................................................... 67
Statistical models ................................................................................................................................ 68
DISCUSSION ......................................................................................................................................... 71
CONCLUSIONS..................................................................................................................................... 77
ACKNOWLEDGEMENTS .................................................................................................................... 78
REFERENCES ....................................................................................................................................... 79
CHAPTER FOUR ................................................................................................................................... 93
COMPARISON OF COVARIATE ADJUSTMENT METHODS USING SPACE-TIME SCAN
STATISTICS FOR FOOD ANIMAL SYNDROMIC SURVEILLANCE ............................................. 93
ABSTRACT ............................................................................................................................................ 93
BACKGROUND .................................................................................................................................... 95
METHODS ............................................................................................................................................. 98
ix
Data source and variables ................................................................................................................... 98
Space-time scan statistic ..................................................................................................................... 99
RESULTS ............................................................................................................................................. 101
“Parasitic liver” data ......................................................................................................................... 102
Pneumonic lung data ......................................................................................................................... 102
DISCUSSION ....................................................................................................................................... 103
CONCLUSIONS................................................................................................................................... 107
ACKNOWLEDGEMENTS .................................................................................................................. 108
REFERENCES ..................................................................................................................................... 109
CHAPTER FIVE .................................................................................................................................. 118
SUITABILITY OF SENTINEL ABATTOIRS FOR SYNDROMIC SURVEILLANCE USING
PROVINCIALLY-INSPECTED BOVINE ABATTOIR CONDEMNATION DATA ....................... 118
ABSTRACT .......................................................................................................................................... 118
BACKGROUND .................................................................................................................................. 121
METHODS ........................................................................................................................................... 123
Data source ........................................................................................................................................ 123
Descriptive analyses .......................................................................................................................... 125
Statistical analyses ............................................................................................................................ 126
RESULTS ............................................................................................................................................. 127
Descriptive Statistics ......................................................................................................................... 127
Negative binomial models ................................................................................................................ 129
DISCUSSION ....................................................................................................................................... 130
CONCLUSIONS................................................................................................................................... 132
ACKNOWLEDGEMENTS .................................................................................................................. 133
REFERENCES ..................................................................................................................................... 134
CHAPTER SIX ..................................................................................................................................... 148
SUMMARY, DISCUSSION, CONCLUSIONS AND RECOMMENDATIONS ............................... 148
REFERENCES ..................................................................................................................................... 155
x
LIST OF TABLES
CHAPTER 2:
Table 2.1 – Number of provincially inspected abattoirs in Ontario 2001 – 2007.........................45
Table 2.2 – Reason for whole carcass condemnation in provincially inspected abattoirs in
Ontario 2001-2007.........................................................................................................................46
Table 2.3 – Summary of number of cattle processed, number weeks open, audit rating and animal
class................................................................................................................................................48
Table 2.4 – Univariable negative binomial models using GEE approach……………………….49
Table 2.5 – Comparison of GEE fitted models using QIC statistic……………………………...51
Table 2.6 – Multivariable negative binomial model using GEE approach………………………52
Table 2.7 – Linear combinations of condemnation rates in cows and calves……………………55
CHAPTER 3:
Table 3.1 – Summary of Ontario provincial abattoir bovine portion condemnations 2001-
2007................................................................................................................................................82
Table 3.2 – Summary of number cattle processed, number weeks open, audit rating and animal
class………………………………………………………………………………………………83
Table 3.3 – Multivariable multi-level Poisson regression using pneumonic lung condemnation
rates………………………………………………………………………………………………84
Table 3.4 – Multivariable multi-level Poisson regression using “parasitic liver” condemnation
rates………………………………………………………………………………………………86
CHAPTER 4:
Table 4.1 – Results of the space-time scan statistic using “parasitic liver” condemnation
data……………………………………………………………………………………………...111
Table 4.2 – Results of the space-time scan statistic using pneumonic lung condemnation
data……………………………………………………………………………………………...114
CHAPTER 5:
Table 5.1 – Summary of sentinel abattoirs selection approaches for Ontario provincial abattoirs
(2001 – 2007)…………………………………………………………………………………...136
xi
Table 5.2 – Percentage of shared abattoirs between sentinel selection approaches……………137
Table 5.3 – Distribution of provincial abattoirs 2001 – 2007 among census agricultural regions in
Ontario based on three sentinel selection approaches…………………………………………..138
Table 5.4 – Negative binomial regression models comparing monthly pneumonic lung
condemnation rates for all abattoirs for each animal class with three methods of sentinel abattoir
selection………………………………………………………………………………………...139
Table 5.5 – Multivariable multilevel negative binomial regression model examining seasonal and
annual variability in monthly pneumonic lung condemnations for three sentinel selection
approaches………………………………………………………………………………………140
xii
LIST OF FIGURES
CHAPTER 2:
Figure 2.1 – Condemnation rates per 1000 cattle from Ontario provincial abattoirs 2001-
2007………………………………………………………………………………………………56
Figure 2.2 – Animal class condemnation rates per 1000 cattle from Ontario provincial abattoirs
2001-2007………………………………………………………………………………………..57
Figure 2.3 – Choropleth map of percentage of abattoirs processing cattle in Ontario per census
agricultural region………………………………………………………………………………..58
CHAPTER 3:
Figure 3.1 – Map of provincially-inspected abattoirs in Ontario 2001-2007……………………88
Figure 3.2 – “Parasitic liver” condemnation rates per 1000 slaughtered cattle from Ontario
provincial abattoirs 2001-2007…………………………………………………………………..89
Figure 3.3 - Pneumonic lung condemnation rates per 1000 slaughtered cattle from Ontario
provincial abattoirs 2001-2007…………………………………………………………………..90
Figure 3.4 – Model expected pneumonic lung condemnation rates based on multi-level Poisson
model……………………………………………………………………………………………..91
Figure 3.5 – Model expected “parasitic liver” condemnation rates based on multi-level Poisson
model……………………………………………………………………………………………92
CHAPTER 4:
Figure 4.1 – Results of space-time scan statistic using “parasitic liver” portion condemnation
data from Ontario provincial abattoirs 2001 – 2007 using four approaches for covariate
adjustment compared to the unadjusted data including: A) Unadjusted Poisson space-time scan
test, B) Animal-adjusted Poisson space-time scan test, C) Model-adjusted space-time scan test
and D) Space-time permutation scan test………………………………………………………116
Figure 4.2 – Space-time scan statistic using pneumonic lung condemnation data from Ontario
provincial abattoirs 2001 -2007 using four approaches for covariate adjustment compared to the
unadjusted data including: A) Unadjusted Poisson space-time scan test, B) Animal-adjusted
Poisson space-time scan test, C) Model-adjusted space-time scan test and D) Space-time
permutation scan test and E) Space-time scan test of residuals………………………………...117
xiii
CHAPTER 5:
Figure 5.1 – Pneumonic lung condemnation rates per 1000 slaughtered cattle for each animal
class for all abattoirs processing cattle throughout the study period…………………………...142
Figure 5.2 – Boxplots comparing the pneumonic lung condemnation rates per 1000 slaughtered
cattle for each sentinel abattoir selection approach…………………………………………….143
Figure 5.3 – Comparison of pneumonic lung condemnation rates in calves for sentinel site
selection approaches and full dataset 2001 – 2007……………………………………………..144
Figure 5.4 - Comparison of pneumonic lung condemnation rates in cows for sentinel site
selection approaches and full dataset 2001 – 2007……………………………………………..145
Figure 5.5 - Comparison of pneumonic lung condemnation rates in heifers for sentinel site
selection approaches and full dataset 2001 – 2007……………………………………………..146
Figure 5.6 - Comparison of pneumonic lung condemnation rates in steers for sentinel site
selection approaches and full dataset 2001 – 2007……………………………………………147
1
CHAPTER ONE:
INTRODUCTION, LITERATURE REVIEW, STUDY RATIONALE AND
OBJECTIVES
INTRODUCTION
With the emergence and re-emergence of novel infectious and zoonotic diseases
in recent years, there is an increased concern and awareness surrounding novel methods
for disease surveillance for the purposes of outbreak detection at the human-animal
interface [1]. Syndromic surveillance is one such novel approach that has been used to a
greater extent in various public health applications [2-4] in the past 20 years, but has been
explored more recently for the detection of disease outbreaks in animal health
applications [5-7]. In an effort to improve the early detection of animal disease outbreaks
in Ontario, the Ontario Ministry of Agriculture and Food (OMAF) was interested in
determining the suitability of utilizing novel data sources for animal disease surveillance
[7-9], including the use of bovine abattoir condemnation data for food animal syndromic
surveillance and the use of cluster detection methods for the rapid identification of
potential outbreaks of infectious animal diseases.
A great deal of research has been conducted on the application of quantitative
methods for disease surveillance [10]. However, it is important to remember that animal
health data, and in particular, provincial abattoir condemnation data have unique
characteristics which can create “noise” in the data. In provincial abattoir condemnation
data, for instance, non-disease factors such as season, price of animal class, animal
throughput, and processing capacity of abattoir need to be considered in understanding
2
the factors that influence condemnation rates. The selection, application and
interpretation of quantitative methods for disease surveillance needs to account for both
disease and non-disease specific factors that can lead to biases in the surveillance system.
This is particularly important when using animal health data to minimize the occurrence
of false alarms, and to avoid missing true disease outbreaks.
This thesis focuses on the novel approach of investigating the use of bovine
abattoir condemnation data from provincially-inspected abattoirs in Ontario for
syndromic surveillance of emerging infectious and zoonotic diseases, and assesses the
suitability of various quantitative cluster detection methods for integration into a food
animal syndromic surveillance system. The aim of the following narrative review is to
highlight recent examples from the past 20 years of veterinary syndromic surveillance
and key considerations for the application of quantitative cluster detection methods for
Ontario provincial abattoir condemnation data through the proceeding topics:
1. Overview of the Ontario provincial abattoir system and abattoir surveillance in
Ontario;
2. Current and historical applications of abattoir data;
3. Overview of surveillance approaches; and
4. Quantitative methods for disease surveillance.
3
LITERATURE REVIEW
Overview of the Ontario provincial abattoir system and abattoir surveillance in
Ontario
According to the 2011 Census of Agriculture, the Ontario cattle population
consisted of approximately 1.7 million cattle from 20,349 farms, representing
approximately 14% of the Canadian cattle population and 24% of Canadian farms.
Ontario has two types of abattoirs: provincial and federal [11]. The provincial abattoirs
process livestock for consumption within provincial boundaries [12]. These abattoirs are
generally small to medium sized establishments in terms of processing capacity, yet more
abundant in numbers compared to federal abattoirs. Federal abattoirs process livestock
for distribution inter-provincially and internationally [12]. As of September 2013, there
were seven federally-inspected [13] and 100 provincially inspected [14] meat
establishments processing cattle in Ontario and 95% of Canadian cattle were slaughtered
at federal abattoirs [15].
Meat inspection is the largest component of Canada’s food processing industry
[15]. Meat inspection regulations in Ontario are enforced by both federal and provincial
jurisdictions. Since 1997, federal abattoirs are licensed and regulated by the Canadian
Food Inspection Agency (CFIA). The CFIA continually performs inspection activities in
all federally registered abattoirs to ensure that meat products are processed according to
the Meat Inspection Act and Meat Inspection Regulations. Alternatively, the Food
Inspection Branch of the Ontario Ministry of Agriculture and Food (OMAF) regulates
meat inspection at provincial abattoirs through the Food Safety and Quality Act
(“FSQA”) and Regulation 31/05 (“Regulation”) [16]. OMAF has inspectors who inspect
the abattoirs and free standing meat plants and auditors who audit the plants. The meat
4
inspection regulations stipulate that every carcass and half carcass at the plant is subject
to a post-mortem inspection by a provincial inspector before being approved for use as
food [16]. Slaughter and condemnation data from all provincially-inspected abattoirs are
collected on a daily basis by OMAF using the Food Safety Decision Support System
(FSDSS) [12]. This system represents a potential wealth of information for food animal
syndromic surveillance, yet has been under-utilized to date.
In a variety of countries, abattoirs have previously been utilized for surveillance
of a variety of diseases important to both human and animal health [17-22]. While
abattoirs have been previously used to monitor animal diseases and lesions, these
programs were mainly designed to provide feedback to producers and stakeholders about
production performance, disease occurrence, and the impact of disease at the herd and
national levels [23]. However, little research has been done to examine the usefulness of
abattoir condemnation data for application in a food animal syndromic surveillance
system, particularly in cattle in Ontario, with the specific purpose to outbreak detection.
Compared to federal abattoirs, provincial abattoir condemnation data present a unique
opportunity to be utilized for surveillance as these data appear to be spatially sensitive.
Anecdotal evidence has suggested that cattle shipped to provincial abattoirs originate
from relatively local farms [12]. However, as farm of origin data are not currently
associated with Ontario provincial abattoir condemnation data this has never been
substantiated in the literature. This is an important characteristic of provincial abattoirs in
Ontario and pertinent spatial-temporal surveillance, as detection of disease clusters at the
abattoir may signal problems which originate in the surrounding cattle population. In
5
addition, due to international trade regulations at federal abattoirs, cattle of poorer quality
(i.e., poor-doers) may be more likely to be sent to provincial abattoirs [24].
Current and historical uses of abattoir data
Recent major health events in Canadian animal agriculture
Many countries worldwide, including Canada and specifically Ontario, have
experienced a re-emergence of infectious diseases in food animal populations over the
past decade. This includes outbreaks of a new strain of porcine circovirus type II (PCV-2)
in swine in 2004 [25] and outbreaks of bovine viral diarrhea (BVD) with enhanced
virulence in cattle from 1993-1995 [26]. BVD is an acute, highly contagious viral disease
in cattle. Infected animals usually exhibit signs of fever, diarrhea and mucosal lesions
[26]. The disease was first described in the literature in 1946 in New York State [27].
Although some reports have described the disease to be severe, prior to the early 1990’s,
the disease was considered to be mild, short in duration and have low mortality rates in
cattle of all ages [28, 29]. However, in 1993, bovine viral diarrhea virus (BVDV) with
enhanced virulence caused unprecedented outbreaks of severe BVD in Ontario and
Quebec. The outbreak lasted 3 years and caused an estimated $40,000 - $100,000 in
economic losses per herd due to lost animals, decreased milk production and abortion
[26]. A retrospective review of the outbreak data in Ontario by Carman et al. found
unusual and significantly higher median crude mortality rates in adult dairy cattle of
infected herds [26]. This finding suggests that perhaps the monitoring of dead stock or
mortality rates in cattle could have been used as an early warning system for the
outbreak.
6
Animal disease outbreaks are not only detrimental to the animals; they can have
devastating and lasting effects on the producers and processors, particularly if they
impact international trade. This type of event was seen when a case of bovine spongiform
encephalopathy (BSE), was detected in Alberta, Canada in 2003. Bovine spongiform
encephalopathy is a progressive and fatal prion infection in cattle causing damage to the
central nervous system. It belongs to a larger family of diseases called transmissible
spongiform encephalopathies (TSEs) which includes scrapie in sheep and goats, chronic
wasting disease in deer, elk and moose, and classic and variant Creutzfeldt-Jakob disease
in humans [30]. The incubation period for BSE is quite long with most cattle exhibiting
signs approximately 3-6 years after infection. Signs of BSE include behavioural changes,
coordination problems, weight loss and decreased milk production [30]. Though much
research has been completed, currently there is no test to diagnose BSE in live animals.
Diagnosis is generally made through clinical signs and confirmed under Western Blot for
the presence of prions in the brain tissue [31].
Bovine spongiform encephalopathy is speculated to have originated in the early
1970’s from the use of rendered ruminant protein containing tissue of infected animals in
the United Kingdom [31]. While there is still more to be learned about the transmission
of BSE, it is believed the primary source of infection is through contaminated feed
through the use of meat and bone meal containing protein from rendered infected
ruminants [30]. Within hours of the confirmation of the first BSE case in Canada, the
United States government announced an immediate ban of all imports of Canadian beef
[31]. The closure of international trade borders to Canadian cattle lasted for 26 months.
This caused the price of cattle to drop significantly throughout the country, and it took
7
several years to recover [32]. These economic impacts may be associated with the quality
of animals being sent to slaughter, as producers consider shipping costs against possible
return of shipping an animal of suspect health, thus impacting condemnation rates. The
story of BSE in Canada raises the importance of early identification and prevention of
infectious and zoonotic diseases in food animals, as they have the ability to have lasting
and far-reaching effects on the agricultural industry and international trade.
Use of abattoirs for surveillance
The use of abattoirs for disease surveillance is not a novel concept. Abattoirs have
been utilized successfully as a source of data collection and surveillance for many years.
Historically, abattoirs have been used primarily for targeted surveillance of a specific
pathogen, or to quantify the presence/absence of certain lesions or condemnations [33-
39]. Studies by Van Donkersgoed et al. [38] and Schamber et al. [39] in the 1980’s and
1990’s described the condemnations in cattle presented for slaughter at federally-
inspected abattoirs in Canada and the United States. These studies highlighted liver [38]
and pneumonic lung condemnations [39] as being among the most frequently reported
reasons for condemnation of cattle. Similar studies have also been conducted using other
species including broilers in Danish abattoirs [34], or sheep and goats from abattoirs in
Tanzania [36].
Other studies using abattoir data have focused on specific pathogen surveillance
to improve food safety. For example, a study by Alban et al. [40] used data from the
Danish Salmonella surveillance and control programme for finisher pigs and found that
meat-juice samples taken from finisher pigs at the time of slaughter was an effective way
to identify high-risk herds for Salmonella. Similarly, a study by Osaili et al. [35],
8
collected samples from cattle slaughtered in Amman to determine the prevalence of
Escherichia coli O157:H7 and respective antimicrobial susceptibility.
Types of surveillance systems
Active vs. passive surveillance
Disease surveillance is the ongoing systematic monitoring of health of a
population, based on the collection, interpretation and dissemination of health data and
information [41]. Disease surveillance is a broad term which encompasses many different
approaches for surveillance. Each surveillance approach has advantages and
disadvantages which make them suitable for certain circumstances. These are defined by
the overall objective, context and the available resources under which a surveillance
program is conducted [42, 43]. Broadly, surveillance systems are classified into active
and passive. Passive surveillance involves the reporting of a particular disease as it is
diagnosed. Health authorities do not stimulate reporting by reminding health care workers
to report disease or provide feedback to individual health care workers. For example, in
human medicine, Canada has a published list of notifiable diseases and health care
workers are supposed to report a case of a notifiable disease to the local health
department on a case-by-case basis [44]. Similarly, in veterinary medicine, the Canadian
Food Inspection Agency (CFIA) has a list of reportable diseases outlined in the Health of
Animals Act and Reportable Disease Regulations which are to be reported to a CFIA
district veterinarian on a case-by-case basis [45]. The advantage of this approach to
surveillance is that it is simple and not burdensome to the health authority; however,
there can be an issue with incompleteness and variability in reporting. As a result, passive
9
surveillance systems may not be representative and may not always be sensitive enough
to identify outbreaks. In contrast, active surveillance involves regular outreach to data
reporters, such as health care workers and stimulates the reporting of specific diseases
[43, 44]. Active surveillance systems can be very costly and labour-intensive; therefore,
active surveillance systems are often used for limited time periods and for specific
purposes. Other major descriptors of surveillance include syndromic and sentinel
surveillance which may also be described as active or passive [43, 44].
Syndromic surveillance
Syndromic surveillance involves the grouping of large numbers of similar
signs/symptoms or surrogate data utilizing non-traditional data sources. The groupings of
signs/symptoms are loosely designated as ‘syndromes’. This information can be used to
track disease trends in a population and signal an aberration that may warrant further
investigation [46]. The objective of syndromic surveillance is to detect clusters of
symptoms earlier than traditional surveillance data sources (e.g., the detection of specific
pathogens through laboratory surveillance). Syndromic surveillance is a relatively new
tool for animal and public health. The concept was derived from a gastrointestinal disease
outbreak in Milwaukee in 1993. The outbreak was eventually traced to Cryptosporidium
in the water supply. However, after the outbreak was over, it was discovered that the
sales of over-the-counter anti-diarrheal medications had increased more than three-fold
weeks before health officials were alerted to the outbreak [47]. Since 1993, much
advancement has been made in the area of syndromic surveillance, in terms of
quantitative methods and practical applications of the methodology. Systems have been
developed to monitor data on school absenteeism, sales of over-the-counter products,
10
calls to nurse hotlines and counts of hospital emergency room admissions [2-4]. In
theory, syndromic surveillance shows great promise for the detection of disease outbreaks
in a variety of settings. However, in practical applications, syndromic surveillance
systems can be burdened with continuous false-alarms [48].
While syndromic surveillance has been utilized in public health settings for nearly
two decades, the application of such methods in an animal health context have only
recently been explored within the last few years using a variety of novel data inputs [5-8].
Amezcua et al. conducted a pilot study to determine the feasibility of utilizing a
veterinary-based syndromic surveillance system to detect disease outbreaks in Ontario
swine [7]. The surveillance system was evaluated in terms of timeliness, compliance,
geographic coverage and data quality. The study found that obtaining information from
veterinarians can be valuable, reliable and relatively inexpensive for the collection of data
from a large proportion of Ontario’s swine farms. However, it was found that the
timeliness of data transmission was too slow to identify disease trends on a weekly basis
and more work needed to be done to improve timeliness of data and drop-offs in
compliance [7]. Similarly, clinical data from veterinary practitioners have been
investigated in other countries including France, New Zealand, United States, The
Netherlands and United Kingdom [49].
Recent veterinary syndromic surveillance research investigated the use of novel
data sources for the timely detection of disease outbreaks, as well as employing
laboratory submission requests, rather than the actual laboratory results in animal disease
syndromic surveillance. Studies by Dórea et al. [50-52], have been investigating the use
of laboratory submission counts for veterinary syndromic surveillance based on
11
submissions to the Animal Health Laboratory at the University of Guelph. As there is no
standard classification for veterinary syndromes, the researchers created 17 syndromic
groups based on the type of sample submitted and diagnostic test requested by the
veterinarian for cattle [50]. The researchers also investigated the performance of various
time-series algorithms for use in detection of potential outbreaks using simulated data and
found the developed system provided high sensitivity for detection of potential outbreak
signals for cattle [52]. Similarly, a study by O’Sullivan et al. [9], investigated the use of
veterinary diagnostic laboratory data from the Animal Health Laboratory at the
University of Guelph, with a focus on monitoring negative test results in swine. The
study hypothesized that the presence of a disproportionate number of negative test results
for a specific disease could be an indication of a novel disease outbreak. The study found
that during the PCV-2 associated disease outbreak which occurred in Ontario from
December 2005 – May 2006, the probability of a positive porcine reproductive and
respiratory syndrome virus (PRRSV) polymerase chain reaction at the Animal Health
Laboratory decreased. The study concluded that in order to detect a new disease agent
emerging in a particular population, it may be advantageous to monitor negative test
results of commonly used first-order tests for a known disease [9].
Other researchers have used alternative approaches to veterinary-based
surveillance and have investigated the use of clinical observations of animals entering a
livestock auction. A study by Van Metre et al. used visual inspection of animals outside
pens at a livestock auction as a potential data source for syndromic surveillance [5]. This
study utilized data from multiple species including cattle, sheep, goats, horses and pigs.
Livestock were visually inspected for clinical signs of disease by a veterinarian and
12
clinical signs were categorized into 12 pre-determined disease syndromes. The
frequencies of clinical signs and disease syndromes were then calculated. The most
commonly observed disease syndromes for all animal species were respiratory tract
diseases, thin body condition, and abnormal ambulation or posture, and together
accounted for approximately 93% of all clinical signs observed. Further studies are
needed to determine whether these methods are sensitive and specific enough to detect
disease outbreaks [5].
Ontario provincial abattoir condemnation data show great potential for the
monitoring of emerging infectious and zoonotic diseases, however, these data have been
rarely used, particularly for cattle in Ontario. A study by Thomas-Bachli et al.
investigated the suitability and limitations of using abattoir condemnation data for
syndromic surveillance of emerging diseases of swine in Ontario [8]. This study found
clusters of high condemnation rates for kidneys with nephritis in time and space-time
which preceded the timeframe during which case clusters of PCV-2 were detected using
traditional laboratory data. In addition, the trends found in kidney condemnation rates
related to nephritis lesions in Eastern Ontario were consistent with documented disease
outbreaks during that time period [8]. The findings from this study support the further
investigation of using abattoir condemnation data for food animal syndromic
surveillance. In addition, studies in the United States have also utilized abattoir data for
syndromic surveillance. A study by Engle [53], used condemnation data from the
electronic Animal Disposition Reporting System (eADRS) and found that a swine
erysipelas outbreak in Iowa and Minnesota during July 2001 could have been identified
approximately 10 months earlier if an automated surveillance system had been in place.
13
A similar study in the United States evaluated the use of condemnation data from swine
to set up an animal health monitoring system [53]. More recently, studies in Switzerland
[33] and France [54], have investigated the use of both portion and whole carcass
condemnations from abattoirs for syndromic surveillance purposes. No studies have
investigated the use of bovine abattoir condemnation data from provincially-inspected
abattoirs in Ontario.
False-alarm rate is an issue with livestock data, and in particular for Ontario
provincial abattoir condemnation data. There are many characteristics, such as the
changing background population and small throughput of animals at certain facilities,
which can contribute to false-alarms if there are predictable covariates that are not
corrected for in quantitative cluster detection methods. These false-alarms can mask the
presence of a true outbreak and use additional governmental time and resources for
investigation. The study by Thomas-Bachli et al. identified a number of non-disease
factors associated with lung and kidney condemnation rates, including abattoir processing
capacity, agricultural region and season which should be taken into account when
applying cluster detection methods for abattoir condemnation data [8].
Sentinel surveillance
Sentinel surveillance is a form of surveillance which involves a limited number of
recruited participants or organizations, such as farms, veterinarians, abattoirs, healthcare
providers or hospitals, which report on specific health events to identify disease presence
and/or trends in the general population [40]. Sentinel surveillance is an alternative to
population-based surveillance and is often used when population-based data collection is
not feasible [44]. Sentinel surveillance enables the collection of timely information in a
14
relatively inexpensive manner, rather than sampling information from the general
population. Sentinel surveillance has been previously used in both human and animal
health settings for a variety of health outcomes. Sentinel surveillance has been used to
monitor or identify outbreaks of infectious diseases and to monitor the activity of certain
health conditions which can change due to environmental conditions. Some notable
examples in human health include the Canadian Primary Care Sentinel Surveillance
Network (CPCSSN), which is Canada’s first multi-disease electronic record sentinel
surveillance system funded by the Public Health Agency of Canada. CPCSSN includes
10 practice-based research networks associated with departments of family medicine
across the country [55]. This system collects information on chronic diseases from a
select group of family doctors. A similar approach utilizing sentinel practitioners has
been employed in the United States by the United States Influenza Sentinel Physicians
Surveillance Network which provides weekly reports on the total number of patients seen
and the number of patients with influenza-like illness by age group [43]. Though used
less often in animal health applications, sentinel surveillance has been used successfully
for surveillance in various applications. For example, following the emergence of
Bluetongue virus serotype 8 in Central Europe in 2006, causing a large scale outbreak in
several countries in Europe in 2007, a Bluetongue sentinel surveillance program was
established in Belgium in 2010. This surveillance program intended to demonstrate the
absence of Bluetongue virus [56]. This program randomly selected a total of 300 dairy
herds, with 30 herds selected from each of the Belgian provinces. The criteria for
selection of herds was based on dairy herds that were expected to have a minimum of 15
animals present between 4 and 12 months of age at the start of the sentinel program [56].
15
Other studies have combined both sentinel and syndromic surveillance and utilized data
collected from sentinel veterinarians or veterinary practices on farms for syndromic
surveillance of emerging infectious diseases in animals [7, 49].
Quantitative methods for disease surveillance
With the emergence of previously unidentified infectious and zoonotic diseases in
recent years in both animal and public health, the need for sensitive and specific
quantitative methods for disease surveillance is growing. A variety of quantitative
methods for disease surveillance exist depending on the available data and the question to
be answered. Quantitative methods for spatio-temporal disease surveillance generally
seek to determine whether the incidence of the disease of interest in a spatially and
temporally defined subset is unusual compared to the incidence in the rest of the study
population [10]. There are various broad classifications for quantitative methods
available including: tests for space-time interaction, cumulative sum methods, scan
statistics and model-based approaches [10].
Tests for space-time interaction
Tests for space-time interaction involve the number of cases of disease that are
related in space and time and compare this to a null hypothesis of no interaction [57].
Examples of tests for space-time interaction include the Knox test [58], Mantel’s test [59]
and the space-time K-function [60]. The space-time interaction tests are only suitable for
testing cases of a single disease and can only report the presence/absence of spatio-
temporal dependence or clustering in the data [10].
16
Cumulative sum (CUSUM) methods
Cumulative sum (CUSUM) methods maintain a running total of the deviations
between observed and expected values [10]. If this total deviates beyond pre-determined
threshold levels, then an alarm will indicate a potential disease outbreak [46]. CUSUM
methods are designed to detect sudden changes in the mean value and originated from
industrial process control applications to monitor production quality [46]. This statistic
was expanded by Rogerson and Yamada in 2004 [61] to create a multivariate CUSUM
where the statistic is derived for a singular disease over multiple sub-regions, but could
also be used to monitor multiple diseases over multiple sub-regions. CUSUM methods
are effective for prospective disease surveillance; however, due to the sum of deviations,
given a long enough time period false alarms are inevitable. The false-positive rate is
controlled by the expected time it takes for a false alarm to be signalled (in-control
average run length), however, in practice this value can be difficult to specify and
remains a key problem with CUSUM methods [10].
Scan statistics
Cluster detection methods have become popular due to the increased availability
of geo-referenced health data and the desire to detect not only when a potential disease
outbreak is occurring, but also the geographical region where it is occurring. One such
method is the space-time scan statistic; which has been previously utilized in a variety of
human and animal health-related applications [62-64]. The scan statistic detects a local
excess of events and then tests if this excess may have occurred due to chance [46]. The
17
scan statistic was first studied with time-series data [65]. Since this time, the scan
statistic method has been extended to include the 2-dimensional scan statistics to detect
purely spatial disease clusters [66], and 3-dimensional scan statistics to detect disease
clusters in space and time [63]. The space-time scan statistic uses a cylindrical window
with a circular base corresponding to space and the height corresponding to time [63].
The cylinder moves in space and time so that for each possible geographic location and
size, it also visits each possible time interval. This creates a number of overlapping
cylinders of different sizes covering the entire study area. The scanning window
compares the number of cases inside potential cluster space to the expected number of
cases outside that space based on the population at risk and estimates a likelihood for
each cylinder. The test is based on the assumption of constant risk inside the scanning
window compared to outside the scanning window under the null hypothesis. Statistical
significance of the cluster is determined by Monte Carlo-based simulations [63]. The
scan statistic can be conducted using a variety of probability models including: Poisson,
Bernoulli, multinomial, ordinal, exponential and normal depending on the type of data
being used.
Model-based approaches
While spatial, temporal and spatio-temporal scan statistics have been widely used
in the literature, limitations in the ability of these statistical methods to adjust for
covariates remains an issue. Currently, the implementations of the scan statistic allow for
the adjustment of categorical covariates without interactions. Model-based quantitative
methods for disease surveillance address the current limitations of spatio-temporal cluster
detection methods and allow for the inclusion of further covariates that may influence the
18
incidence of the disease of interest. For example, the incidence of some diseases can vary
with age and gender and would need to be accounted for in the model [10]. Regression
models allow for the incorporation of continuous and categorical confounding variables,
as well as, complex interaction terms and adjust the disease risk in space and time. In
addition, the covariates can be evaluated for statistical significance with the outcome of
interest to ensure only statistically significant variables are being included in the model.
Generalized linear mixed models (GLMMs) are one example of a broad classification of
statistical models that have been used for space-time disease surveillance [10]. The
GLMM is a statistical regression-based approach to model an outcome of interest, such as
disease counts or rates using a variety of statistical distributions [10]. The GLMM is an
extension of the generalized linear model that includes random effects to account for one
or more levels of clustering or spatial hierarchy within the data [67]. A study by
Kleinman et al. proposed a general approach to evaluate whether observed counts in a
relatively small area are larger than would be expected on the basis of a history of
naturally occurring disease utilizing a generalized linear mixed model [68]. The statistical
methods described in the above study were of particular use for syndromic surveillance
of bioterrorism agents, specifically anthrax, where the initial symptoms are difficult to
distinguish from those of naturally occurring disease [68]. However, in comparison to
spatio-temporal cluster detection methods, model-based surveillance approaches require
additional tests to determine if incoming data differ from modelled pattern of cases and
are unable to pin-point the location of a significant cluster. Thus, surveillance workers
may use models to estimate a relative risk and then combine the results with specific
cluster detection methods. An example of this multiple-method approach is illustrated by
19
Kleinman et al. [48]. Their study introduced a method to adjust for naturally occurring
temporal trends or geographical patterns in lower respiratory tract complaints in humans
reported to a large group medical practice. The study demonstrated the importance of
adjusting for predictable non-disease factors such as day of week, month, holidays, as
well as disease factors related to the local history of disease, and found the number of
false alarms could be reduced by adjusting for “noise” in the data [48]. However, these
methods have never been applied to bovine abattoir condemnation data, and specifically
in Ontario, and may be helpful at reducing the background “noise” in abattoir
condemnation data and improve its use in a food animal syndromic surveillance system.
STUDY RATIONALE AND OBJECTIVES
In order to conduct sensitive surveillance of infectious and zoonotic diseases and
identify potential threats to animal and public health, government surveillance agencies
are seeking novel ways to detect disease outbreaks in a timely manner. Syndromic
surveillance has been demonstrated in human health settings to address the issue of
timely identification of disease outbreaks and now animal health surveillance workers
have started to follow suit. However, there has been little investigation into the suitability
of bovine abattoir condemnations for use in a food animal syndromic surveillance system
and the quantitative methods necessary for this type of system.
This thesis focuses on the application of quantitative methods for food animal
syndromic surveillance utilizing bovine abattoir condemnation data as a case study to
illustrate approaches to using Ontario provincial abattoir data. The following objectives
will be addressed to assess the usage and analysis of these data for syndromic
surveillance:
20
1) Estimate approximate distance cattle are shipped for slaughter to ensure cattle at
abattoirs are originating from ‘local’ farms (Chapter 2);
2) Identify potential predictable covariates in abattoir condemnation rate data and
their potential impact on food animal syndromic surveillance (Chapter 2, 3);
3) Compare the impact of correcting for predictable covariates during quantitative
cluster detection for syndromic surveillance using abattoir condemnation data
(Chapter 4); and
4) Assess the performance of sentinel site selection approaches in terms of regional
representation, and capturing temporal and demographic patterns (Chapter 5).
21
REFERENCES
1. Brown C: Emerging zoonoses and pathogens of public health significance - an
overview. Rev sci tech Off int Epiz 2004, 23:435-442.
2. Edge V, L., Pollari F, Ng LK, Michel P, McEwen SA, Wilson JB, Jerrett M, Sockett
PN, Martin SW: Syndromic surveillance of Norovirus using over-the-counter sales of
medications related to gastrointestinal illness. Can J Infect Dis Med Microbiol 2006,
17:235-241.
3. van Dijk A, McGuinness D, Rolland E, Moore KM: Can Telehealth Ontario
respiratory call volume be used as a proxy for emergency department respiratory
visit surveillance by public health? CJEM 2008, 10:18-24.
4. Hope K, Durrheim DN, Muscatello D, Merritt T, Zheng W, Massey P, Cashman P,
Eastwood K: Identifying pneumonia outbreaks of public health importance: can
emergency department data assist in earlier identification? Aust N Z J Public Health
2008, 32:361-364.
5. Van Metre DC, Barkey DQ, Salman MD, Morley PS: Development of a syndromic
surveillance system for detection of disease among livestock entering an auction
market. J Am Vet Med Assoc 2009, 234:658-664.
6. Vourc'h G, Bridges VE, Gibbens J, De Groot BD, McIntyre L, Poland R, Barnouin J:
Detecting emerging diseases in farm animals through clinical observations. Emerg
Infect Dis 2006, 12:204-210.
7. Amezcua M, Pearl D, Friendship R, McNab W: Evaluation of a veterinary-based
syndromic surveillance system implemented for swine. CJVR 2010, 74:241-250.
8. Thomas-Bachli AL, Pearl DL, Friendship RM, Berke O: Suitability and limitations
of portion-specific abattoir data as part of an early warning system for emerging
diseases of swine in Ontario. BMC Vet Res 2012, 8:3.
9. O'Sullivan T, Friendship R, Pearl D, McEwen B, Dewey C: Identifying an outbreak
of a novel swine disease using test requests for porcine reproductive and respiratory
syndrome as a syndromic surveillance tool. BMC Vet Res 8:192.
10. Robertson C, Nelson TA, MacNab YC, Lawson AB: Review of methods for space–
time disease surveillance. Spat Spattemporal Epidemiol 2010, 1:105-116.
11. Farm and farm operator data 2011[http://www29.statcan.gc.ca/ceag-
web/eng/community-agriculture-profile-profil
agricole?geoId=350000000&selectedVarIds=239%2C240%2C241%2C232%2C234%2C
242%2C233%2C236%2C235%2C238%2C237%2C]
22
12. Haines RJ: Report of the Meat Regulatory and Inspection Review: Farm to Fork - A
Stratgey for Meat Safety in Ontario. Toronto, ON: Ontario Ministry of the Attorney
General; 2004.
13. List of Federally Registered Meat Establishments and their Licensed Operators
[http://active.inspection.gc.ca/scripts/meavia/reglist/reglist.asp?lang=e]
14. Provincially Licensed Meat Plants
[http://www.omafra.gov.on.ca/english/food/inspection/meatinsp/licenced_operators_list.h
tm]
15. Canada's Meat Inspection System [http://inspection.gc.ca/food/information-for-
consumers/fact-sheets/specific-products-and-risks/meat-and-poultry-products/meat-
inspection-system/eng/1374559586662/1374559587537]
16. Meat Plant Guidelines
[http://www.omafra.gov.on.ca/english/food/inspection/meatinsp/manual/introduction.htm
]
17. Vilas VJDR, Bohning D, Kuhnert R: A comparison of the active surveillance of
scrapie in the European Union. Vet Res 2008, 39:37-52.
18. Hadorn DC, Stark KDC: Evaluation and optimization of surveillance systems for
rare and emerging infectious diseases. Vet Res 2008, 39:57-69.
19. Karama M, Johnson RP, Holtslander R, McEwen SA, Gyles CL: Prevalence and
characterization of verotoxin-producing Escherichia coli (VTEC) in cattle from an
Ontario abattoir. Can J Vet Res 2008, 72:297-302.
20. Kaneene JB, Miller R, Meyer RM: Abattoir surveillance: The U.S. experience. Vet
Microbiol 2006, 112:273-282.
21. Ansong-Danquah J: A survey of carcass condemnation at a poultry abattoir and
its application to disease management. Can Vet J 1987, 28:53-56.
22. Benschop J, Stevenson MA, Dahl J, Morris RS, French NP: Temporal and
longitudinal analysis of Danish Swine Salmonellosis Control Programme data:
implications for surveillance. Epidemiol Infect 2008, 136:1511-1520.
23. Christensen J, Ellegaard B, Kirkegaard Petersen BK, Willeberg P, Mousing J: Pig
Health and production surveillance in Denmark: sampling design, data recording,
and measures of disease frequency. Prev Vet Med 1994, 20:47-61.
24. Alton GD, Pearl DL, Bateman KG, McNab WB, Berke O: Factors associated with
whole carcass condemnation rates in provincially-inspected abattoirs in Ontario
23
2001-2007: implications for food animal syndromic surveillance. BMC Vet Res 2010,
6:42.
25. Poljak Z, Dewey CE, Rosendal T, Friendship RM, Young B, Berke O: Spread of
porcine circovirus associated disease (PCVAD) in Ontario (Canada) swine herds:
Part I. Exploratory spatial analysis. BMC Vet Res 2010, 5:59.
26. Carman S, van Dreumel T, Ridpath J, Hazlett M, Alves D, Dubovi E, Tremblay R,
Bolin S, Godkin A, Anderson N: Severe acute bovine viral diarrhea in Ontario, 1993-
1995. J Vet Diagn Invest 1998, 10:27-35.
27. Olafson P, MacCallum A, Fox F: An apparently new transmissible disease of
cattle. Cornell Vet 1946, 36:205-213.
28. Duffel S, Harkness J: Bovine virus diarrhea-mucosal disease infection in cattle.
Vet Rec 1985, 117:240-245.
29. Radostits O, Littlejohns I: New concepts in the pathogenesis, diagnosis and control
of diseases caused by the bovine viral diarrhea virus. Can Vet J 1988, 29:513-528.
30. Animal Disease Information
[http://www.aphis.usda.gov/animal_health/animal_diseases/bse/index.shtml]
31. Overview of Canada's BSE Safeguards
[http://www.inspection.gc.ca/animals/terrestrial-
animals/diseases/reportable/bse/safeguards/eng/1363896195473/1363896681768]
32. Beef Farmers of Ontario [http://www.cattle.guelph.on.ca/statistics/f-
p_inspectcattle.asp]
33. Vial F, Reist M: Evaluation of Swiss slaughterhouse data for integration in a
syndromic surveillance system. BMC Vet Res 2014, 10:33.
34. Lund VP, Kyvsgaard NC, Christensen JP, Bisgaard M: Pathological manifestations
observed in dead-on-arrival broilers at a Danish abattoir. Br Poult Sci 2013, 54:430-
440.
35. Osaili TM, Alaboudi AR, Rahahlah M: Prevalence and antimicrobial susceptibility
of Escherichia coli O157:H7 on beef cattle slaughtered in Amman abattoir. Meat Sci
2013, 93:463-468.
36. Mellau LS, Nonga HE, Karimuribo ED: A slaughterhouse survey of lung lesions in
slaughtered stocks at Arusha, Tanzania. Prev Vet Med 2010, 97:77-82.
24
37. Dorjee S, H., Dorjee S, Heuer C, Jackson R, West D, Collins-emerson J, Midwinter
A, Ridler A: Prevalence of pathogenic Leptospira spp. in sheep in a sheep-only
abattoir in New Zealand. N Z Vet J 56:164-170.
38. Van Donkerdgoed J, Jewison G, Mann M, Cherry B, Altwasser B, Lower R, Wiggins
K, Dejonge R, Thorlakson B, Moss E, Mills C, Grogan H: Canadian beef quality audit.
Can Vet J 1997, 38:217-225.
39. Schamber G, Olson C, Witt L: Condemnations in Cattle Presented for Slaughter.
Bovine Practice 1981, 2:19-21.
40. Alban L, B., Alban L, Baptista FM, Møgelmose V, Sørensen LL, Christensen H,
Aabo S, Dahl J: Salmonella surveillance and control for finisher pigs and pork in
Denmark - A case study. Food Res Int 45:656-665.
41. Teutsch SM, Churchill RE: Principles and Practice of Public Health Surveillance:
2nd
Ed. Oxford ; New York: Oxford University Press; 2000.
42. Halliday Jeb, Halliday JEB: Animal sentinel surveillance: evaluating domestic
dogs as sentinels for zoonotic pathogen surveillance.
http://hdl.handle.net.subzero.lib.uoguelph.ca/ 2010, :.
43. Salman MD: Animal Disease Surveillance and Survey Systems Methods and
Applications: Ames, Iowa : Iowa State Press; 2003.
44. Lee LM, Teutsch SM, Thacker BS, St. Louis ME: Principles and Practice of Public
Health Surveillance: 3rd
Ed. Oxford; New York: Oxford University Press; 2010.
45. Reportable Diseases [http://www.inspection.gc.ca/animals/terrestrial-
animals/diseases/reportable/eng/1303768471142/1303768544412]
46. Lawson AB, Kleinman K: Spatial and Syndromatic Surveillance for Public Health:
Chichester, West Sussex ; Hoboken, NJ: J. Wiley; 2005.
47. Stoto MA: Syndromic surveillance. Issues in Science & Technology 2005, 21:49-56.
48. Kleinman K, Abrams A, Kulldorff M, Platt R: A model-adjusted space-time scan
statistic with an application to syndromic surveillance. Epidemiol Infect 2005,
133:409-419.
49. Dórea F, Sanchez J, Revie CW: Veterinary syndromic surveillance: Current
initatives and potential for development. Prev Vet Med 2011, 101:1-17.
50. Dórea FC, Muckle CA, Kelton D, Mcclure JT, Mcewen BJ, Mcnab WB, Sanchez J,
Revie CW: Exploratory analysis of methods for automated classification of
25
laboratory test orders into syndromic groups in veterinary medicine. PloS one
8(3):e57334.
51. Dórea FC, Revie CW, McEwen BJ, McNab WB, Kelton D, Sanchez J: Retrospective
time series analysis of veterinary laboratory data: Preparing a historical baseline
for cluster detection in syndromic surveillance. Prev Vet Med 2013, 1:219-227.
52. Dórea FC, Mcewen BJ, Mcnab WB, Sanchez J, Revie CW: Syndromic surveillance
using veterinary laboratory data: algorithm combination and customization of
alerts. PloS one 8(12):e82183.
53. Weber WD: Development of an animal monitoring system based on slaughter
condemnation data. Miami: Proceedings of the Eighth International Society for Disease
Surveillance Conference; 2009. 3-4 December.
54. Dupuy C, Morignat E, Maugey X, Vinard J-L, Hendrikx P, Ducrot C, Calavas D, Gay
E: Defining syndromes using cattle meat inspection data for syndromic surveillance
purposes: a statistical approach with the 2005-2010 data from ten French
slaughterhouses. BMC Vet Res 9:88.
55. Canadian Primary Care Sentinel Surveillance Network
[http://www.cpcssn.ca/cpcssn/home-e.asp]
56. Vangeel ID, Vangeel I, De Leeuw I, Meroc E, Vandenbussche F, Riocreux F,
Hooyberghs J, Raemaekers M, Houdart P, Van dS, De Clercq K: Bluetongue sentinel
surveillance program and cross-sectional serological survey in cattle in Belgium in
2010-2011. Prev Vet Med 106:235-243.
57. Kulldorff M, Hjalmars U: The Knox Method and Other Tests for Space-Time
Interaction. Biometrics 1999, 55:544-552.
58. Knox E: The detection of space-time interactions. Appl Stat 1964, 13:25-29.
59. Mantel N: The detection of disease clustering and a generalized regression
approach. Cancer Res 1967, 27:209-220.
60. Diggle P, Chetwynd A, Haggkvist R, Morris S: Second-order analysis of space-time
clustering. Stat Methods Med Res 1995, 4:124-136.
61. Rogerson P, Yamada I. Statistical detection and surveillance of geographic
clusters. J Reg Sci 51:195.
62. Pearl DL, Louie M, Chui L, Dore K, Grimsrud KM, Leedell D, Martin SW, Michel P,
Svenson LW, McEwen SA: The use of outbreak information in the interpretation of
clustering of reported cases of Escherichia coli 0157 in space and time in Alberta,
Canada, 2000 - 2002. Epidemiol Infect 2006, 134:699-711.
26
63. Kulldorff M, Athas W, Feuer E, Miller B, Key C: Evaluating cluster alarms: A
space-time scan statistic and brain cancer in Los Alamos. Am J Public Health 1998,
88:1377-1380.
64. Perez AM, Zeng D, Tseng C, Chen H, Whedbee Z, Paton D, Thurmond MC: A web-
based system for near real-time surveillance and space-time cluster analysis of foot-
and-mouth disease and other animal diseases. Prev Vet Med 2009, 91:39-45.
65. Naus J: The distribution of the size of maximum cluster of points on a line. J Am
Stat Assoc 1965, 60:532-538.
66. Kulldorff M: A spatial scan statistic. Commun Statist - Theory Meth 1997, 26:1481-
1496.
67. McCulloch CE, Searle SR: Generalized Linear Mixed Models (GLMMs). In
Generalized, Linear, and Mixed Models. Edited by Anonymous New Jersey, USA: John
Wiley & Sons; 2001:220-246.
68. Kleinman K, Lazarus R, Platt R: A generalized linear mixed model approach for
detecting incident clusters of diease in small areas, with an application to biological
terrorism. Am J Epidemiol 2004, 159:217-224.
27
CHAPTER TWO:
FACTORS ASSOCIATED WITH WHOLE CARCASS CONDEMNATION
RATES IN PROVINCIALLY-INSPECTED ABATTOIRS IN ONTARIO 2001 –
2007: IMPLICATIONS FOR FOOD ANIMAL SYNDROMIC SURVEILLANCE
(Alton et al. as published in BMC Veterinary Research 2010; 6:42)
ABSTRACT
Ontario provincial abattoirs have the potential to be important sources of
syndromic surveillance data for emerging diseases of concern to animal health, public
health and food safety. The objectives of this study were to: (1) describe provincially
inspected abattoirs processing cattle in Ontario in terms of the number of abattoirs, the
number of weeks abattoirs process cattle, geographical distribution, types of whole
carcass condemnations reported, and the distance animals are shipped for slaughter; and
(2) identify various seasonal, secular, disease and non-disease factors that might bias the
results of quantitative methods, such as cluster detection methods, used for food animal
syndromic surveillance.
Data were collected from the Ontario Ministry of Agriculture, Food and Rural
Affairs and the Ontario Cattlemen’s Association regarding whole carcass condemnation
rates for cattle animal classes, abattoir compliance ratings, and the monthly sales-yard
price for various cattle classes from 2001-2007. To analyze the association between
condemnation rates and potential explanatory variables including abattoir characteristics,
season, year and commodity price, as well as animal class, negative binomial regression
models were fit using generalized estimating equations (GEE) to account for
autocorrelation among observations from the same abattoir. Results of the fitted model
28
found animal class, year, season, price, and audit rating are associated with condemnation
rates in Ontario abattoirs. In addition, a subset of data was used to estimate the average
distance cattle are shipped to Ontario provincial abattoirs. The median distance from the
farm to the abattoir was approximately 82 km, and 75% of cattle were shipped less than
100 km.
The results suggest that secular and seasonal trends, as well as some non-disease
factors will need to be corrected for when applying quantitative methods for syndromic
surveillance involving these data. This study also demonstrated that animals shipped to
Ontario provincial abattoirs come from relatively local farms, which is important when
considering the use of spatial surveillance methods for these data.
29
INTRODUCTION
The monitoring and surveillance of emerging infectious and zoonotic diseases in
food animals are important components of our food safety system. In recent years,
emerging zoonotic diseases have been of increased concern to both public and animal
health, following the emergence of H5N1 influenza and bovine spongiform
encephalopathy (BSE) [1]. Consequently, researchers are turning their attention to novel
approaches, such as syndromic surveillance, for detecting emerging diseases in food
animals at various points along the farm-to-fork continuum [2, 3].
Abattoirs have played an important role in the surveillance of various diseases of
human and animal health importance [4-6]. Surveillance at the abattoir allows for all
animals passing into the human food chain to be examined for unusual signs, lesions or
specific diseases. For instance, a study evaluating surveillance systems for bovine
tuberculosis in Switzerland found that surveillance during meat inspection at the
slaughterhouse had the highest sensitivity for identifying the disease compared to passive
clinical surveillance of humans or cattle on farms [5]. A relatively new application of
surveillance methods for animal health, food safety and public health is syndromic
surveillance of food animals.
Syndromic surveillance is the grouping of large numbers of signs/symptoms and
data regarding non-traditional sources of information. These groups of signs/symptoms
are loosely designated as ‘syndromes’. This information is then used to track disease
trends in a population and signal putative outbreaks that warrant further investigation [7].
Syndromic surveillance has been primarily used in public health practice [8-10] and has
had some success at the early detection of disease outbreaks in humans [11]. Recently,
30
syndromic surveillance has been applied to animal health using data from farms, sales-
yards, veterinary practitioners, and abattoir condemnation data [2, 3, 12]. Changes in the
incidence of lesions at slaughter may provide important information for syndromic
surveillance of diseases of animal, public health, and food safety significance.
The application of appropriate quantitative methods is important for any
surveillance system. A variety of statistical methods has been developed and is frequently
used for disease surveillance, including spatial, temporal and spatiotemporal methods [7].
In general, these methods use various statistical detection algorithms to analyze a
continuous stream of data and raise an alarm when the count is significantly greater than
expected, suggesting a possible disease outbreak [13]. Depending on the method used and
the data collected, one might be able to identify the area and/or time of the disease
outbreak. Although, these methods have been shown to be useful, surveillance systems
are only as good as the data provided to the system. Consideration of the quality of the
data and naturally occurring covariates need to be taken into account in the selection,
application and interpretation of quantitative methods for disease surveillance.
Recent literature has suggested the need for model-based approaches for
surveillance in order to include other variables into the specification of expected disease
incidence [14]. For many diseases, the incidence may vary with biological factors such as
sex and season. In addition, factors associated with the reporting of disease may also
impact the apparent incidence of disease. For example, price of the commodity may be
associated with the quality of animals being shipped to slaughter, which then in turn will
affect the condemnation rate in abattoirs. Being able to account for these known factors
prior to the application of cluster detection methods, may improve the sensitivity and
31
specificity of a quantitative syndromic surveillance system [15, 16]. The quality of spatial
data is also important to consider prior to the application of space or space-time cluster
detection methods [17]. These issues may be particularly true for Ontario provincial
abattoir data where factors, such as the capacity of an abattoir may be correlated with the
quality of animals they receive, and the location of the abattoir can only approximately
reflect the spatial location of an animal’s farm of origin.
It was hypothesized that a variety of factors may be associated with the
condemnation rates seen in abattoirs, and that these effects may vary in space and/or
time. Abattoir characteristics, such as the number of weeks abattoirs processed animals
and the number of animals processed, may be associated with condemnation rates, as the
speed of processing may impact inspection. An abattoir’s audit rating may reflect a
plant’s compliance to regulations and/or willingness to accept animals of poorer quality.
Region of the abattoir may be associated with condemnation rates in provincial abattoirs
due to regional differences in animal density and disease prevalence. Season and year
may also be associated with condemnation rates as many diseases in animals are thought
to have seasonal and secular variability. Animal class may be associated with
condemnation rates, as older animals are generally at higher risk of disease. Economic
factors, such as commodity price fluctuate greatly and may be associated with the quality
of animals being sent to slaughter, as producers consider shipping costs against the
possible return of shipping an animal of suspect health.
Consequently, the objectives of this study were to identify biological and non-
biological factors that may be associated with abattoir condemnations and possibly
influence cluster detection methods for quantitative syndromic surveillance systems.
32
Specifically, provincially inspected abattoirs which slaughter cattle in Ontario were
characterized in terms of number of abattoirs, the number of weeks abattoirs processed
cattle, geographical distribution, types of condemnations reported, and distance animals
are shipped to provincial abattoirs. Secondly, this study will determine how abattoir
characteristics, season, year and commodity price, and animal class may be associated
with whole carcass condemnation rates in provincial abattoirs. In addition, the suitability
of Ontario provincial abattoir data for spatial and spatio-temporal analyses will be
considered based on the results.
METHODS
Data source and variables
Whole carcass condemnation data were obtained from the Food Safety
Decision Support System (FSDSS) database maintained by the Ontario Ministry of
Agriculture, Food and Rural Affairs (OMAFRA). Data were extracted from the database
for cattle animal classes: bulls, calves, cows, heifers and steers from January 1, 2001 to
December 31, 2007. Missing geographical coordinates for abattoirs were approximated
using postal codes and/or addresses with the address geocoding software GeoPinpoint
Suite 6.4 (DMTI Spatial Inc., Markham, Ontario, Canada). Using the FSDSS database,
the following information was extracted for each month: abattoir identification number,
geographical coordinates of abattoir, year, season, number of weeks an abattoir was open
each year, total number of whole carcasses condemned, total number of cattle processed
each year, and animal class. Season was categorized by 3 month groupings as follows:
winter (December – February), spring (March – May), summer (June – August) and fall
(September – November). Animal class included five categories: bulls, cows, calves,
33
heifers, and steers. Bulls were excluded from subsequent statistical analyses due to
missing data and inconsistencies in the use of this classification. The number of weeks an
abattoir was open each year was determined by the total number of weeks in which at
least one bovine animal was processed. The total number of animals processed each year
was calculated by adding the total number of condemned cattle and the total number of
cattle fit for consumption each year. Linearity of continuous variables was assessed by
plotting the log of the condemnation rate against the covariate using a lowess smoother.
If there was no visible linear relationship between the outcome and the covariate, and the
association could not be adequately modeled with a quadratic term, or transformation,
then the variable was categorized.
Abattoir audit ratings were obtained for all abattoirs from the abattoir audit
program administered through OMAFRA. The audit program assesses each facility’s
food safety performance and compliance with the Ontario Meat Inspection Act.
Audits are conducted once a year and evaluate each premise on 14 food safety areas
based on the Standards of Compliance relating to food safety, animal welfare and
occupational health and safety with a letter grade given for each abattoir [18]. Annual
OMAFRA audit ratings were obtained for all abattoirs in the audit program from 2001-
2007. Abattoir audit ratings were classified according to the letter grade received from
best to poorest as follows: AAA, AA, A, B or C and unrated for abattoirs that had
missing data.
The price of cattle was obtained from the Ontario Cattlemen’s Association market
reports for 2001-2007. Prices were calculated to be the average price (in Canadian
dollars) per lbs based on sales records from Ontario sales-yards. A price was assigned for
34
each month and year by animal class. The most appropriate weight category was selected
to represent each animal class based on an average animal at the time of slaughter.
The agricultural region where the abattoir was located was classified as: central,
eastern, northern, southern or western Ontario using the Census Agricultural Region
boundaries (Statistics Canada, Census Agricultural Regions, Census year 2001). The
regional location of each abattoir was determined using the point-in-polygon technique
with geographic information system software ArcGIS 9.2 (ESRI, Redlands, California,
USA).
Travel distance between the animals’ farm and the abattoir was estimated using
data obtained from OMAFRA. Because farm location is not routinely recorded with
condemnation data, a subset of cattle, in which a sample was sent for laboratory testing,
were used to obtain geo-location information for the abattoir and farm. Like abattoir
location, owner address information was geo-coded according to the owner postal code
using geocoding software GeoPinpoint Canada. Distance from the farm to abattoir was
calculated using the Haversine distance formula, which calculates the great-circle
distances between two points on a sphere using their longitudes and latitudes [19].
Data from all sources were merged into one master dataset using Stata 10.1 (Stata
Corp., College Station, Texas, USA).
Statistical analysis
To model and evaluate their association with monthly whole carcass cattle
condemnation rates, the effect of year, season, annual audit rating, number of weeks
open, number of cattle processed, census agricultural region, animal class and sales price
of animal class were included in the model. All covariates were evaluated for statistical
significance individually and then in a multivariable model using generalized estimating
35
equations (GEE) to fit a negative binomial regression model with an exchangeable
correlation structure to account for repeated measurements from each abattoir. Wald tests
were performed on each covariate in the model to estimate the significance of each
categorical variable as a group. Non-significant covariates (p ≥ 0.05) based on the Wald
test were removed from the model. All excluded covariates were evaluated for their
potential confounding effect by evaluating if their removal produced a 20% or greater
change in the coefficient of significant variables in the model. Interactions between price
and year, year and number of animals processed, year and animal class, as well as season
and animal class were investigated. In addition, the covariates included in the model were
then fitted using GEE with both Poisson and negative binomial distributions and the
following correlation structures: exchangeable, first order autoregressive structure,
second order autoregressive structure, non-stationary, and stationary. All resulting models
were evaluated for how well the model fit the data using a quasi-log-likelihood under the
independence model information criterion (QIC) statistic for model selection [20]. The
model with the lowest QIC was selected as the final model. Robust standard errors were
used for all GEE fitted models. All statistical analyses were performed using Stata 10.1.
RESULTS
Descriptive statistics
There were 207 provincially-inspected abattoirs processing a total of 1,162,410
cattle from 2001 – 2007 with the following animal class distribution: 5.4% bulls, 12.6%
cows, 19.8% heifers, 30.8% calves and 31.3 % steers. The number of abattoirs processing
at least one bovine animal each year varied over the study period (Table 2.1). Of the total
number of processed cattle, 6875 carcasses were condemned for various reasons with
septicaemia and/or toxaemia being typically the most common reason for condemnation
36
(Table 2.2). The condemnation rate per 1000 animals fluctuated over the study period
with the most prominent decrease occurring during 2004-2005 (Figure 2.1). Average
overall condemnation rates were much greater in cows compared to other cattle classes
(Figure 2.2). However, the overall decreasing trend in condemnation rates was most
marked in cows in 2006 (Figure 2.2). The total number of animals being processed
peaked in cows, heifer and steers at some point during the 2004-2006 period (Figure 2.2).
The quartiles of total number of animals and the corresponding number of
processing abattoirs were tabulated for each year of the study period (Table 2.3a). With
the exception of 2004, 2005 and 2007, most of the abattoirs processed fewer than 500
cattle per year. The quartiles of the total number of weeks each year an abattoir processed
at least one animal and the corresponding number of abattoirs was tabulated for each year
of the study period (Table 2.3b). Over the study period, there was an increasing trend in
the number of abattoirs processing cattle more than 49 weeks per year. On average,
approximately 1.5% of abattoirs processed cattle only 1 week per year, 7% processed
cattle a quarter of the year or less, and 20% processed cattle up to half of the year. During
the study period only 19% of abattoirs processed cattle 52 weeks per year. The annual
OMAFRA audit rating scores and the corresponding number of abattoirs receiving those
scores are shown in Table 2.3c for each year during the study period. Throughout this
period, the majority of rated abattoirs were given an “A” rating. The median sales price
of each animal class was calculated for each year during the study period (Table 2.3d).
The median sales-price in all cattle classes was lowest in 2004. No continuous variables
were found to have a linear relationship with cattle carcass condemnation rates, therefore,
quartiles of the empirical distribution were used to categorize the total number of animals
37
processed and the number of weeks an abattoir was open (Table 2.3a and Table 2.3b,
respectively). Price was categorized into a dichotomous variable according to the yearly
median sales price for each animal class (Table 2.3d).
Provincial abattoirs are located throughout Ontario with the majority situated in
southern and western Ontario and fewest located in northern Ontario (Figure 2.3). The
distance between abattoir and farm was calculated for 2456 samples sent for laboratory
testing from 107 of the 207 abattoirs processing cattle from 2001-2007. Results indicated
that the median distance between the farm and abattoir was 82 km, with 25% of all farms
located within 34 km of the abattoir, and 75% within 94 km of the abattoir.
Statistical models
Results of the univariable GEE modeling approach indicated that animal class
(χ2 = 147.39, p < 0.001), region (χ2 = 12.83, p = 0.012), audit rating (χ2 = 351.23, p
<0.001), season (χ2 = 10.41, p = 0.015), and price (χ2 = 4.05, p = 0.044) all had
statistically significant associations with the outcome according to the Wald test
performed to determine the significance of the entire variable in the model (Table 2.4).
Year, the number of weeks an abattoir was open, and the total number of animals
processed were not significant based on Wald tests (χ2 = 11.82, p = 0.066; χ2 = 3.95, p =
0.267; and χ2 = 1.21, p = 0.751, respectively).
Multivariable Poisson and negative binomial models fit by GEE were investigated
using a variety of correlation structures (Table 2.5). Based on the QIC statistic, the best
fitting model was a multivariable negative binomial regression model using an
exchangeable correlation structure. Animal class, year, season, price and audit rating of
abattoirs were the only statistically significant variables in the model. The interaction
38
between animal class and year was the only statistically significant interaction term.
There was no evidence that excluded variables confounded the remaining variables. The
fitted model indicated that during 2005 - 2006 cows had the most prominent decrease in
condemnation rates in abattoirs compared to calves in 2001, based on the size of the cows
x 2005 and cows x 2006 interaction terms (Table 2.6). Condemnation rates for cows
ranged from approximately 3 to 8 times greater than calves throughout the study period
(Table 2.7). Condemnation rates were significantly lower in heifers during 2002
compared to calves in 2001 (Table 2.6). In comparison to winter, condemnation rates
were significantly lower in the summer and fall (Table 2.6). Condemnation rates were
also significantly higher in C rated abattoirs compared to higher rated abattoirs.
Condemnation rates were higher in cattle when the sales price of the animal class was
above the yearly median (Table 2.6).
DISCUSSION
Provincial abattoir data are useful for surveillance because they can provide a
more regionally specific picture of emerging diseases in Ontario. However, various
biological and non-biological factors were found to have an effect on condemnation rates.
Consequently, careful consideration should be given to how these factors may influence
quantitative methods designed for outbreak detection. If these variables are ignored,
quantitative methods designed to identify trends or disease clusters may not provide valid
results.
Results from this study regarding the distance animals are shipped to slaughter
indicated that the majority of cattle are shipped less than 100 km. Within the spatial scale
of the province of Ontario, which is approximately one million square kilometres [21],
39
we can conclude that cattle sent to provincial abattoirs come from relatively local farms.
This result is important for the application of quantitative spatial surveillance methods, as
the assumption that abattoir data reflects disease rates among locally slaughtered animals
appears to be valid.
Seasonal effects were noted in the results of the multivariable analysis with
summer and fall having lower condemnation rates compared to winter. This may be due
to a change in the quality of animals being submitted over the year. Animals which are
shipped during the winter may reflect those animals which grew slower due to certain
health issues causing delayed market readiness, thus, resulting in more condemnations at
slaughter during this season. More research on quality of animals and point in production
cycle is needed to confirm if this trend reflects “poor-doers”. It is important to identify
seasonal trends before the application of quantitative surveillance methods, as many
diseases have seasonal variability, and without correcting for this trend, any results from
temporal or spatial-temporal quantitative methods will be biased.
Commodity class and year were found to be associated with condemnation rates
in provincial abattoirs. It is suspected that the discovery of BSE in Alberta, Canada in
May 2003 [22] had an impact on the patterns of condemnations in Ontario provincial
abattoirs. Within hours of the confirmation of the first Canadian case, the United States
(US) government announced an immediate ban of all imports of Canadian beef. On July
24, 2003, new processing regulations were implemented in Canadian abattoirs outlining
that specified risk materials (SRMs), such as brain and spinal cord, must be removed
from cattle older than 30 months. After 26 months, the US ban on Canadian cattle
imports was lifted [23]. The effects of BSE in Canada, and subsequent changes to trade
40
and processing regulations are mirrored in the descriptive data, which showed a
decreasing trend over the study period in the number of abattoirs processing cattle, as
well as overall condemnation rates and condemnation rates by animal class. The
decreasing trend levels off in 2004 – 2006 where rates begin to increase again.
Condemnation rates in cows showed the largest drop, coinciding with regulations
implemented for cattle over 30 months [22]. These trends are also mirrored in the
interaction effect between year and animal class seen in the multivariable model.
Condemnation rates for cows ranged from approximately 3 to 8 times greater than calves
throughout the study period, with cows during 2005 – 2006 having the most prominent
decrease in condemnation rates in abattoirs compared to calves in 2001. Being older,
cows are at higher risk for disease and thus are condemned more frequently. The decline
in cow condemnation rates during this period occurred while the total number of cows
being processed by provincial abattoirs increased. In the case of cows, the decline in the
rate of condemnations most likely resulted from younger and healthier cows being
shipped for slaughter to provincial abattoirs, which ship their products intra-provincially,
rather than federal plants, which ship their products inter-provincially and internationally.
Commodity class is an important factor, which must be accounted for in a food animal
syndromic surveillance system. Older animals are not only more likely to be condemned
due to an increased incidence of disease, but are more likely to be processed in certain
abattoirs that specialize in this animal class. Consequently, accounting for animal class is
important in determining the expected or baseline condemnation rates at an abattoir.
Economic factors, such as commodity sales price, also appear to play a role in
condemnation rates, as we found that condemnation rates in cattle were higher when the
41
sales price was above the yearly median. This may be due to a difference in the quality of
animals being sent to slaughter depending on the sales price. If the price were too low,
the cost of shipping animals of questionable quality may have exceeded the potential
return for the producer. There was also an association found between condemnation rates
and audit rating. Condemnation rates tended to be higher in C-rated abattoirs compared to
those abattoirs with higher ratings. This may be due to certain abattoirs accepting a larger
proportion of older or poorer quality cattle; however, there were only a small number of
C-rated abattoirs during the study period. The unrated abattoirs were unlikely to have
influenced the model since the condemnation rates for abattoirs in this category were not
significantly different from abattoirs with other ratings. These results demonstrate the
need to consider not only biological factors, but also non-biological factors, which may
be associated with condemnation rates.
This study did not find a statistical association between condemnation rates and
factors dealing with abattoir throughput. However, it is also important to consider abattoir
characteristics, such as the number of weeks and/or the number of animals processed at
abattoirs. There was great variability in the number of weeks and the number of animals
processed in Ontario provincial abattoirs, which may affect quantitative data analysis. For
instance, a spatio-temporal surveillance method such as the spatio-temporal scan statistic
using a space-time permutation model assumes the background population remains
relatively stable over time [24]. This is not the case in Ontario provincial abattoir data, as
only 19% of abattoirs consistently process cattle throughout the year. If the background
population increases or decreases faster in certain areas, there is a risk of population shift
bias, which can cause biased p-values [24]. This bias may cause abattoirs to be identified
42
in a high rate cluster solely due to the irregular timing of processing at certain abattoirs,
not due to a true disease outbreak. It may be more appropriate to select sentinel abattoirs
from regions throughout Ontario, which have a more consistent and stable processing
capacity throughout the year to more accurately capture disease clusters among abattoirs.
It is important to understand the factors affecting and even biasing abattoir
surveillance data, before any quantitative methods can be chosen in the design of a
surveillance system. This study identified and discussed the implications of biological
and non-biological factors, which may affect quantitative surveillance methods. Once
these factors are identified, appropriate adjustments can be made to the quantitative
methods being used for outbreak detection. For instance, a study by Kleinman et al. [15]
compared the performance of the space-time scan statistic using unadjusted data for
lower respiratory complaints and model-adjusted data for day of week, month, holidays
and local history of illness. The study found significantly lower false detection rates in
the model-adjusted analysis compared to the unadjusted.
CONCLUSIONS
This study has identified various seasonal, secular, biological and non-biological
factors that may be associated with the expected incidence of disease and indicates the
potential importance of adjusting for these factors when applying quantitative methods
for any disease surveillance system. Specifically, this study found that animal class, year,
season, price, and audit rating impact condemnation rates from provincially inspected
cattle abattoirs in Ontario. Similarly, the quality of spatial data should also be part of the
assessment of potential biases that may arise from surveillance data prior to decisions
concerning the application of spatial and/or spatial-temporal surveillance techniques.
43
REFERENCES
1. Brown C. Emerging zoonoses and pathogens of public health significance - an
overview. Rev sci tech Off int Epiz 2004; 23:435-442.
2. Vourc'h G, Bridges VE, Gibbens J, De Groot BD, McIntyre L, Poland R, Barnouin J.
Detecting emerging diseases in farm animals through clinical observations. Emerg
Infect Dis 2006; 12:204-210.
3. Van Metre DC, Barkey DQ, Salman MD, Morley PS. Development of a
syndromic surveillance system for detection of disease among livestock entering
an auction market. J Am Vet Med Assoc 2009; 234:658-664.
4. Vilas VJDR, Bohning D, Kuhnert R. A comparison of the active surveillance of
scrapie in the European Union. Vet Res 2008; 39:37-52.
5. Hadorn DC, Stark KDC. Evaluation and optimization of surveillance systems
for rare and emerging infectious diseases. Vet Res 2008; 39:57-69.
6. Karama M, Johnson RP, Holtslander R, McEwen SA, Gyles CL. Prevalence and
characterization of verotoxin-producing Escherichia coli (VTEC) in cattle from
an Ontario abattoir. Can J Vet Res 2008; 72:297-302.
7. Lawson AB, Kleinman K. Spatial and syndromic surveillance for public health.
Chichester,West Sussex; Hoboken,NJ: J. Wiley; 2005.
8. Edge VL, Pollari F, Ng LK, Michel P, McEwen SA, Wilson JB, Jerrett M, Sockett
PN, Martin SW: Syndromic surveillance of Norovirus using over-the-counter sales
of medications related to gastrointestinal illness. Can J Infect Dis Med Microbiol
2006; 17:235-241.
9. Hope K, Durrheim DN, Muscatello D, Merritt T, Zheng W, Massey P, Cashman P,
Eastwood K: Identifying pneumonia outbreaks of public health importance: can
emergency department data assist in earlier identification? Aust N Z J Public
Health 2008; 32:361-364.
10. van Dijk A, McGuinness D, Rolland E, Moore KM: Can Telehealth Ontario
respiratory call volume be used as a proxy for emergency department
respiratory visit surveillance by public health? CJEM 2008; 10:18-24.
11. Wagner MM, Moore AW, Aryel RM. Handbook of Biosurveillance. Amsterdam;
Boston: Academic Press; 2006.
12. Weber WD: Development of an animal health monitroing system based on
slaughter condemnation data. In: Proceedings of the first International Society for
Disease Surveillance Conference: 3-4 December 2009; Miami.
44
13. Stoto MA, Fricker RDJ, Jain A, Diamond A, Davies-Cole JO, Glymph C, Kidane G,
Lum G, Jones L, Dehan K, Yuan C. Evaluating Statistical Methods for
Syndromic Surveillance. In: Statistical Methods in Counterterrorism. Edited by
Wilson AG, Wilson GD, Olwell DH. New York: Springer; 2006, pp 141-172.
14. Robertson C, Nelson TA, MacNab YC, Lawson AB: Review of methods for
space-time disease surveillance. Spat Spattemporal Epidemiol 2010; 1:105-116.
15. Kleinman K, Abrams AM, Kulldorff M, Platt R: A model-adjusted space-time
scan statistic with an application to syndromic surveillance. Epidemiol Infect
2005; 133:409-419.
16. Klassen AC, Kulldorff M, Curriero F: Geographical clustering of prostate
cancer grade and stage at diagnosis, before and after adjustment for risk factors. Int J Health Geogr 2005; 4:1.
17. Teutsch SM, Churchill RE: Principles and practice of public health surveillance:
2nd Ed. Oxford; New York: Oxford University Press; 2000.
18. Audit Ratings of Provincially Licensed Meat Plants [Available from
http://www.omafra.gov.on.ca/english/food/inspection/meatinsp/abattoir_audit_ratings.ht
m]
19. Gottwald S, Hirsch KA, Reichardt H: VNR Concise Encyclopedia of Mathematics:
2nd ed. New York: Van Nostrand Reinhold; 1989.
20. Pan W: Akaike's information criterion in generalized estimating equations.
Biometrics 2001; 57:120-125.
21. Land and freshwater area, by province and territory
[http://www40.statcan.gc.ca/101/cst01/phys01-eng.htm]
22. Bovine Spongiform Encephalopathy (BSE) in North America
[http://www.inspection.gc.ca/english/anima/heasan/disemala/bseesb/bseesbindexe.
Shtml]
23. Timeline of BSE in Canada and the U.S.
[http://www.cbc.ca/news/background/madcow/timeline.html]
24. Kulldorff M, Hjalmars U: The Knox Method and other tests for space-time
interaction. Biometrics 1999; 55:544-552.
45
Table 2.1- Number of provincially inspected abattoirs in Ontario 2001-2007
Year Number of abattoirs reporting bovine carcass counts
2001 163
2002 158
2003 147
2004 143
2005 148
2006 139
2007 129
Total 207
Number of provincially inspected abattoirs in Ontario processing at least one bovine
animal per year and the total number of provincially inspected abattoirs processing cattle
in Ontario from 2001-2007.
46
Table 2.2 - Reason for whole carcass condemnation in provincially inspected
abattoirs in Ontario 2001-2007
Condemnation reason Percentage of total yearly condemnations
Year
2001 2002 2003 2004 2005 2006 2007
Abscess 7.4 9.4 7.3 8.8 9.8 10.5 9.8
Animal arrived dead1 0 0 2.7 0.4 1.1 0.8 0.6
Animal found dead2 4.7 5.8 4.9 4.4 7.5 7.2 6.8
Arthritis 3.5 2.4 2.9 2.9 1.9 4.0 3.2
Bruising 9.1 9.8 7.8 3.0 2.0 1.4 1.1
Chemical Residue 0 0.1 0.0 0.0 0.0 0.0 0.0
Contamination 0.1 0.1 0.2 0.1 0.2 0.2 0.3
Cystericercus bovis 0.0 0 0.0 0.2 0.0 0.0 0.0
Drug residue 3.0 2.0 1.5 0.0 0.1 0.3 3.1
Edema 2.3 3.7 3.4 3.4 1.3 1.2 0.8
Emaciation 6.6 6.0 7.7 6.6 10.7 8.4 9.7
Icterus 0.6 1.2 1.5 1.6 1.7 1.2 1.5
Inadequate bleeding 0.4 0.1 0.9 3.0 1.5 0.5 0.8
Lymphadenitis 1.5 2.2 5.3 6.7 4.5 4.6 3.4
Lymphosarcoma neoplasm 12.6 13.1 12.4 6.6 4.4 2.7 4.3
Mastitis 0.0 0.1 0.2 0.2 0.6 0.2 0.3
Metritis 0.4 0.1 0.1 0.4 0.7 0.2 0.3
Moribund 1.4 1.3 0.7 1.0 1.1 0.9 1.8
Myositis 0.7 0.2 0.2 0.2 0.6 1.1 0.8
Neoplasm other 3.0 1.9 2.7 1.8 2.1 3.0 2.5
Nephritis 0.8 0.7 0.5 0.3 1.2 0.9 0.5
Odour 2.2 1.2 1.5 0.8 0.3 0.2 0.2
47
Operator requested
condemnation
0.0 0.0 0.0 4.5 10.6 7.9 8.9
Other disease 6.9 6.2 5.3 8.7 5.1 4.9 3.4
Pericarditis 2.0 1.7 1.1 1.3 1.9 2.1 1.7
Peritonitis 6.5 6.2 6.0 6.1 6.1 7.6 6.9
Pleuritis 0.4 0.8 0.6 0.5 1.3 1.4 0.6
Pneumonia 2.7 3.1 2.2 4.1 5.5 4.6 4.8
Pyelonephritis 0.3 0.2 0.9 0.0 1.2 1.4 2.5
Rabies 0.0 0.1 0.0 0.0 0.0 0.0 0.0
Septicaemia and/or toxemia 11.7 15.6 15.0 17.7 11.6 14.0 15.8
Squamous cell carcinoma
neoplasm
1.2 0.4 0.9 1.3 0.8 1.1 1.1
Toxaemia 6.0 3.2 2.5 2.3 0.7 4.4 0.5
Uremia 2.3 1.2 1.5 1.3 1.8 1.4 2.2
Total number of condemnations
per year
1047 1203 1236 1188 895 656 650
1 Refers to animals which are dead when the truck pulls into the plant premises
2 Inspector finds a dead animal in the pen that had arrived alive during ante-mortem
inspection in the barn
Reason and percentage of whole cattle carcasses condemned in provincially inspected
abattoirs in Ontario from 2001-2007.
48
Table 2.3 - Summary of number of cattle processed, number weeks open, audit
rating and animal class
a) Number of
animals processed
Number of Abattoirs
2001 2002 2003 2004 2005 2006 2007
1 - 286 73 72 36 24 29 40 38
287 - 498 42 35 40 28 34 34 25
499 - 841 25 24 36 40 43 32 36
842 - 27 286 22 26 33 49 41 32 30
b) Number of
weeks open
Number of Abattoirs
1 - 43 66 67 44 32 43 38 38
44 - 49 58 46 48 45 30 39 29
50 - 51 17 27 25 31 41 31 32
52 21 17 28 33 33 30 30
c) Audit Rating Number of Abattoirs
AAA 1 0 0 0 0 1 1
AA 6 6 7 13 16 23 28
A 61 72 79 78 78 78 72
B 15 8 8 5 8 6 9
C 2 1 0 0 1 0 0
Unrated 78 71 53 47 45 31 19
d) Animal Class Median sales-price of animal class per year 2001-2007
Calves 136.53 117.95 112.19 83.61 107.14 119.69 108.49
Cows 64.28 58.64 27.11 19.70 25.38 31.84 34.77
Heifers 112.01 101.22 82.16 73.61 87.74 90.87 91.09
Steers 113.24 102.57 83.15 75.60 91.07 93.55 92.56
Summary of the a) quartiles of the total number of cattle processed, b) quartiles of the number of weeks at least one bovine animal was
processed, c) annual OMAFRA audit rating, and d) median sales price of calves, cows, heifers and steers 2001-2007.
49
Table 2.4 - Univariable negative binomial models using GEE1
approach
Covariate IRR2 Robust Standard
Error
P-value 95% CI
Animal class
Calves --- --- --- ---
Cows 4.83 1.35 < 0.001 2.80 - 8.34
Heifers 0.43 0.09 < 0.001 0.28 - 0.64
Steers 0.51 0.11 0.01 0.34 - 0.77
Year
2001 --- --- --- ---
2002 1.02 0.14 0.90 0.77 - 1.34
2003 0.90 0.14 0.49 0.67 - 1.22
2004 0.70 0.15 0.10 0.46 - 1.06
2005 0.69 0.14 0.07 0.47 - 1.03
2006 0.75 0.16 0.17 0.50 - 1.13
2007 0.73 0.17 0.18 0.46 - 1.16
Season
Winter --- --- --- ---
Spring 0.94 0.05 0.20 0.85 - 1.03
Summer 0.86 0.06 0.02 0.76 - 0.98
Fall 0.76 0.07 0.01 0.64 - 0.91
3Audit rating
C --- --- --- ---
AAA 0.71 0.53 0.64 0.16 - 3.07
AA 0.90 0.37 0.80 0.40 - 2.01
A 1.19 0.48 0.67 0.54 - 2.61
B 1.01 0.42 0.98 0.45 - 2.28
50
unrated 2.69 1.09 0.01 1.22 - 5.94
Number of weeks open
1 - 43 --- --- --- ---
44 - 49 0.72 0.15 0.10 0.48 - 1.08
50 - 51 0.71 0.16 0.13 0.46 - 1.11
52 0.92 0.29 0.78 0.49 - 1.70
Number of animal processed
1 - 286 --- --- --- ---
287 - 498 1.01 0.15 0.95 0.75 - 1.35
499 - 841 1.14 0.19 0.44 0.82 - 1.58
842 - 27286 1.29 0.42 0.43 0.68 - 2.46
Region
Central --- --- --- ---
Eastern 9.88 6.83 0.01 2.55 - 38.26
Northern 3.78 2.49 0.04 1.04 - 13.74
Southern 7.91 6.56 0.01 1.56 - 40.21
Western 1.60 2.31 0.75 0.09 - 27.23
Price
Below median --- --- --- ---
Above median 1.11 0.06 0.04 1.00 - 1.23
1 Generalized estimating equation using an exchangeable correlation structure to
accommodate repeated measurements among provincial abattoirs 2
Incidence rate ratio 3
Model would not converge therefore results represent univariable negative binomial
model
Univariable negative binomial models using a GEE1
approach modeling the association
between condemnation rates in Ontario provincial abattoirs and animal class, year,
season, audit rating, the number of weeks an abattoir was open each year, the number of
cattle an abattoir processed each year, census agricultural region and median yearly sales-
price for animal class.
51
Table 2.5 - Comparison of GEE1 fitted models using QIC
2 statistic
Model Correlation Structure QIC
Negative binomial GEE2 Exchangeable 11328
Non-stationary 11626
2
nd order autoregressive 11647
Stationary 11664
1
st order autoregressive 11668
Poisson GEE2 2
nd order autoregressive 12007
Stationary 12056
1
st order autoregressive 12409
Exchangeable 13605
Non-stationary Did not converge
1Generalized estimating equation
2Quasi-log- likelihood under the independence model information criterion
Comparison of QIC2
statistic values for negative binomial or Poisson models using a
GEE1
approach to investigate the association between condemnation rates in Ontario
Provincial abattoirs and year, season, animal class, price, year-animal class interaction
and audit rating.
52
Table 2.6 - Multivariable negative binomial model using a GEE1 approach
Covariate IRR2
Robust Standard
Error P-value 95% CI
Animal class
Calves --- --- --- ---
Cows 6.73 2.40 < 0.001 3.35 - 13.52
Heifers 0.47 0.10 < 0.001 0.31 - 0.72
Steers 0.46 0.12 0.01 0.29 - 0.74
Year
2001 --- --- --- ---
2002 1.37 0.21 0.04 1.02 - 1.84
2003 0.70 0.16 0.11 0.45 - 1.09
2004 0.56 0.13 0.02 0.36 - 0.90
2005 0.93 0.20 0.74 0.61 - 1.43
2006 1.05 0.23 0.80 0.69 - 1.63
2007 1.16 0.29 0.56 0.71 - 1.88
Season
Winter --- --- --- ---
Spring 0.88 0.07 0.07 0.76 - 1.01
Summer 0.82 0.06 0.01 0.71 - 0.95
Fall 0.77 0.05 < 0.001 0.69 - 0.87
Audit rating
C --- --- --- ---
AAA 0.32 0.30 0.22 0.05 - 2.01
AA 0.43 0.12 0.01 0.25 - 0.74
A 0.64 0.11 0.01 0.46 - 0.89
B 0.56 0.11 0.01 0.39 - 0.82
53
unrated 1.41 0.35 0.17 0.87 - 2.31
Price
Below median --- --- --- ---
Above median 1.12 0.05 0.02 1.02 - 1.23
Year x animal class
Calves and 2001 --- --- --- ---
Cows and 2002 0.61 0.12 0.01 0.42 - 0.88
Cows and 2003 1.13 0.34 0.69 0.63 - 2.05
Cows and 2004 0.83 0.30 0.61 0.40 - 1.72
Cows and 2005 0.48 0.14 0.01 0.27 - 0.86
Cows and 2006 0.45 0.12 0.01 0.27 - 0.78
Cows and 2007 0.51 0.19 0.06 0.25 - 1.04
Heifers and 2002 0.61 0.14 0.03 0.39 - 0.95
Heifers and 2003 1.46 0.44 0.21 0.81 - 2.65
Heifers and 2004 1.05 0.30 0.85 0.61 - 1.83
Heifers and 2005 0.88 0.24 0.63 0.51 - 1.50
Heifers and 2006 0.93 0.22 0.77 0.59 - 1.48
Heifers and 2007 0.71 0.23 0.29 0.39 - 1.33
Steers and 2002 0.74 0.14 0.12 0.51 - 1.08
Steers and 2003 1.40 0.43 0.28 0.77 - 2.54
Steers and 2004 1.63 0.50 0.11 0.89 - 2.99
Steers and 2005 1.02 0.28 0.94 0.60 - 1.73
Steers and 2006 1.42 0.44 0.26 0.78 - 2.61
Steers and 2007 0.90 0.32 0.77 0.45 - 1.81
Correlation value = 0.067
1 Generalized estimating equation using an exchangeable correlation structure to account
for repeated measurements among provincial abattoirs
54
2 Incidence rate ratio
Multivariable negative binomial models using a GEE1
approach investigating the
association between condemnation rates in Ontario provincial abattoirs and animal class,
year, season, price, year-animal class interaction and audit rating based on the
multivariable GEE model.
55
Table 2.7 - Linear combinations of condemnation rates in cows and calves
Contrast IRR P-value 95% CI
Cows 2001 vs. Calves 2001 6.73 < 0.001 3.35 - 13.52
Cows 2002 vs. Calves 2002 4.09 < 0.001 2.23 - 7.50
Cows 2003 vs. Calves 2003 7.61 < 0.001 4.00 - 14.46
Cows 2004 vs. Calves 2004 5.56 < 0.001 2.81 - 10.99
Cows 2005 vs. Calves 2005 3.23 < 0.001 1.90 - 5.50
Cows 2006 vs. Calves 2006 3.06 < 0.001 1.86 - 5.01
Cows 2007 vs. Calves 2007 3.41 < 0.001 1.97 - 5.90
Linear combinations of condemnation rates in cows compared to calves in Ontario
Provincial abattoirs 2001-2007 based on model in Table 2.6.
56
Figure 2.1 - Condemnation rates per 1000 cattle from Ontario provincial abattoirs
2001 – 2007
57
Figure 2.2 - Animal class condemnation rates per 1000 cattle from Ontario provincial abattoirs 2001 - 2007
58
Figure 2.3 - Choropleth map of percentage of abattoirs processing cattle in Ontario
per census agricultural region
59
CHAPTER THREE:
SUITABILITY OF BOVINE PORTION CONDEMNATIONS AT
PROVINCIALLY-INSPECTED ABATTOIRS IN ONTARIO CANADA FOR
FOOD ANIMAL SYNDROMIC SURVEILLANCE
(Alton et al. as published in BMC Veterinary Research 2012; 8:88)
ABSTRACT
Abattoir condemnations may play an important role in a food animal syndromic
surveillance system. Portion condemnation data may be particularly useful, as these data
can provide more specific information on health outcomes than whole carcass
condemnation data. Various seasonal, secular, disease, and non-disease factors have been
previously identified to be associated with whole carcass condemnation rates in Ontario
provincial abattoirs; and if ignored, may bias the results of quantitative disease
surveillance methods. The objective of this study was to identify various seasonal,
secular, and abattoir characteristic factors that may be associated with bovine portion
condemnation rates and compare how these variables may differ from previously
identified factors associated with bovine whole carcass condemnation rates.
Data were collected from the Ontario Ministry of Agriculture, Food and Rural
Affairs (OMAFRA) and the Ontario Cattlemen’s Association regarding “parasitic liver”
and pneumonic lung condemnation rates for different cattle classes, abattoir compliance
ratings, and the monthly sales-yard price for commodity classes from 2001-2007. To
control for clustering by abattoirs, multi-level Poisson modeling was used to investigate
the association between the following variables and “parasitic liver” as well as
pneumonic lung condemnation rates: year, season, annual abattoir audit rating,
60
geographic region, annual abattoir operating time, annual total number of animals
processed, animal class, and commodity sales price.
In this study, “parasitic liver” condemnation rates were associated with year,
season, animal class, audit rating, and region. Pneumonic lung condemnation rates were
associated with year, season, animal class, region, audit rating, number of cattle
processed per year, and number of weeks abattoirs processed cattle. Unlike previous
models based on whole carcass condemnations, commodity price was not associated with
partial condemnations in this study. The results identified material-specific predictor
variables for condemnation rates. This is important for syndromic surveillance based on
abattoir data and should be modeled and controlled for during quantitative surveillance
analysis on a portion specific basis.
BACKGROUND
Animal disease outbreaks can have a devastating effect, not only on animals, but
to the food-animal industry, public, economy, and international trade [1-3]; therefore,
research and development of novel animal disease surveillance systems is extremely
important. In recent years, Ontario has experienced the emergence of various infectious
animal diseases including a new strain of porcine circovirus type II (PCV-2) in 2004[4],
outbreaks of bovine viral diarrhea (BVD) with enhanced virulence in cattle in the early
90’s [3], and the impact of the identification of a small number of cases of bovine
spongiform encephalopathy (BSE) in Alberta, Canada in 2003[2]. The confirmation of
BSE in May 2003 is just one of many examples of how the emergence of an infectious
animal disease can cripple an industry and have profound lasting effects [2]. Though all
61
BSE cases in Canada were only identified in Alberta, Canada, closure of international
trade borders to Canadian cattle caused the price of cattle to drop significantly throughout
the country and took several years to recover [5]. New approaches to animal disease
surveillance, such as syndromic surveillance of condemnation data may be important for
the timely identification of infectious animal and zoonotic disease events in the future.
Syndromic surveillance, though an increasingly popular tool in public health
surveillance research [6-9] has only recently been explored as an option for animal health
surveillance. One novel way to target surveillance of infectious animal and zoonotic
diseases would be through the food system. In the past, abattoirs have been the focus for
surveillance at this human-animal interface, usually involving targeted surveillance of a
specific disease [10-13]. In recent years, research has been expanded to include
surveillance of food animal data from a variety of sources including on-farm surveillance
[14,15], sales-yard surveillance [16], as well as syndromic surveillance using abattoir
condemnation data [17-19].
Abattoir condemnation data have the potential to provide early warning of
emerging animal and zoonotic diseases, particularly provincial abattoir data and yet these
data have been under-utilized in the past. Portion condemnations and whole carcass
condemnation data have been previously described in the literature [20-22]; however, few
reports of potential usages of these data for syndromic surveillance purposes have been
implemented. Changes in portion and whole carcass condemnation rates could be
monitored over time and space, and when the condemnation rate reaches a certain
threshold it may signal a potential outbreak or quality control problem within an abattoir
62
and/or region. In Ontario, condemnation data from provincially inspected abattoirs are
particularly useful for syndromic surveillance, as they give a relatively local perspective
on the health of animals within the province. Ontario provincial abattoirs only distribute
their products within the province, compared to federal abattoirs, which ship their
products inter-provincially and internationally [23]. Anecdotal evidence has suggested
that cattle shipped to Ontario provincial abattoirs originate from relatively local farms.
This was confirmed by a previous study which found that 75% of cattle from Ontario
provincial abattoirs originated from farms less than 94 kilometres from the abattoir [17].
Previous research has investigated the use of whole carcass condemnation data for
syndromic surveillance [17]. It is unclear whether organ/body system data may be better
suited for syndromic surveillance, as these data may provide more specific information
on health outcomes than whole carcass condemnation data. By having a more specific
outcome, it is hypothesized that portion condemnation data should be more sensitive than
whole carcass data because inspectors condemn a carcass for one reason. However,
bovine carcasses may have several disease conditions causing the condemnation of the
carcass; whereas organs are less likely to have more than one reason for condemnation
and reflect diseases found in a specific organ system.
Generalized linear mixed models (GLM) have been previously used for human
disease surveillance [24], as well as clustered GLM’s fit by generalized estimating
equations (GEE) specifically for whole carcass condemnation data [17]. Various
seasonal, secular, disease, and non-disease factors were previously found to have a
statistically significant association with bovine whole carcass condemnation rates, and
should be controlled for in the application of quantitative surveillance methods to prevent
63
biased results (e.g., false alarms) [17]. It was unclear whether the same factors would also
be significantly associated with condemnation data when applied to portion
condemnation data.
The objective of this study was to identify various seasonal, secular, and abattoir
characteristic factors that may be associated with bovine portion condemnation rates and
compare how these may differ from factors associated with bovine whole carcass
condemnation rates. “Parasitic liver” and pneumonic lung condemnations were used in
this case study as these condemnation classes were a rich source of data.
METHODS
Data source and variables
Portion condemnation data were obtained from the Food Safety Decision Support
System (FSDSS) database maintained by the Ontario Ministry of Agriculture, Food and
Rural Affairs (OMAFRA). The database contains information regarding the number and
reason for daily organs/body systems condemnations in provincially inspected abattoirs
in Ontario. The condemnation categories of pneumonic lungs and “parasitic livers” were
selected for this analysis, as these categories were among the most frequently reported
portion condemnations by provincial inspectors during the study period and represent
potential animal and public health concern. “Parasitic livers” is an inspection term used to
label bovine livers considered unfit for human consumption, and thus condemned due to
pathologies such as necrosis, fibrosis cirrhosis, atrophy, telangiectasia, and adhesions.
Although the term “parasitic liver” suggests truly parasitic infections such as fascioliasis,
the term covers non-parasitic conditions as well (personal communication Ab Rehmtulla,
64
DVM, OMAFRA, Stone Road, Guelph, Ontario). Pneumonic lung condemnation refers
to bovine lungs which were condemned for lesions indicative of a previous localized and
resolved antero-ventral pneumonia infection. Data were extracted from the database for
cattle animal classes: bulls, calves, cows, heifers, and steers from January 1, 2001 to
December 31, 2007. Missing geographical coordinates for abattoirs were approximated
using postal codes and/or addresses with the address geocoding software GeoPinpoint
Suite 6.4 (DMTI Spatial Inc., Markham, Ontario, Canada). Using the FSDSS database,
the following variables were created for each month: abattoir identification number,
geographical coordinates of abattoir, year, season, number of weeks an abattoir was
operating each year, total number of “parasitic liver” and pneumonic lung
condemnations, total number of cattle processed each year, and animal class. Season was
categorized by 3 month groupings as follows: winter (December - February), spring
(March - May), summer (June - August), and fall (September - November). Animal class
included five categories: bulls, cows, calves, heifers, and steers. Bulls were excluded
from subsequent analyses due to missing data and inconsistencies in the use of this
classification. The number of weeks an abattoir was operating each year was determined
by the total number of weeks in which at least one bovine animal was processed. The
total number of animals processed each year was calculated from the total number of
condemned cattle plus the number of cattle fit for consumption.
Abattoir audit ratings were obtained for all abattoirs through the abattoir audit
program administered through OMAFRA. The audit program assesses each facility’s
food safety performance and compliance with the Ontario Meat Inspection Act. Audits
are conducted once a year and evaluate each premise on 14 food safety areas based on the
65
Standards of Compliance relating to food safety, animal welfare and occupational health
and safety with a letter grade given for each abattoir [25]. Annual OMAFRA audit ratings
were obtained for all abattoirs in the audit program from 2001-2007. Abattoir audit
ratings were classified according to the letter grade received from best to poorest as
follows: AAA, AA, A, B or C and unrated for abattoirs that had missing data.
The price of cattle was obtained from the Ontario Cattlemen’s Association market
reports for 2001-2007. Prices are calculated to be the average price (in Canadian dollars)
per 100 lbs based on sales records from Ontario sales-yards. A price was assigned to each
month and year by animal class. The most appropriate weight category was selected to
represent each animal class based on an average animal at the time of slaughter.
The agricultural region where an abattoir was located was classified as: central,
eastern, northern, southern or western Ontario using the Ontario Census Agricultural
Region boundaries (Statistics Canada, Census Agricultural Regions, Census year 2001).
The regional location of each abattoir was determined using the point-in-polygon
technique with geographic information system software ArcGIS 9.2 (ESRI, Redlands,
California, USA).
Data from all sources were merged into one master dataset using Stata 10.1 (Stata
Corp., College Station, Texas, USA).
Statistical analysis
Multilevel Poisson regression modeling was used to evaluate the association of
monthly condemnation rates of pneumonic lungs and “parasitic livers” with the above
66
mentioned predictor variables. To model and evaluate their association with monthly
“parasitic liver” and pneumonic lung condemnation rates, the effect of year, season,
annual audit rating, number of weeks in operation, number of cattle processed, census
agricultural region, animal class, sales price of animal class were included in the model.
Linearity of continuous variables was assessed by plotting the log of the condemnation
rate per slaughtered cattle for both liver and lung condemnations against the covariate
using a locally weighted regression (lowess smoother) approach. If there was no visible
linear relationship between the outcome and the covariate, and the association could not
be adequately modeled with a quadratic term, or transformation, then the variable was
categorized. All covariates were evaluated for statistical significance individually and
then in a multivariable model using a multilevel Poisson regression model accounting for
clustering of observations within abattoirs. Backward selection was based on Wald tests
and non-significant covariates were removed from the model (α = 0.05). All excluded
covariates were evaluated for their potential confounding effect by evaluating if their
removal produced a 20% or greater change in the coefficient of the remaining variables in
the model. Interactions between region and animal class, year and animal class, as well as
season and animal class were investigated. The covariates included in the model were
fitted using a multilevel Poisson penalized quasi-likelihood model using a 2nd order
Taylor Series approximation (PQL-2). If convergence issues prohibited the multilevel
Poisson model being fit using the PQL-2 algorithm, a 1st order Taylor series
approximation (PQL-1) was employed. The premise’s identification number was used to
account for clustering at the level of the abattoir in both models. Fit of the models were
assessed visually by plotting the upper level residuals (at the level of the abattoir), also
67
known as the best linear unbiased predictors (BLUP’s) against the normal scores to
assess normality, as well as comparing the BLUP’s to the predicted outcome to assess
variance homogeneity. All multi-level statistical analyses were performed using MLwiN
2.17 (Centre for Multilevel Modeling, London, UK).
RESULTS
Descriptive Statistics
There were 211 provincially-inspected abattoirs slaughtering a total of 1,155,535
cattle from 2001-2007 (Figure 3.1). Of the total number of slaughtered cattle, 403,290
organs/portions were condemned for various reasons with the top four reasons being
evidence of pneumonia, liver abscesses, nephritis and “parasitic livers”. There were 14
different organs/body systems investigated for portion condemnations and 13
condemnation reasons (Table 3.1). Overall, the condemnation rate for “parasitic livers”
per 1000 slaughtered cattle increased over the study period, particularly in cows (Figure
3.2). In contrast, the condemnation rate for lungs per 1000 slaughtered cattle decreased
over the study period, especially in calves (Figure 3.3).
The quartiles of total number of animals and the corresponding number of
processing abattoirs were tabulated for each year of the study period (Table 3.2a). With
the exception of 2004 and 2005, most of the abattoirs processed fewer than 500 cattle per
year. The quartiles of the total number of weeks each year an abattoir processed at least
one animal and the corresponding number of abattoirs was tabulated for each year of the
study period (Table 3.2b). Over the study period, there was an increasing trend in the
number of abattoirs processing cattle more than 49 weeks per year. The annual
OMAFRA audit rating scores and the corresponding number of abattoirs receiving those
68
scores are shown in Table 3.2c for each year of the study. Throughout the study period,
the majority of rated abattoirs were given an “A” rating. The median sales price of each
animal class was calculated for each month during the study period (Table 3.2d). The
median sales-prices for all cattle classes were lowest in 2004. No continuous variables
were found to have a linear relationship with cattle carcass condemnation rates, therefore,
quartiles were used to categorize the total number of animals processed and number of
weeks an abattoir was open (Tables 3.2a and 3.2b, respectively). Price was categorized
into a dichotomous variable according to whether the price was less than or equal to the
annual median sales price for each animal class (Table 3.2d).
Statistical models
Pneumonic lung condemnation model
Results of the univariable Poisson regression models (PQL-2) indicated that
animal class (p < 0.01), year (p < 0.01), season (p < 0.01), agricultural region (p = 0.02),
number of weeks abattoirs processed cattle (p < 0.01) all had statistically significant
association with the condemnation rate of lungs according to the Wald test for the
variable. The number of animals processed per year (p < 0.01) was statistically
significant according to the Wald test for the variable using a MQL estimation method, as
the PQL algorithm would not converge. Price was not significantly associated with lung
condemnations (p = 0.79). Due to the convergence issues in the model fitting process the
univariable association between the outcome and audit rating could not be assessed.
Animal class, year, season, agricultural region, audit rating, number of weeks
abattoirs processed cattle, and number of animal processed per year were found to have a
statistically significant association with the outcome in the multivariable model (PQL-2)
69
(Table 3.3). There was no evidence that the excluded variable price confounded these
variables. Statistically significant interactions were found to exist between animal class
and season, animal class and region, as well as, animal class and year. The fitted model
indicated that lung condemnations tended to be lower in higher rated abattoirs compared
to C rated abattoirs (Table 3.3). Condemnation rates tended to be lower in abattoirs
processing a larger number of cattle each year compared to smaller processing abattoirs
(Table 3.3). Lung condemnation rates were higher in abattoirs open throughout the year
compared to abattoirs open fewer weeks during the year (Table 3.3). Due to the
complexity of the interaction terms between animal class and season, agricultural region
and year, relationships among these variables based on the predicted rates from the fitted
model were explored (Figure 3.4). According to predicted lung condemnation rates for
calves from the multilevel model, condemnation rates were highest in eastern, western
and central Ontario regions (Figure 3.4), with the highest condemnation rates found in
calves in eastern Ontario compared to all other regions and animal classes. A decreasing
trend in condemnation rates in calves was also seen in these same regions with the
exception of a small peak in 2003. The same decreasing trend and peak in 2003 was also
evident in heifers and steers in the same regions noted above (Figure 3.4). The same trend
was evident in cows, with the exception of a peak in 2003 and 2004 (Figure 3.4). In
comparison, calves, cows, heifers and steers in northern and southern Ontario regions had
consistently lower lung condemnation rates throughout the entire study period (Figure
3.4). The best linear unbiased predictors (BLUP’s) were visually inspected for the multi-
level model; there was no evidence to reject the assumptions of normally distributed
residuals and homogeneity of variance.
70
“Parasitic liver” condemnation model
Results of the univariable multi-level Poisson regression model (PQL-2) indicated
that animal class (p < 0.01), year (p < 0.01), season (p < 0.01), agricultural region (p <
0.01), audit rating (p < 0.01), the number of weeks an abattoir processed cattle (p < 0.01),
total number of cattle processed per year (p < 0.01) and price (p < 0.01) all had
statistically significant associations with “parasitic liver” condemnation rate according to
the Wald test for the variable.
In the multivariable multilevel Poisson model (PQL-1), year, animal class, season,
agricultural region, and audit rating were found to have a statistically significant
association with the outcome (Table 3.4). There was no evidence that the excluded
variables confounded the variables included in the model. Statistically significant
interactions were found to exist between animal class and season, animal class and
region, as well as animal class and year (Table 3.4). The fitted model indicated that liver
condemnation rates were higher in C-rated abattoirs compared to higher rated abattoirs
(Table 3.4). Due to the complexity of the interaction terms between animal class and
season, agricultural region and year, relationships among these variables based on
predicted rates from the fitted model were explored (Figure 3.5). According to predicted
liver condemnation rates for steers from the multilevel model, condemnation rates were
highest in eastern Ontario. Rates amongst steers tended to be higher in fall and winter.
The lowest rates in steers were found in northern Ontario (Figure 3.5). Liver
condemnation rates for calves were highest in central Ontario and lowest in eastern and
northern Ontario. Similar to steers, rates amongst calves tended to be higher in fall and
winter throughout the study period. Liver condemnation rates in cows had the highest
71
rates compared to all other animal classes in all regions of Ontario throughout the study
period (Figure 3.5). Condemnation rates in cows were highest in central and eastern
Ontario and lowest in southern Ontario. Condemnation rates amongst cows remained
fairly stable throughout the seasons; however, an increasing secular trend was seen in
cows with peaks in 2004 and 2007 for all regions of Ontario. Liver condemnation rates
for heifers were highest in eastern and western Ontario and lowest in northern and
southern Ontario (Figure 3.5). Condemnation rates in heifers remained fairly stable
within a region throughout the study period. The predicted random effects were subjected
to a visual diagnostic analysis and no evidence against the assumption of normality and
variance homogeneity was found.
DISCUSSION
Portion and whole carcass condemnation data may have an important role in the
development of a food animal syndromic surveillance system. These data provide insight
into lesions on carcasses and organs, which may lead to early detection of emerging
animal and zoonotic diseases. This study builds upon previous research investigating
biological and non-biological factors associated with bovine whole carcass condemnation
rates in Ontario provincial abattoirs during the same study period [17]. As with whole
carcass data, various seasonal, secular and abattoir characteristic factors were found to
have an association with liver and lung portion condemnations, and need to be taken into
account in the application of quantitative methods, such as cluster detection for disease
surveillance involving these data. In addition, the results show differences in the models
constructed for liver and lung portion condemnations, as well as between portion and
previously explored whole carcass condemnation data [17]. These findings suggest that
72
different variables may be associated with condemnation rates depending on the type of
material being condemned, and should be modeled and controlled for during quantitative
surveillance on a portion-specific basis. Previous studies have demonstrated the
importance of identification of potential confounding variables and different methods of
controlling for these variables in cluster detection methods for disease surveillance [26,
27]. For example, a study by Kleinman et al. [26] compared the performance of the
space-time scan statistic using unadjusted and covariate-adjusted respiratory complaint
data in humans to account for confounding temporal factors such as day of the week,
month and holidays. The study concluded that failure to adjust for confounding variables
can produce many false alarms and/or mask potential outbreaks.
Pneumonic lungs and “parasitic” livers were used as examples to explore the
modeling of biological and non-biological factors associated with portion condemnations.
These condemnation designations were selected as they represented two of the most
frequently reported reasons for portion condemnations by inspectors at provincial
abattoirs during the study period. Syndromic surveillance is based on non-traditional data
sources. Though this allows for early warning about potential disease outbreaks these
systems are generally less sensitive than traditional laboratory based surveillance systems
[28]. Therefore it is of utmost importance to (i) use highly predictive models (adjusted for
known risk factors and confounders), and (ii) preserve the robustness of the models by
adhering to the principle of parsimony. These are well known contradicting goals in
predictive modeling that make sufficient/large sample sizes an important requirement.
Therefore, pneumonic lung and “parasitic liver” condemnation rates were selected as they
were a rich data source from which to estimate trends. Livers are important from a public
73
health and economic standpoint, as they are a common edible portion in cattle and
represent a possible food safety concern. Lungs, though generally not consumed by the
average Ontarion, are also an important animal and public health concern, as lesions such
as tuberculosis granulomas may be found in inedible organs/tissues such as the lungs.
These are important factors to consider when selecting portion condemnation
designations for syndromic surveillance; a previous study by Thomas et al. [19],
investigating the use of portion condemnations in market hogs, noted that the quality of
data recording was poor for organs that were not considered to be economically
important or a concern for food safety.
It was interesting to find similarities and differences in terms of the significant
variables and the impact the variables had on condemnation rates in both the liver and
lung portion models, as well as the previously described models for whole carcass
condemnations [17]. The variables year, animal class, season and annual audit rating
were found to be significantly associated with condemnation rates in both portion models
as well as whole carcasses. It is not surprising that season and animal class were found to
be significant factors associated with abattoir condemnation rates in all three models, as
many animal diseases tend to have a distinct seasonality and high risk age groups
associated with the disease. For example, bovine respiratory disease complex more
commonly infects calves following a stressful event, such as sudden change in weather
conditions [29], and older cattle are generally at higher risk for being culled due to
disease and thus condemned more frequently. In the whole carcass condemnation data,
the variable year, identified patterns assumed to be associated with the discovery of
Bovine Spongiform Encephalopathy (BSE) in Alberta, Canada in 2003 [17]. In the
74
portion models, year also appeared to have a significant decreasing trend in calves for
pneumonic lung condemnations. It is suspected that these temporal trends in liver and
lung condemnations also stem from regulation changes due to BSE. Prior to BSE in
Canada, it was legal to load and transport downer animals with a veterinary certification
of fitness for slaughter. However, during 2004, changes were made to federal cattle
transportation regulations which forbid the transportation and slaughter of non-
ambulatory animals [30]. Ambulatory animals may be less likely to have lung pathology
than compromised and downer animals and are likely reflected in the marked decrease in
condemnation rates in pneumonic lungs (personal communication Ab Rehmtulla, DVM,
OMAFRA, Stone Road, Guelph, Ontario). In contrast, “parasitic liver” condemnations
increased in cows over the study period and may also reflect the type and quality of cattle
being sent to slaughter; however, it is unclear why an increase was seen over the study
period. Audit rating appeared to be an important variable for provincial abattoir
condemnation data. It was hypothesized that audit rating may reflect an abattoir’s
compliance to regulations and/or willingness to accept animals of poorer quality.
Although, different trends were found in these variables between all 3 models, the overall
statistically significant association of these variables with condemnation rates appears to
be “universal” within Ontario provincial abattoir data, and should be controlled for in
quantitative cluster detection methods for surveillance. Although the AAA and C
categories represent a small number of abattoirs over the study period, we felt that it was
important to not collapse the categories, as these abattoirs point to unique qualities among
these establishments. Since this study was conducted, the audit rating system was
simplified in September 2010 to a 3-grade system including pass, conditional pass and
75
fail to better reflect systems used in several jurisdictions for evaluating food safety at
restaurants [25]. This change in reporting may need to be accounted for in future studies
and/or in attempts to control for non-disease issues when conducting space-time cluster
analyses.
Agricultural region was found to be significantly associated with the portion
condemnation rates but not with whole carcass rates. It was interesting to note that
overall, predicted condemnation rates for bovine lungs and liver were lower in northern
and southern Ontario regions compared to other regions for all animal classes throughout
the study period. This pattern was particularly evident in the lung dataset. This regional
difference of condemnation rates in pneumonic lungs may reflect a genuine regional
difference; however, it likely stems from the quality of animals being sent to slaughter in
these regions. Abattoirs in southern and northern Ontario did not specialize and rarely
received non-ambulatory cattle. Prior to 2004, most of the so-called “downer plants”
were located in southwestern, central and eastern Ontario. While the regional differences
in “parasitic” livers may be due to a smaller concentration of dairy cattle in northern and
to some extent southern Ontario, corresponding with the lower condemnation rates in
these areas; it was surprising that variables related to abattoir processing capacity, such as
number of cattle processed each year and number of weeks an abattoir processed cattle
was only significant in the lung condemnation model. This may reflect issues with
processing speeds within abattoirs, perhaps the speeds impact lung inspection more than
livers since livers are more commonly considered to be an edible portion of cattle and
thus more carefully inspected.
76
It is important to identify and understand the factors which may cause “noise” in
the data before any quantitative methods can be chosen for disease surveillance. This
extra work beforehand can save valuable time and resources investigating “false alarms”
after the application of quantitative methods. Although differing results were found in the
two portion models as well as the previously described whole carcass model, common
themes arose from the results. Bovine abattoir condemnation data are sensitive to the
effects of regulatory and economic changes in the industry. Therefore it is important to
adjust models as regulations change over time. In addition, seasonal, secular, and non-
disease factors, such as commodity class, abattoir rating and processing capacity also
seemed to be important factors for bovine abattoir condemnation data and should be
adjusted in subsequent cluster detection analyses to prevent biased results. All analyses
thus far have been conducted retrospectively, facilitating the use of historical data to
highlight variables which need to be taken into account prior to applying outbreak
detection methods. However, the practical application of disease surveillance would be
conducted prospectively, and would have implications on the models. For example, the
variable year was found to be a very important variable in all of the models; however this
variable would not be applicable in a prospective analysis. The secular trend effect would
have to be accounted for in the analyses in some other way, for example using a trend
polynomial or trend filter. In addition, we were only able to account for clustering by
abattoir. However, it would be useful to explore the effect of clustering by inspector as
well. Unfortunately, these data were not available to explore the effect of inspector.
Specific training of inspectors may also improve the ability of a syndromic surveillance
system to detect unusual events. For example, the broad use of the category “parasitic
77
liver” may be more useful at detecting changes in disease if inspectors used more specific
criteria when condemning organs.
The suitability of portion vs. whole carcass condemnation data for syndromic
surveillance is difficult to ascertain in this study. It is suspected that for syndromic
surveillance purposes, portion condemnation data will be more specific and sensitive at
detecting changes in condemnation rates. However, validation of this conclusion is
difficult with no major documented disease outbreaks in Ontario cattle during the study
period. Nevertheless, it is encouraging that regulatory changes surrounding the
identification of BSE in Alberta, Canada was identified in both the whole carcass and
portion condemnation data. A similar study in pigs using Ontario provincial abattoir data
found that whole hog carcass condemnation data performed better than portion carcass
condemnation data at detecting disease clusters consistent with a documented porcine
circovirus-associated disease outbreak in Ontario [19]. Further investigations, perhaps
using simulated data, are needed to determine which types of data are more suitable for
the syndromic surveillance of specific types of diseases.
CONCLUSIONS
Findings from this study suggest that there are “universal” factors associated with
condemnation data, such as animal class, year, season and audit rating, additional
material-specific covariates are important and should be modeled and controlled for in
quantitative methods for disease surveillance, such as in cluster detection methods.
Validation of results is an issue with bovine portion abattoir condemnation data, as there
were no documented outbreaks in Ontario cattle during the study period. Further
investigations are needed to determine whether portion data would perform better than
78
whole carcass condemnation data for disease surveillance, and whether certain portions
are more appropriate than whole carcass for the surveillance of specific diseases.
ACKNOWLEDGEMENTS
The authors would like to acknowledge the following organizations for their
support for infrastructure, data retrieval and funding for this project: Canada Foundation
for Innovation, the Ontario Research Fund, the Ontario Graduate Scholarship,
OMAFRA/University of Guelph Research Program and Ontario Ministry of Agriculture,
Food and Rural Affairs (OMAFRA).
79
REFERENCES
1. Bovine spongiform encephalopathy (BSE)
[http://www.omafra.gov.on.ca/english/food/inspection/meatinsp/abattoir_audit.htm].
2. Timeline of BSE in Canada and the U.S
[http://www.cbc.ca/news/background/madcow/timeline.html].
3. Carman S, van Dreumel T, Ridpath J, Hazlett M, Alves D, Dubovi E, Tremblay R,
Bolin S, Godkin A, Anderson N: Severe acute bovine viral diarrhea in Ontario, 1993-
1995. J Vet Diagn Invest 1998, 10:27–35.
4. Poljak Z, Dewey CE, Rosendal T, Friendship RM, Young B, Berke O: Spread of
porcine circovirus associated disease (PCVAD) in Ontario (Canada) swine herds:
Part I. Exploratory spatial analysis. BMC Vet Res 2010, 6:59.
5. Ontario Cattlemen's Association - Market Information Service
[http://www.cattle.guelph.on.ca/markets/marketinfo.asp].
6. Steiner-Sichel L, Greenko J, Heffernan R, Layton M, Weiss D: Field investigations of
emergency department syndromic surveillance signals. New York City: MMWR;
2004:184–189.
7. Heffernan R, Mostashari F, Das D, Besculides M: New York City syndromic
surveillance systems. New York City: MMWR; 2004:25–27.
8. Edge VL, Pollari F, Ng LK, Michel P, McEwen SA, Wilson JB, Jerrett M, Sockett PN,
Martin SW: Syndromic surveillance of Norovirus using over-the-counter sales of
medications related to gastrointestinal illness. Can J Infect Dis Med Microbiol 2006,
17:235–241.
9. van Dijk A, McGuinness D, Rolland E, Moore KM: Can Telehealth Ontario
respiratory call volume be used as a proxy for emergency department respiratory
visit surveillance by public health? CJEM 2008, 10:18–24.
10. Hadorn DC, Stark KDC: Evaluation and optimization of surveillance systems for
rare and emerging infectious diseases. Vet Res 2008, 39:57–69.
11. Karama M, Johnson RP, Holtslander R, McEwen SA, Gyles CL: Prevalence and
characterization of verotoxin-producing Escherichia coli (VTEC) in cattle from an
Ontario abattoir. Can Vet Res 2008, 72:297–302.
12. Vilas VJDR, Bohning D, Kuhnert R: A comparison of the active surveillance of
scrapie in the European Union. Vet Res 2008, 39:37–52.
13. Kaneene JB, Miller R, Meyer RM: Abattoir surveillance: The U.S. experience. Vet
Microbiol 2006, 112:273–282.
80
14. Vourc'h G, Bridges VE, Gibbens J, De Groot BD, McIntyre L, Poland R, Barnouin J:
Detecting emerging diseases in farm animals through clinical observations. Emerg
Infect Dis 2006, 12:204–210.
15. Amezcua M, Pearl D, Friendship R, McNab W: Evaluation of a veterinary-based
syndromic surveillance system implemented for swine. Can J Vet Res 2010, 74:241–
250.
16. Van Metre DC, Barkey DQ, Salman MD, Morley PS: Development of a syndromic
surveillance system for detection of disease among livestock entering an auction
market. JAVMA 2009, 234:658–664.
17. Alton GD, Pearl DL, Bateman KG, McNab WB, Berke O: Factors associated with
whole carcass condemnation rates in provincially-inspected abattoirs in Ontario
2001-2007: implications for food animal syndromic surveillance. BMC Vet Res 2010,
6:42.
18. Weber WD: Development of an animal health monitoring system based on slaughter
condemnation data. Miami: Proceedings of the Eighth International Society for Disease
Surveillance Conference; 2009. 3-4 December.
19. Thomas-Bachli AL, Pearl DL, Friendship RM, Berke O: Suitability and limitations
of portion-specific abattoir data as part of an early warning system for emerging
diseases of swine in Ontario. BMC Vet Res 2012, 6:3.
20. Van Donkersgoed J, Jewison G, Mann M, Cherry B, Altwasser B, Lower R, Wiggins
K, Dejonge R, Thorlakson B, Moss E, Mills C, Grogan H: Canadian beef quality audit.
Can Vet J 1997, 38:217–225.
21. Ansong-Danquah J: A survey of carcass condemnation at a poultry abattoir and
its application to disease management. Can Vet J 1987, 28:53-56.
22. Schamber G, Olson C, Witt L: Condemnations in cattle presented for slaughter.
Bovine Practice 1981, 2:19–21.
23. Your responsibilities under the meat inspection act (Ontario)
[https://ospace.scholarsportal.info/bitstream/1873/4777/2/10308452.pdf].
24. Robertson C, Nelson TA, MacNab YC, Lawson AB: Review of methods for space-
time disease surveillance. Spat Spattemporal Epidemiol 2010, 1:105–116.
25. The Meat Plant Audit Program
[http://www.omafra.gov.on.ca/english/food/inspection/meatinsp/abattoir_audit.htm].
26. Kleinman K, Abrams A, Kulldorff M, Platt R: A model-adjusted space-time scan
statistic with an application to syndromic surveillance. Epidemiol Infect 2005,
133:409–419.
81
27. Klassen A, Kulldorff M, Curriero F: Geographical clustering of prostate cancer
grade and stage at diagnosis, before and after adjustment for risk factors. Int J
Health Geogr 2005, 4:1.
28. Lawson AB, Kleinman K: Spatial and Syndromatic Surveillance for Public Health:
Chichester, West Sussex; Hoboken. NJ: J. Wiley; 2005.
29. Ogilvie TH: Large Animal Internal Medicine: 1st Edition ed. Baltimore, Maryland:
Williams and Wilkens; 1998.
30. Overview of Canada's BSE Safeguards
[http://www.inspection.gc.ca/english/anima/disemala/bseesb/bseesbfs2e.shtml].
82
Table 3.1 – Summary of Ontario provincial abattoir bovine portion condemnations
2001 -2007.
Organ or body system
(# of condemnations / %)
Reason for condemnation Number of portion
condemnations (%)
Front quarter Arthritis 427 (0.11 %)
(28 318 / 7.02 %) Inflammation 701 (0.17 %)
Abscess 955 (0.24 %)
Bruising 12 947 (3.21 %)
Contamination 13 288 (3.29 %)
Head
(including tongue)
Xanthomatosis 40 (0.01%)
(9519 / 2.36 %) Actino 670 (0.17 %)
Contamination 2421 (0.60 %)
Abscess
Erosion
3723 (0.92 %)
2665 (0.66%)
Heart Adhesion 8261 (2.05 %)
Hind quarter Abscess 174 (0.04 %)
Hind quarter Inflammation 516 (0.13 %)
(49 858 / 12.36 %) Arthritis 2132 (0.53 %)
Abscess 4349 (1.08 %)
Contamination 14 053 (3.48 %)
Bruising 28 808 (7.14 %)
Kidneys Cystic 31 570 (7.83 %)
(93 589 / 23.21 %) Nephritis 62 019 (15.48% )
Liver Cirrhosis 1666 (0.41 %)
(168 183 / 41.70 %) Melanosis 8368 (2.07 %)
Adhesion 29 125 (7.22 %)
Abscess 57 511 (14.26 %)
Parasitic 71 513 (17.73 %)
Loin Abscess 456 (0.11 %)
Ribs Abscess 46 (0.01 %)
Shoulder/limb Arthritis 214 (0.05 %)
Stifle joint Arthritis 693 (0.17 %)
Trim Bruising 7096 (1.76 %)
Lungs Pneumonia 36 883 (9.15 %)
Total portions condemned 403 290
Total cattle slaughtered 1 155 535
83
Table 3.2 – Summary of number of cattle processed, number of weeks open, audit
rating and animal class
a) Number of
animals
processed
Number of Abattoirs
2001 2002 2003 2004 2005 2006 2007
1 – 286 73 72 36 24 30 40 38
287 – 498 43 35 40 28 36 34 26
499 – 841 24 24 36 40 40 32 35
842 – 27 286 22 26 33 49 41 32 30
b) Number of
weeks open
Number of Abattoirs
1 – 43 66 67 44 32 43 38 38
44 – 49 56 45 48 45 30 39 29
50 – 51 18 28 25 31 41 31 32
52 22 17 28 33 33 30 30
c) Audit
Rating
Number of Abattoirs
AAA 1 0 0 0 0 1 1
AA 6 6 7 13 15 22 28
A 61 71 77 76 78 78 72
B 14 8 8 5 8 6 9
C 2 1 0 0 1 0 0
Unrated 78 71 53 47 45 31 19
d) Animal
Class
1Median sales-price of animal class per year 2001-2007
Calves 136.53 117.95 112.19 83.61 107.14 119.69 108.49
Cows 64.28 58.64 27.11 19.70 25.38 31.84 34.77
Heifers 112.01 101.22 82.16 73.61 87.74 90.87 91.09
Steers 113.24 102.57 83.15 75.60 91.07 93.55 92.56 1Median sales-price in Canadian dollars per 100 lbs.
Summary of the a) quartiles of the number of cattle processed, b) quartiles of the number
of weeks at least one bovine animal was processed, c) annual OMAFRA audit rating, and
d) median sales price of calves, cows, heifers, steers 2001–2007 for portion
condemnations.
84
Table 3.3 – Multivariable multi-level Poisson regression using pneumonic lung
condemnation rates
Variable Categories IRR1 Std. Err. P-value 95% CI
Year 2001 --- --- --- --- ---
2002 0.83 0.02 < 0.01 0.79 0.86
2003 1.14 0.02 < 0.01 1.09 1.19
2004 0.80 0.03 < 0.01 0.76 0.84
2005 0.75 0.03 < 0.01 0.71 0.79
2006 0.45 0.03 < 0.01 0.43 0.49
2007 0.31 0.04 < 0.01 0.28 0.34
Animal class Calves --- --- --- --- ---
Cows 0.30 0.23 < 0.01 0.19 0.47
Heifers 0.39 0.11 < 0.01 0.32 0.49
Steers 0.16 0.07 < 0.01 0.14 0.19
Season Winter --- --- --- --- ---
Spring 1.07 0.02 < 0.01 1.03 1.11
Summer 0.96 0.02 0.03 0.93 0.99
Fall 1.10 0.02 < 0.01 1.06 1.15
Region Central --- --- --- --- ---
Eastern 1.88 0.54 0.24 0.66 5.39
Northern 0.02 1.31 < 0.01 0.001 0.20
Southern 0.75 0.47 0.54 0.30 1.90
Western 0.98 0.46 0.97 0.40 2.42
Rating C --- --- --- --- ---
AAA 0.24 1.26 0.26 0.02 2.85
AA 0.18 0.64 0.01 0.05 0.62
A 0.11 0.63 < 0.01 0.03 0.37
B 0.15 0.63 < 0.01 0.04 0.50
unrated 0.29 0.64 0.048 0.08 0.99
# of animals 1-286 --- --- --- --- ---
287-498 0.70 0.12 < 0.01 0.56 0.89
499-841 0.36 0.15 < 0.01 0.27 0.48
842-27286 0.27 0.14 < 0.01 0.20 0.35
# of weeks 1-43 --- --- --- --- ---
44-49 1.72 0.07 < 0.01 1.50 1.97
50-51 2.84 0.06 < 0.01 2.55 3.16
52 2.31 0.05 < 0.01 2.11 2.54
Season x animal class Winter x calves --- --- --- --- ---
Spring x heifers 0.93 0.05 0.14 0.85 1.02
Summer x heifers 0.92 0.05 0.10 0.83 1.02
Fall x heifers 0.83 0.05 < 0.01 0.76 0.92
Spring x steers 0.91 0.04 0.01 0.84 0.98
Summer x steers 0.94 0.04 0.10 0.88 1.01
85
Fall x steers 0.82 0.04 < 0.01 0.77 0.88
Spring x cows 0.78 0.09 0.01 0.66 0.93
Summer x cows 0.73 0.09 < 0.01 0.61 0.88
Fall x cows 0.65 0.09 < 0.01 0.55 0.78
Region x animal class Central x calves --- --- --- --- ---
South x heifers 0.37 0.09 < 0.01 0.31 0.45
West x heifers 0.65 0.09 < 0.01 0.53 0.78
North x heifers 6.58 1.25 0.13 0.56 76.70
East x heifers 0.72 0.16 0.05 0.54 0.99
South x steers 0.68 0.07 < 0.01 0.60 0.78
West x steers 2.81 0.06 < 0.01 2.52 3.14
North x steers 67.90 1.06 < 0.01 8.44 546.45
East x steers 1.64 0.14 < 0.01 1.25 2.15
South x cows 0.26 0.27 < 0.01 0.15 0.45
West x cows 1.60 0.20 0.02 1.09 2.34
North x cows 7.39 1.49 0.18 0.40 136.25
East x cows 1.05 0.24 0.85 0.66 1.66
Year x animal class 2001 x calves --- --- --- --- ---
2002 x heifers 1.84 0.08 < 0.01 1.59 2.13
2003 x heifers 1.53 0.08 < 0.01 1.32 1.78
2004 x heifers 1.64 0.08 < 0.01 1.41 1.92
2005 x heifers 1.24 0.08 < 0.01 1.06 1.45
2006 x heifers 2.23 0.08 < 0.01 1.90 2.63
2007 x heifers 2.68 0.09 < 0.01 2.25 3.20
2002 x steers 2.05 0.05 < 0.01 1.85 2.27
2003 x steers 2.27 0.05 < 0.01 2.06 2.49
2004 x steers 2.44 0.05 < 0.01 2.20 2.69
2005 x steers 1.64 0.05 < 0.01 1.48 1.83
2006 x steers 2.60 0.06 < 0.01 2.32 2.92
2007 x steers 2.86 0.06 < 0.01 2.54 3.23
2002 x cows 1.07 0.19 0.74 0.73 1.55
2003 x cows 3.09 0.14 < 0.01 2.35 4.08
2004 x cows 3.63 0.14 < 0.01 2.76 4.77
2005 x cows 1.57 0.15 < 0.01 1.17 2.12
2006 x cows 2.57 0.16 < 0.01 1.89 3.49
2007 x cows 2.08 0.18 < 0.01 1.46 2.98 1Incidence rate ratio.
Multivariable multi-level Poisson regression model (PQL-2) investigating the association
between pneumonic lung condemnation rates in Ontario provincial abattoirs and year-
animal class interaction, season-animal class interaction, region-animal class interaction,
number of animals, number of weeks of processing and audit rating.
86
Table 3.4 – Multivariable multi-level Poisson regression using pneumonic lung
condemnation rates
Variable Categories IRR1 Std.
Err.
P-value
95% CI
Year 2001 --- --- --- --- ---
2002 1.08 0.04 0.08 0.99 1.17
2003 1.73 0.04 < 0.01 1.60 1.88
2004 1.81 0.04 < 0.01 1.67 1.96
2005 2.06 0.04 < 0.01 1.90 2.24
2006 2.29 0.04 < 0.01 2.12 2.49
2007 1.71 0.04 < 0.01 1.57 1.86
Animal class Calves --- --- --- --- ---
Cows 1.00 0.06 1.00 0.89 1.13
Heifers 1.45 0.05 < 0.01 1.30 1.60
Steers 1.29 0.08 < 0.01 1.10 1.50
Season Winter --- --- --- --- ---
Spring 0.88 0.03 < 0.01 0.83 0.94
Summer 1.08 0.03 < 0.01 1.02 1.14
Fall 1.22 0.03 < 0.01 1.15 1.29
Region Central --- --- --- --- ---
Southern 0.49 0.18 < 0.01 0.35 0.69
Western 0.61 0.18 < 0.01 0.43 0.86
Northern 0.22 0.31 < 0.01 0.12 0.41
Eastern 0.23 0.22 < 0.01 0.15 0.35
Rating C --- --- --- --- ---
AAA 0.53 0.13 < 0.01 0.41 0.69
AA 0.72 0.09 < 0.01 0.60 0.85
A 0.72 0.09 < 0.01 0.60 0.85
B 0.74 0.09 < 0.01 0.62 0.88
unrated 0.58 0.10 < 0.01 0.48 0.71
Season x animal class Winter x calves --- --- --- --- ---
Spring x heifers 1.04 0.04 0.37 0.96 1.12
Summer x heifers 0.96 0.04 0.33 0.90 1.04
Fall x heifers 0.90 0.04 0.00 0.84 0.97
Spring x steers 0.96 0.04 0.30 0.90 1.03
Summer x steers 0.89 0.03 < 0.01 0.84 0.95
Fall x steers 0.95 0.03 0.11 0.89 1.01
Spring x cows 1.13 0.04 < 0.01 1.05 1.21
Summer x cows 0.87 0.04 < 0.01 0.81 0.93
Fall x cows 0.75 0.03 < 0.01 0.70 0.80
Region x animal class Central x calves --- --- --- --- ---
South x heifers 1.90 0.05 < 0.01 1.71 2.10
West x heifers 1.82 0.05 < 0.01 1.65 2.00
North x heifers 2.96 0.17 < 0.01 2.11 4.15
East x heifers 7.34 0.09 < 0.01 6.10 8.82
South x steers 1.55 0.05 < 0.01 1.42 1.69
87
West x steers 1.76 0.04 < 0.01 1.63 1.91
North x steers 2.90 0.17 < 0.01 2.09 4.01
East x steers 6.09 0.09 < 0.01 5.10 7.28
South x cows 0.96 0.06 0.45 0.85 1.07
West x cows 1.38 0.05 < 0.01 1.26 1.57
North x cows 3.12 0.17 < 0.01 2.24 4.34
East x cows 5.96 0.09 < 0.01 4.96 7.17
Year x animal class 2001 x calves --- --- --- --- ---
2002 x heifers 0.88 0.06 0.02 0.79 0.98
2003 x heifers 0.52 0.06 < 0.01 0.47 0.58
2004 x heifers 0.57 0.05 < 0.01 0.51 0.63
2005 x heifers 0.58 0.05 < 0.01 0.53 0.65
2006 x heifers 0.55 0.05 < 0.01 0.50 0.61
2007 x heifers 0.72 0.05 < 0.01 0.65 0.80
2002 x steers 0.74 0.05 < 0.01 0.67 0.82
2003 x steers 0.65 0.05 < 0.01 0.59 0.72
2004 x steers 0.63 0.05 < 0.01 0.58 0.69
2005 x steers 0.60 0.05 < 0.01 0.55 0.66
2006 x steers 0.50 0.05 < 0.01 0.45 0.54
2007 x steers 0.64 0.05 < 0.01 0.59 0.71
2002 x cows 1.09 0.09 0.32 0.92 1.29
2003 x cows 1.06 0.08 0.46 0.91 1.23
2004 x cows 1.41 0.07 < 0.01 1.23 1.63
2005 x cows 0.91 0.07 0.17 0.79 1.04
2006 x cows 1.01 0.07 0.88 0.88 1.16
2007 x cows 2.12 0.07 < 0.01 1.83 2.44
1Incidence rate ratio
Multivariable multi-level Poisson regression model (PQL-1) investigating the association
between “parasitic liver” condemnation rates in Ontario provincial abattoirs and year-animal class
interaction, season-animal class interaction, region-animal class interaction and audit rating.
88
Figure 3.1 - Map of provincially-inspected abattoirs in Ontario
89
Figure 3.2 - “Parasitic liver” condemnation rates per 1000 slaughtered cattle from
Ontario provincial abattoirs 2001 – 2007
90
Figure 3.3 - Pneumonic lung condemnation rates per 1000 slaughtered cattle from
Ontario provincial abattoirs 2001 -2007
91
Figure 3.4 - Model expected pneumonic lung condemnation rates based on multi-level Poisson model
92
Figure 3.5 - Model expected “parasitic liver” condemnation rates based on multi-level Poisson model
93
CHAPTER FOUR:
COMPARISON OF COVARIATE ADJUSTMENT METHODS USING SPACE-
TIME SCAN STATISTICS FOR FOOD ANIMAL SYNDROMIC
SURVEILLANCE
(As published in BMC Veterinary Research 2013; 9:231 with minor editorial
revisions)
ABSTRACT
Abattoir condemnation data show promise as a rich source of data for syndromic
surveillance of both animal and zoonotic diseases. However, inherent characteristics of
abattoir condemnation data can bias results from space-time cluster detection methods for
disease surveillance, and may need to be accounted for using various adjustment
methods. The objective of this study was to compare the space-time scan statistics with
different abilities to control for covariates and to assess their suitability for food animal
syndromic surveillance.
Four space-time scan statistic models were used including: animal class adjusted
Poisson, space-time permutation, multi-level model adjusted Poisson, and a weighted
normal scan statistic using model residuals. The scan statistics were applied to monthly
bovine pneumonic lung and “parasitic liver” condemnation data from Ontario provincial
abattoirs from 2001–2007.
The number and space-time characteristics of identified clusters often varied
between space-time scan tests for both “parasitic liver” and pneumonic lung
condemnation data. While there were some similarities between isolated clusters in
space, time and/or space-time, overall the results from space-time scan statistics differed
substantially depending on the covariate adjustment approach used.
94
Variability in results among methods suggests that caution should be used in
selecting space-time scan methods for abattoir surveillance. Furthermore, validation of
different approaches with simulated or real outbreaks is required before conclusive
decisions can be made concerning the best approach for conducting surveillance with
these data.
95
BACKGROUND
With the development and availability of geographic information systems (GIS),
there has been an increasing trend in human and animal disease surveillance towards
capturing both temporal and spatial data for health and disease outcomes. Spatio-
temporal scan statistics are one of the most widely used methodologies [1] for
surveillance and have been shown to be useful for surveillance and outbreak detection in
both human and animal health applications [2-7]. Space-time scan statistics are one type
of spatio-temporal surveillance method which uses a cylindrical scanning window to scan
spatially by varying the size of the cylinder radius and scan temporally by varying the
height of the cylinder. Statistical significance of the cluster is determined by Monte Carlo
based simulations. Analysis can be conducted both retrospectively, as well as
prospectively, making it suitable for disease surveillance [8].
Syndromic surveillance is the amalgamation of signs/symptoms using data from
non-traditional sources [9]. The sign/symptom groupings are loosely designated as
‘syndromes’, and are used to track disease trends in populations and signal a possible
outbreak that warrants further investigation [9]. Historically, syndromic surveillance has
primarily been applied to human health data [10-12]. However, in recent years there has
been a growing trend towards the application of these methods for animal health
surveillance data [13-17]. Abattoir condemnation data are a rich source of information for
syndromic surveillance, and have the potential to provide early warning of emerging
animal and zoonotic disease but have been under-utilized in the past. Ontario provincial
abattoir data are particularly advantageous for syndromic surveillance and the application
96
of spatio-temporal methods, as they represent a fairly local picture of animal health
events, with cattle being shipped to abattoirs originating from farms less than 100 km
away [17].
Scan statistics identify the approximate locations of disease clusters in space and
time, and make use of a variety of statistical models [1,2], making them useful for a
variety of data. However, space-time scan statistics and current available software do
have some limitations and assumptions, which may be violated by the inherent
characteristics of provincial abattoir data. For example, while the space-time scan statistic
has the ability to control for covariates, at this time, this is only applicable to categorical
variables, thus limiting the type of variables one can control for in the analysis [1]. The
space-time permutation model inherently corrects for purely spatial and purely temporal
clusters, however, the expected rates of disease are dependent on a relatively stable
background population [18]. While this is generally true for human populations and
periods of a few years, this is generally not the case with abattoir data, where animal
population sizes can vary by season.
In recent years, model-based approaches have emerged to account for covariates
such as age, gender, and seasonality in expected rates of disease, in response to the
limited ability of space-time scan statistic software to include these covariate data [1].
Statistical modeling allows for adjustment of disease risk for both categorical and
continuous variables in space and time. By combining both methods, surveillance
researchers have the ability to account for relevant covariates, while locating clusters in
space and time. A study by Kleinman et al. [6], used this approach by conducting model-
97
adjusted space-time scan tests for syndromic surveillance of lower respiratory complaints
in a human health care setting. The study controlled for non-disease factors such as day
of week, month, and holidays and found that the number of false alarms could be reduced
by removing the “noise” of predictable covariates. However, this method has not been
applied to animal condemnation data for disease surveillance. Previous studies by Alton
et al. and Thomas et al. found that various seasonal, secular, disease and abattoir
characteristic factors were associated with condemnation rates in Ontario provincial
abattoirs; they stressed that these might be accounted for in the application of quantitative
space-time cluster detection methods for disease surveillance involving abattoir data [17-
19]. This study also highlights the importance of thinking beyond the typical age and sex
covariate adjustment and controlling for disease and non-disease factors such as animal
throughput at the abattoir and sales price of the animal class which may have a
considerable impact on the results, particularly for abattoir condemnation data.
Due to the variety of methods available for covariate-adjustment in cluster
detection, and their varying level of complexity in terms of analysis, a comparison study
of the space-time scan statistic on four different approaches at controlling for covariates
was used. If similarities in results were found between multiple approaches, then the most
parsimonious model could be recommended. Four covariate adjusted scan tests were
compared to results from the unadjusted space-time scan test including: 1) categorical
variable adjustment which stratifies on the covariate variable of interest within the space-
time scan statistic, 2) space-time permutation model which uses only case data and
inherently controls for purely spatial and purely temporal clusters, 3) multi-level model
adjusted approach which allows for adjustment of both categorical and continuous
98
variables, and 4) multi-level model residual-adjusted approach which uses the
standardized residuals from the above mentioned multi-level model to control for
covariates.
The objective of this study was to demonstrate four commonly used space-time
scan statistic approaches with different abilities to control for covariates that animal
health surveillance workers might consider when using statistical methods to identify
outbreaks of disease using abattoir condemnation data and assess their suitability for food
animal syndromic surveillance involving Ontario provincial abattoir condemnation data.
METHODS
Data source and variables
Data regarding bovine “parasitic liver” and pneumonic lung condemnations were
extracted from the Food Safety Decision Support System (FSDSS) database maintained
by the Ontario Ministry of Agriculture, Food and Rural Affairs (OMAFRA). The
database contains information regarding the number and reason for daily portion
condemnations in provincially inspected abattoirs in Ontario. These portion
condemnation categories were selected for this analysis as an example dataset, as they
were among the most frequently reported portion condemnations by provincial inspectors
during the study period [19]. Additionally, as bovine livers are an edible portion, these
data may represent a potential food safety and/or quality concern. “Parasitic liver” is an
inspection term used to label bovine livers considered unfit for human consumption, due
to lesions such as necrosis, fibrosis, cirrhosis, atrophy, telangiectasia, and adhesions.
Although the term “parasitic liver” suggests truly parasitic infections such as fascioliasis,
99
the term covers non-parasitic conditions as well (personal communication Ab Rehmtulla,
DVM, OMAFRA, Stone Road, Guelph, Ontario). Pneumonic lung condemnation refers
to bovine lungs which were condemned for lesions indicative of a previous localized and
resolved antero-ventral pneumonia infection.
Data were extracted from the database for cattle animal classes: bulls, calves,
cows, heifers and steers from January 1, 2001 – December 31, 2007. Data from 45,148
bulls were excluded from subsequent analyses due to missing data and inconsistencies in
the use of this classification. Missing geographical coordinates for 54 abattoirs (26%)
were approximated using postal codes (76%) and/or addresses (24%) with the address
geocoding software GeoPinpoint Suite 6.4 (DMTI Spatial Inc., Markham, Ontario,
Canada).
Space-time scan statistic
The space-time scan statistic was used to identify abattoirs with high and low
“parasitic liver” and pneumonic lung condemnation rates in space-time using SaTScan
v8.0 (Kulldorff M. and Information Management Services Inc., 2009.), and were
visualized on maps using ArcGIS 9.2 (ESRI, Redlands, California, USA). Four different
approaches to control for confounding variables were compared to each other and an
unadjusted Poisson scan statistic including: (1) animal class adjusted Poisson scan
statistic, (2) space-time permutation, (3) multi-level model adjusted Poisson scan statistic,
and (4) a weighted normal scan statistic using model residuals. For all scan tests, latitude
and longitude coordinates for each abattoir, and premise identification number were used
to create the coordinates file. A maximum spatial cluster size of 50% of the population at
100
risk and maximum temporal cluster size of 50% of the study period were used. For all
scan tests, 9999 Monte Carlo replications were used to estimate the significance levels of
the clusters. For all analyses, the most likely (based on the size of the log-likelihood
ratio), non-overlapping in space-time, statistically significant (α = 0.05) clusters are
presented. Secondary clusters were set to allow some overlap as long as the secondary
cluster and a previously reported cluster did not both contain each other’s centroid. By
allowing some overlap, we were able to identify space-time clusters that overlapped in
space, but not time. Only the most likely non-overlapping clusters were reported to
simplify the presentation of the results. All the tests were run as two-sided tests scanning
for both high and low levels of disease to identify disease clusters as well as abattoirs
with unusually low condemnation rates.
For the unadjusted scan statistic, monthly raw counts of “parasitic
liver”/pneumonic lung condemnations and monthly number of cattle slaughtered were
used to create the case and population files respectively using a Poisson distribution. In
the animal-adjusted scan statistic, monthly raw counts of “parasitic liver”/pneumonic
lung condemnations and monthly number of cattle slaughtered were used to create the
case and population files respectively using a Poisson distribution. Cattle animal class
(e.g., calves, cows, heifers, steers) were adjusted with the space-time scan statistic by
stratifying on the variable within the case file. For the space-time permutation model, raw
case counts of “parasitic liver”/pneumonic lung condemnations were used. For the
model-adjusted scan test, a multi-level model was previously created to identify
economic, seasonal and abattoir processing capacity characteristics associated with
“parasitic liver” and pneumonic lung condemnation rates. The model identified year,
101
season, animal class, audit rating and region to be statistically associated with “parasitic
liver” condemnation rates [19]. For the pneumonic lung condemnation rates year, season,
animal class, region, audit rating, number of cattle processed per year, and number of
weeks abattoirs processed cattle were found to have a statistically significant association
[19]. For the model-adjusted scan statistic, standardized morbidity ratios were used in the
space-time scan statistic based on the observed number of condemnations and the model
predicted counts. For the model residual scan test, the observation level standardized
residuals from the multi-level model were analyzed. A Poisson model was used for the
model-adjusted scan test and a weighted normal model was used for the scan test using
the multi-level model residuals (at the level of the observation). The normal model
assumes that the quantitative variable (i.e., standardized residual) is independent and
identically distributed under the null hypothesis and therefore has the same variance.
Since the population at abattoirs changes greatly, the varying sample size at each abattoir
will cause the variance to be different for different abattoirs, thus a weighted normal
model was used to take into account the uncertainty of the observed rate [20]. The total
number of cattle slaughtered at each abattoir at each time period was used to account for
the variability.
RESULTS
A total of 211 provincially-inspected abattoirs, slaughtering a total of 1,155,535
cattle from 2001-2007 were included in this study. Provincially-inspected abattoirs can be
found throughout Ontario; however, over 80% of abattoirs processing cattle are located in
Southern, Western and Central Ontario regions. “Parasitic liver” and pneumonic lungs
condemnations were among the most frequently condemned portions and accounted for
102
approximately 18% and 9% of total condemned portions for the study period,
respectively. A complete description of these data and abattoir locations can be found in
Alton et al., 2012 [69].
“Parasitic liver” data
Results of the unadjusted Poisson space-time scan statistic identified 1 high rate
cluster from July 2004 – December 2007, and 1 low rate cluster from March 2002 –
August 2005 (Figure 4.1a and Table 4.1a). Results of the animal class adjusted space-
time scan statistic identified 2 high rate clusters from August 2003 – January 2007 and
December 2006 – Dec 2007, and 1 low rate cluster from December 2001 – May 2005
(Figure 4.1b and Table 4.1b). The model-adjusted space-time scan identified 3 high rate
clusters from September 2001 – December 2001, March 2001 – June 2002, and
September 2003 – May 2004, and 3 low rate clusters from January 2001 – April 2001,
April 2004 – November 2006 and April 2005 – January 2006 (Figure 4.1c and Table
4.1c). The space-time permutation model identified 2 high rate clusters from January
2001 – September 2002 and December 2006 – December 2007, and 1 low rate cluster
from January 2001 – September 2002 (Figure 4.1d and Table 4.1d). Lastly, the space-
time scan test applied to Poisson model residuals did not identify any statistically
significant high or low rate clusters.
Pneumonic lung data
Results of the unadjusted Poisson space-time scan statistic identified 1 high rate
cluster from January 2001 – June 2004, and 1 low rate cluster from June 2004 to
103
November 2007 (Figure 4.2a and Table 4.2a). Results of the animal class adjusted space-
time scan statistic identified 2 high rate clusters during January 2001 – June 2004 and
October 2001 to March 2005, and 1 low rate cluster during March 2004 – August 2007
(Figure 4.2b and Table 4.2b). The model-adjusted space-time scan statistic identified 1
high rate cluster during January 2001 – March 2002 and 3 low rate clusters during March
2004 – August 2007, January 2001 – March 2003 and April 2005 – November 2007
(Figure 4.2c and Table 4.2c). The space-time permutation model identified 1 high rate
cluster during Jan 2001 – February 2003, and 2 low rate clusters during January 2001 –
February 2003 and February 2005 – December 2007 (Figure 4.2d and Table 4.2d).
Lastly, the space-time scan statistic using multi-level model residuals identified 3 high
rate clusters during January 2001 – March 2003, July 2003 – June 2004 and January 2001
– February 2003, and 1 low rate cluster during January 2001 – March 2003 (Figure 4.2e
and Table 4.2e).
DISCUSSION
Provincial abattoir condemnation data may be useful for integration into a food
animal syndromic surveillance system; however, there are inherent characteristics in the
data which can bias the results of quantitative cluster detection methods. The results of
this methodological comparison study found differing results depending on the type of
covariate adjustment method used. By not considering or accounting for certain
covariates, such as non-disease factors (i.e. price, throughput), or selecting an
inappropriate statistical model for the data (i.e. space-time permutation model) the
subsequent results can lead to very different conclusions. These results suggest caution
should be exercised when arbitrarily selecting a space-time scan statistic model for
104
disease surveillance involving these data and highlight the importance of preliminary
validation studies using simulated or documented outbreak reports before a standard
cluster detection method is adopted by a surveillance system. A proper choice of the
method needs to take into account the properties of the sample data (e.g., its distribution,
case only vs. case and control data) and the question to be answered by the statistical test
(e.g., spatial versus space-time cluster location).
Farm location information is not routinely recorded for provincial abattoir
condemnation data. This lack of farm of origin location information for animals being
shipped to Ontario provincial abattoirs is a limitation in conducting spatial-temporal
cluster analyses. However, a previous study by Alton et al. [17], estimated the distance
between the animals’ farm and the abattoir using a subset of cattle, in which a sample
was sent for laboratory testing. The authors found cattle were shipped less than 100 km to
Ontario provincial abattoirs, and given the spatial scale of Ontario (1,000,000 km2),
abattoirs are considered to give an appropriate approximation of the disease rates among
locally slaughtered cattle. Over 75% of the abattoirs were geo-located by OMAFRA is
the FSDSS dataset; approximately 25% had missing co-ordinates and had to be geo-
located using addresses and/or postal code information. Of these, 54 abattoirs with
missing geo-locations, 76% were geo-located to the centroid of a postal code. However,
based on the potential difference between the abattoir location and centroid of its postal
code area in relation to the size of the study area, the impact would be negligible.
While there were differences between the four space-time scan statistic
approaches particularly for the “parasitic liver” condemnation data, there were some
105
similarities between the different approaches. For example, the “parasitic liver” data had
clusters which overlapped in space between the one high rate cluster and one low rate
cluster in the unadjusted and animal class adjusted approaches (Figure 4.1a and b) and
again between 2 high rate clusters and one low rate cluster in the model-adjusted and
space-time permutation model approaches (Figures 4.1c and d), however, none of these
clusters overlapped in time, and were at least a year apart. In contrast, pneumonic lung
data had a similar high rate cluster which overlapped in space and time between the
unadjusted and animal adjusted approaches (Figures 4.2a and b), a low rate cluster
between the model-adjusted and space-time permutation approach (Figures 4.2c and d)
and a low rate cluster between the model-adjusted, space-time permutation and residual
scan approaches (Figures 4.2c, d and e). There was also a similar high rate cluster in
space and time between the space-time permutation model and the residual scan (Figures
4.2d and e). While there were isolated clusters which overlapped between different
adjustment approaches, overall, the results of the scan statistics depicted very different
clusters between the different adjustment methods. The overall differences in results of
the covariate adjustment approaches suggest ignoring covariates beyond the animal level
may be unwise when using abattoir condemnation data for food animal syndromic
surveillance.
While each adjustment approach found differing results, it is also important to
consider assumptions of the scan statistic models in relation to the data being used. For
example, the space-time permutation model, which uses only case data, is advantageous
when population data are missing or difficult to obtain/sample. However, it is likely not
appropriate for use with provincial abattoir data, as the model assumes a stable
106
underlying population, which is not the case with these data. It was hypothesized that
perhaps the majority of the variation in condemnation rates may be attributed to animal
class, and perhaps by simply controlling for this categorical variable (using an adjustment
file when employing SaTScan) one would see similar results to that of the model-
adjusted approach. However, we found these approaches showed very different results.
We suspected that the model-adjusted and model residual adjustment approaches would
be the most appropriate for quantitative cluster detection involving provincial abattoir
condemnation data, as these methods are able to account for both categorical and
continuous variables, making it the most versatile of the adjustment approaches.
However, when the event is rare, as in the present example with condemnation rates, the
precision of the residuals using the normal model is unstable and may give inaccurate
results when applied to the space-time scan statistic [20]. Thus, this approach is not
appropriate for the application of provincial abattoir condemnation data unless counts are
aggregated to a higher temporal and/or spatial level. The model-adjusted approach using
the ratio of observed versus expected condemnations under the Poisson model would be
more appropriate when utilizing relatively rare events, as in the case of abattoir
condemnation data. However, this approach involves more complex analyses than some
of the other approaches. In addition, the current multi-level model includes temporal
variables such as, year and season, which are more conducive to retrospective analyses
and would be difficult to account for prospectively. Ultimately, to ensure the proper use
of these methods, validation of the different approaches with simulated or documented
outbreaks needs to be performed in selecting the most appropriate statistical test for these
data.
107
The comparison of the different methods for covariate adjustments highlights the
variability in the results for both pneumonic lung and “parasitic liver” portion
condemnation data. Overall, the results for both types of portion condemnation data
demonstrate that as you increase the detail in the covariate adjustment information, the
size of the cluster decreases. A study conducted by Kleinman et al. [6] compared the
space-time scan statistic using unadjusted and model-adjusted approaches for syndromic
surveillance of lower respiratory illness to account for confounding temporal and disease
variables, such as day of week, month, holidays and local history of illness. This study
found that during influenza season, large space-time clusters were identified almost every
weekday by the unadjusted approach compared to the model-adjusted approach, making
it unfeasible to investigate all ‘unusual’ events and diminishing the value of the tool for
surveillance. This adjustment effect could also be found in the current study, further
justifying the need for covariate adjustment with the space-time scan statistic for disease
surveillance purposes.
CONCLUSIONS
This study demonstrates the importance of identifying and adjusting for various
disease and non-disease factors which may bias the results of cluster detection methods
for disease surveillance. When selecting an adjustment method, it is important to consider
not only the inherent assumptions in the statistical method, but also these assumptions in
relation to the data being utilized. The variability in results stresses that there are a variety
of methods currently available for covariate adjustment, and that by simply selecting one
such method, without prior research and planning may yield very different and
potentially inaccurate results. Background studies such as this, which identify important
108
confounding factors and effectively correct for them may assist in improving the
sensitivity and specificity of outbreak detection by controlling for predictable clusters
creating false alarms and reducing the amount of time and resources needed for
investigation of potential clusters, as well as, identifying outbreaks that would have been
normally overlooked in the background “noise” of these data. Ultimately, validation of
different approaches with simulated or real outbreaks needs to be performed in selecting
the appropriate statistical test for these data for a food animal syndromic surveillance
context.
ACKNOWLEDGEMENTS
The authors would like to acknowledge Dr. Martin Kulldorff for his helpful
discussions concerning theoretical issues involving space-time scan statistics. The authors
would also like to acknowledge the following organizations for their support for
infrastructure, data retrieval and funding for this project: Canada Foundation for
Innovation (CFI), the Ontario Research Fund, the Ontario Graduate Scholarship,
OMAF/University of Guelph Research Program and Ontario Ministry of Agriculture &
Food (OMAF).
109
REFERENCES
1. Robertson C, Nelson TA, MacNab YC, Lawson AB: Review of methods for space–
time disease surveillance. Spat Spattemporal Epidemiol 2010, 1:105–116.
2. Pearl DL, Louie M, Chui L, Dore K, Grimsrud KM, Leedell D, Martin SW, Michel P,
Svenson LW, McEwen SA: The use of outbreak information in the interpretation of
clustering of reported cases ofEscherichia coli 0157 in space and time in Alberta,
Canada, 2000–2002. Epidemiol Infect 2006, 134:699–711.
3. Odoi A, Martin SW, Michel P, Middleton D, Holt J, Wilson JB: Investigation of
clusters of giardiasis using GIS and a spatial scan statistic. Int J Health Geogr 2004,
3:11.
4. Scott K, Abhinav K, Stanton B, Johnston C, Turner M, Ampong A, Sakel M, Orrell R,
Howard R, Shaw C, Leigh P, Al-Chalabi A: Geogrphical clustering of amyotrophic
lateral sclerosis in south-east England: a population study. Neuroepidemiology 2009,
32:81–88.
5. Klassen A, Kulldorff M, Curriero F: Geographical clustering of prostate cancer
grade and stage at diagnosis, before and after adjustment for risk factors. Int J
Health Geogr 2005, 4:1.
6. Kleinman K, Abrams A, Kulldorff M, Platt R: A model-adjusted space-time scan
statistic with an application to syndromic surveillance. Epidemiol Infect 2005,
133:409–419.
7. Perez AM, Zeng D, Tseng C, Chen H, Whedbee Z, Paton D, Thurmond MC: A web-
based system for near real-time surveillance and space-time cluster analysis of foot-
and-mouth disease and other animal diseases. Prev Vet Med 2009, 91:39–45.
8. Kulldorff M, Athas W, Feuer E, Miller B, Key C: Evaluating cluster alarms: a
space-time scan statistic and brain cancer in Los Alamos. Am J Public Health 1998,
88:1377–1380.
9. Lawson AB, Kleinman K: Spatial and Syndromatic Surveillance for Public Health.
Chichester, West Sussex, Hoboken, NJ: J. Wiley; 2005.
10. Edge VL, Pollari F, Ng LK, Michel P, McEwen SA, Wilson JB, Jerrett M, Sockett
PN, Martin SW: Syndromic surveillance of Norovirus using over-the-counter sales of
medications related to gastrointestinal illness. Can J Infect Dis Med Microbiol 2006,
17:235–241.
11. Hope K, Durrheim DN, Muscatello D, Merritt T, Zheng W, Massey P, Cashman P,
Eastwood K: Identifying pneumonia outbreaks of public health importance: can
110
emergency department data assist in earlier identification? Aust N Z J Public Health
2008, 32:361–364.
12. van Dijk A, McGuinness D, Rolland E, Moore KM: Can telehealth Ontario
respiratory call volume be used as a proxy for emergency department respiratory
visit surveillance by public health? CJEM 2008, 10:18–24.
13. Vourc’h G, Bridges VE, Gibbens J, De Groot BD, McIntyre L, Poland R, Barnouin J:
Detecting emerging diseases in farm animals through clinical observations. Emerg
Infect Dis 2006, 12:204–210.
14. Van Metre DC, Barkey DQ, Salman MD, Morley PS: Development of a syndromic
surveillance system for detection of disease among livestock entering an auction
market. JAVMA 2009, 234:658–664.
15. Vilas VJDR, Bohning D, Kuhnert R: A comparison of the active surveillance of
scrapie in the European Union. Vet Res 2008, 39:37–52.
16. Weber WD: Development of an animal health monitoring system based on slaughter
condemnation data. Miami: Proceedings of the Eighth International Society for Disease
Surveillance Conference; 2009:3–4. December.
17. Alton GD, Pearl DL, Bateman KG, McNab WB, Berke O: Factors associated with
whole carcass condemnation rates in provincially-inspected abattoirs in Ontario
2001–2007: implications for food animal syndromic surveillance. BMC Vet Res 2010,
6:42.
18. Kulldorff M, Heffernan R, Hartman J, Assunção R, Mostashari F: A space-time
permutation scan statistic for disease outbreak detection. PLoS Med 2005, 2:216–
224.
19. Alton GD, Pearl DL, Bateman KG, McNab WB, Berke O: Suitability of portion
condemnations at Ontario provincially-inspected abbatoirs for food animal
syndromic surveillance. BMC Vet Res 2012, 8:88.
20. Huang L, Tiwari R, Zou Z, Kulldorff M, Feuer E: Weighted normal spatial scan
statistic for heterogeneous population data. J Am Stat Assoc 2009, 104(487):886–898.
111
Table 4.1 - Results of the space-time scan statistic using “parasitic liver” condemnation data
Date of
cluster
No. of
abattoirs in
cluster
Latitude
(°N)
Longitude
(°W) Radius (Km) No. of cases
Relative
risk1
P-value
a) Unadjusted Poisson space-time scan test
Jul 2004 - Dec
2007 16 44.52 80.92 62.45 12907 3.66 < 0.001
Mar 2002 -
Aug 2005 26 43.60 79.81 62.52 6385 0.36 < 0.001
b) Animal-adjusted Poisson space-time scan test
Dec 2006 -
Dec 2007 14 44.53 80.99 52.98 4973 3.90 < 0.001
Dec 2001 -
May 2005 51 42.91 79.50 117.97 8732 0.45 < 0.001
Aug 2003 -
Jan 2007 69 44.36 76.35 255.37 13849 1.59 < 0.001
c) Model-adjusted space-time scan test
Mar 2001 - 1 43.90 81.31 0 420 4.91 < 0.001
112
Date of
cluster
No. of
abattoirs in
cluster
Latitude
(°N)
Longitude
(°W) Radius (Km) No. of cases
Relative
risk1
P-value
Jun 2002
Sep 2003 -
May 2004 71 43.18 80.51 98.22 4576 1.52 < 0.001
Sep 2001 -
Dec 2001 16 43.33 79.98 46.01 1335 2.19 < 0.001
Mar 2004 -
Nov 2006 2 44.11 80.80 0 2338 0.62 < 0.001
Jan 2001 -
Mar 2001 72 44.40 77.61 197.95 496 0.49 < 0.001
Mar 2005 -
Jan 2006 1 44.36 76.35 0 56 0.18 < 0.001
d) Space-time permutation scan test
Jan 2001 - Sep
2002 2 43.65 79.86 4 2335 6.76 < 0.001
Jan 2001 - Sep
2003 7 44.13 80.61 29.42 341 0.11 < 0.001
113
Date of
cluster
No. of
abattoirs in
cluster
Latitude
(°N)
Longitude
(°W) Radius (Km) No. of cases
Relative
risk1
P-value
Sec 2006 -
Dec 2007 2 44.11 80.80 0 4431 3.06 < 0.001
e) Space-time scan test of residuals
No statistically significant clusters found
1The values in this column for the space-time permutation model refer to the observed/expected value.
114
Table 4.2 - Results of the space-time scan statistic using pneumonic lung condemnation data
Date of cluster
No. of
abattoirs in
cluster
Latitude
(°N)
Longitude
(°W)
Radius
(Km) No. of cases
Relative
risk1
P-value
a) Unadjusted Poisson space-time scan test
Jan 2001 - Jun 2004 5 43.94 79.69 26.60 14896 4.98 < 0.001
Jun 2004 - Nov 2007 63 49.75 92.96 1149.60 54 0.01 < 0.001
b) Animal-adjusted Poisson space-time scan test
Jan 2001 - Jun 2004 5 43.96 79.69 26.60 14896 3.12 < 0.001
Mar 2004 - Aug 2007 21 43.01 79.97 49.09 61 0.01 < 0.001
Oct 2001 - Mar 2005 1 42.98 80.60 0 4466 5.14 < 0.001
c) Model-adjusted space-time scan test
Mar 2004 - Aug 2007 16 43.01 79.97 36.88 60 0.07 < 0.001
Jan 2001 - Mar 2003 1 43.91 79.82 0 1558 0.44 < 0.001
Jan 2001- Mar 2002 4 43.30 79.90 19.98 477 6.06 < 0.001
115
Table 4.2 - Results of the space-time scan statistic using pneumonic lung condemnation data
Date of cluster
No. of
abattoirs in
cluster
Latitude
(°N)
Longitude
(°W)
Radius
(Km) No. of cases
Relative
risk1
P-value
Apr 2005 - Nov 2007 4 43.96 79.19 20.93 273 0.26 < 0.001
d) Space-time permutation scan test
Jan 2001 - Feb 2003 11 44.14 79.94 28.20 1520 0.31 < 0.001
Jan 2001 - Feb 2003 32 42.95 79.10 121.36 10514 1.55 < 0.001
Feb 2005 - Dec 2007 10 43.96 79.19 38.19 363 0.20 < 0.001
e) Space-time scan test of residuals
Jan 2001 - Mar 2003 1 43.91 79.82 0 110 --- < 0.001
Jan 2001 – Feb 2003 8 43.17 79.73 26.91 370 --- < 0.001
Jul 2003 - Jun 2004 6 43.65 79.86 29.33 109 --- < 0.001
Jan 2001 - Mar 2003 36 43.99 80.67 68.49 2280 --- < 0.001
1The values in this column for the space-time permutation model refer to the observed/expected value.
116
Figure 4.1 - Results of space-time scan statistic using “parasitic liver” portion
condemnation data from Ontario provincial abattoirs 2001 – 2007 using four
approaches for covariate adjustment compared to the unadjusted data including: A)
Unadjusted Poisson space-time scan test, B) Animal-adjusted Poisson space-time
scan test, C) Model-adjusted space-time scan test and D) Space-time permutation
scan test
117
Figure 4.2 - Space-time scan statistic using pneumonic lung condemnation data from
Ontario provincial abattoirs 2001–2007 using four approaches for covariate
adjustment compared to the unadjusted data including: A) Unadjusted Poisson
space-time scan test, B) Animal-adjusted Poisson space-time scan test, C) Model-
adjusted space-time scan test and D) Space-time permutation scan test and e) Space-
time scan test of residuals
118
CHAPTER FIVE:
SUITABILITY OF SENTINEL ABATTOIRS FOR SYNDROMIC
SURVEILLANCE USING PROVINCIALLY-INSPECTED BOVINE ABATTOIR
CONDEMNATION DATA
(Alton et al. as submitted to BMC Veterinary Research, 2014)
ABSTRACT
Sentinel surveillance has previously been used to monitor and identify disease
outbreaks in both human and animal contexts. However, the use of sentinel abattoirs to
monitor provincially-inspected abattoir condemnation rates for food animal syndromic
surveillance has never been explored in Ontario, Canada. Three approaches for the
selection of sentinel sites are proposed and evaluated regarding their ability to capture
respiratory disease trends using provincial abattoir condemnation data for use in a
sentinel syndromic surveillance system. Consideration of how the sentinel abattoirs are
selected is important, as they are meant to represent the disease trends of all provincial
abattoirs. The objectives of this study were the following: (1) determine the suitability of
sentinel abattoirs for food animal syndromic surveillance using provincial abattoir
pneumonic lung portion data as an example; and (2) determine which sentinel abattoir
selection method best represented the pneumonic lung condemnation rate patterns from
all abattoirs operating throughout the study period.
Condemnation data for the bovine pneumonic lung classification from all
provincially-inspected abattoirs in Ontario were extracted from the Food Safety Decision
Support System (FSDSS) maintained by the Ontario Ministry of Agriculture and Food
(OMAF) from 2001 – 2007. Although there were 211 abattoirs processing cattle during
119
the study period, only 98 abattoirs were consistently operating throughout the study
period and were therefore used as the full set of abattoirs on which to make comparisons.
Pneumonic lung condemnation rates were calculated and plotted graphically. The
following three criteria to select sentinel abattoirs were compared to the full data
including: (1) abattoirs processing cattle all weeks of the year; (2) abattoirs processing at
least 6500 cattle per year; and (3) multi-criteria selection approach using data from
abattoirs that met a predefined set of criteria related to abattoir processing capacity,
animal classes represented, and operating over the entire study period. Negative binomial
regression models were used to compare monthly pneumonic lung condemnation rates
from the full set of abattoirs to the three sentinel abattoir selection methods for each
animal class, as well as to compare monthly pneumonic lung condemnation rates for each
sentinel selection approach while controlling for season, animal class and year to
determine if the overall patterns were similar between the sentinel selection approaches.
Descriptive graphs of pneumonic lung condemnation rates under the three sentinel
abattoir selection approaches showed some differences when compared to the
condemnation rates from the full set of abattoirs. However, all three approaches resulted
in the selection of groups of sentinel abattoirs that captured trends in condemnation rates
similar to those reported by the full set of abattoirs. Results of the negative binomial
regression models showed no difference between the 3 sentinel selection approaches and
the full set of abattoirs for calves, however, all selection approaches tended to
overestimate the condemnation rates of the full dataset by 1.4 to as high as 3.8 times for
cows, heifers and steers. The largest overestimation was consistently found with the
selection approach utilizing abattoirs processing at least 6500 cattle per year. In addition,
120
a negative binomial model found that the selection approach using abattoirs open all
weeks had similar temporal trends when compared to the full set of abattoirs.
Sentinel abattoirs show promise for integration into a food animal syndromic
surveillance system using Ontario provincial abattoir condemnation data. The selection of
sentinel abattoirs is extremely important, as they are meant to be representative of the
disease trends in the entire population. While all selection approaches tended to
overestimate the condemnation rates of the full dataset to some degree, the abattoirs open
all weeks selection approach appeared to best capture the overall seasonal and temporal
trends of the full dataset, and differences in condemnation rates among animal classes,
and would be the most suitable approach for sentinel abattoir selection. However, further
studies are needed to examine how the sentinel abattoir selection approaches perform
with simulated and/or historical outbreak data.
121
BACKGROUND
There are various approaches to conduct disease surveillance including sentinel
and syndromic surveillance. Sentinel surveillance is a form of surveillance which
involves a limited number of recruited participants or organizations, such as farms,
veterinarians, abattoirs, healthcare providers or hospitals, which report on certain health
events to give an indication of what may be happening in the general population [43].
Sentinel surveillance is a strategy used to sample timely data in a relatively inexpensive
manner rather than collect information on the general population, when population-based
data collection is unfeasible in a timely or cost-effective manner [44]. Syndromic
surveillance involves the amalgamation of signs/symptoms using data from non-
traditional data sources [46]; the signs/symptoms are grouped into classifications called
‘syndromes’ and are used to track disease trends in populations and signal a possible
outbreak that warrants further investigation [46].
Sentinel surveillance has been previously used in both human and animal health
settings for a variety of disease outcomes. Sentinel surveillance has been used to monitor
or identify outbreaks of infectious diseases and to monitor the activity of certain health
conditions which can change due to environmental conditions. Though used less often in
animal health applications than in human health, sentinel surveillance has been used
successfully for surveillance in various applications. For example, following the
emergence of Bluetongue virus serotype 8 in Central Europe in 2006, causing a large
scale outbreak in 2007 in several countries in Europe, a Bluetongue sentinel surveillance
program was established in Belgium in 2010. This surveillance program was intended to
demonstrate the absence of Bluetongue virus [70]. This program randomly selected a
122
total of 300 dairy herds, with 30 herds selected from each of the Belgian provinces. The
criteria for selection of herds was based on dairy herds that were expected to have a
minimum of 15 animals present between 4 and 12 months of age at the start of the
sentinel program [56]. Other studies have combined both sentinel and syndromic
surveillance and utilize data from sentinel veterinarians or veterinary practices [7, 49].
Animal health data including abattoir condemnation data have emerged recently
as a novel data stream for syndromic surveillance of diseases of animal and public health
importance [7, 9, 24, 53, 69, 71, 72]. Ontario provincial abattoir data have been recently
explored as a potential source of information for food animal syndromic surveillance [24,
69, 71]. However, there are approximately 100-150 provincial abattoirs in Ontario, many
of which are open sporadically and have differing processing capacities. It may be
beneficial to conduct targeted surveillance at select abattoirs in order to gather more
accurate data for syndromic surveillance in a cost-effective manner; sentinel surveillance
can be a cost-effective solution for disease surveillance. In addition, by selecting specific
sentinel abattoirs for inclusion in a sentinel syndromic surveillance system, it would
allow for more intensive and specialized training of inspectors for syndromic
surveillance. Furthermore, if sentinel abattoirs were selected properly, they could help to
reduce the cost of surveillance while still being representative of Ontario provincial
abattoirs. We proposed three sentinel site selection approaches and compared their ability
to detect respiratory disease trends in bovine abattoir condemnation data for use in a
sentinel-based syndromic surveillance system.
Pneumonic lung condemnation rates from all Ontario abattoirs processing cattle
throughout the 2001-2007 period were compared to those collected from sentinel sites
123
based on three selection approaches: (1) abattoirs processing cattle all weeks of the year;
(2) abattoirs processing at least 6500 cattle per year (based on data from abattoirs in the
upper 95th
percentile in processing capacity); and (3) multi-criteria selection approach of
abattoirs that met a predefined set of criteria related to abattoir processing capacity, and
animal classes represented.
The goal of the study was to design a sentinel surveillance system based on
provincially inspected abattoirs in Ontario. Specific objectives of this study were the
following: (1) determine the suitability of sentinel abattoirs for food animal syndromic
surveillance using provincial abattoir pneumonic lung condemnation data as an example;
and (2) determine which design is most efficient and representative for pneumonic lung
condemnation rate in terms of spatial distribution, temporal trends and relative
differences between animal classes when compared to data from all the abattoirs over the
study period.
METHODS
Data source
Pneumonic lung portion condemnation data were obtained from the Food Safety
Decision Support System (FSDSS) database maintained by the Ontario Ministry of
Agriculture, Food and Rural Affairs (OMAFRA). The database contains information
regarding the number and reason for daily organs/body systems condemnations in
provincially inspected abattoirs in Ontario. The condemnation category of pneumonic
lung was selected for these analyses, as this category represents a major health issue for
beef cattle and was among the most frequently reported portion condemnations by
provincial inspectors during the study period. Pneumonic lung condemnation refers to
124
bovine lungs which were condemned for lesions indicative of a previous localized and
resolved antero-ventral pneumonia infection (personal communication Abdul Rehmtulla,
DVM, OMAFRA, Stone Road, Guelph, Ontario).
Data were extracted from the Food Safety Decision Support System (FSDSS)
database for cattle animal classes: bulls, calves, cows, heifers, and steers from January 1,
2001 to December 31, 2007. Missing geographical coordinates for abattoirs were
approximated using postal codes and/or addresses with the address geocoding software
GeoPinpoint Suite 6.4 (DMTI Spatial Inc., Markham, Ontario, Canada). Using the
FSDSS database, further variables were created for each month: geographical coordinates
of abattoir, year, season, number of weeks an abattoir was operating each year, total
number of pneumonic lung condemnations, total number of cattle processed each year,
and animal class. Animal class included five categories: bulls, cows, calves, heifers, and
steers. Bulls were excluded from subsequent analyses due to missing data and
inconsistencies in the use of this classification. The number of weeks an abattoir was
operating each year was determined by the total number of weeks in which at least one
bovine animal was processed. The total number of animals processed each year was
calculated from the total number of condemned cattle plus the number of cattle fit for
consumption. The agricultural region where an abattoir was located was classified as:
central, eastern, northern, southern or western Ontario using the Ontario Census
Agricultural Region boundaries (Statistics Canada, Census Agricultural Regions, Census
year 2001). The regional location of each abattoir was determined using the point-in-
polygon technique with geographic information system software ArcGIS 9.2 (ESRI,
Redlands, California, USA). Abattoirs were excluded from subsequent analyses if the
125
abattoir did not slaughter at least 1 bovine animal for each year of the study period (2001
– 2007).
Descriptive analyses
Monthly bovine pneumonic lung condemnation rates were calculated using data
from all abattoirs slaughtering cattle during the study period of January 1, 2001 –
December 31, 2007. Condemnation rates were calculated by dividing the total number of
pairs of lungs condemned under the pneumonic lung classification each month by the
total number of slaughtered bovines for each animal class (e.g., calves, cows, heifers and
steers).
The number of abattoirs in operation during the study period varied from year to
year. There were 98 abattoirs consistently in operation during the study period and were
therefore used to represent the full set of abattoirs for this study. Three different design
approaches for a sentinel syndromic surveillance system were compared to all data from
the full set of 98 abattoirs. The first sentinel selection approach, which we refer to as
abattoirs open all weeks, uses data from abattoirs processing cattle 52 weeks per year.
The second approach, which we refer to as large abattoirs; uses data from abattoirs in the
upper 95th
percentile in processing capacity (processing at least 6500 cattle per year). The
third approach, which we refer to as multi-criteria approach, uses data from abattoirs that
met the following criteria for each year of the study period: processed at least 499 cattle
per year (representing the median total number of cattle processed based on data from all
211 abattoirs open from 2001 – 2007), processed at least 1 bovine carcass 44 weeks or
more each year (representing the median number of weeks cattle were processed based
126
on data from all 211 abattoirs open from 2001 - 2007), and processed cattle representing
all animal classes (calves, cows, heifers, and steers).
Data obtained using the three design approaches were summarized graphically
using the monthly condemnation rates of each animal class, as well as boxplots of the
rates. In addition, data were summarized in terms of the number of abattoirs included in
each selection approach, percentage of shared abattoirs between each selection approach,
and geographical representativeness of each selection approach according to the
distribution of abattoirs among census agricultural regions in Ontario.
Statistical analyses
A univariable negative binomial model was used to compare monthly pneumonic
lung condemnation rates from all sentinel site selection approaches and the full set of
abattoirs using a categorical variable for each of the 3 sentinel design approaches.
In addition, a separate multi-level negative binomial regression model was used with a
random intercept for abattoir using the xtnbreg command in Stata. This model allows the
random effect to follow a beta distribution and for the overdispersion parameter to vary
by abattoir. The negative binomial model was used to evaluate the association of monthly
condemnation rates of pneumonic lungs for each sentinel selection approach with year,
season and animal class. All covariates were evaluated for statistical significance
individually and then in a multivariable multi-level negative binomial model. Wald tests
were used to determine the significance of covariates. All covariates were forced into the
models regardless of univariable significance, due to the identification of these variables
127
as important predictors for pneumonic lung condemnation rates in a previous study [69]
and to evaluate overall temporal trends in the data.
For all regression models, the decision to use a negative binomial model instead
of a Poisson regression model was based on evaluating the Akaike Information Criterion
(AIC) value of both models and the significance of the over-dispersion term of the
negative binomial model. The offset of the negative binomial model was the natural log
of the total number of slaughtered cattle for the abattoirs included for each sentinel site
selection approach for each animal class and month-year period during the study period.
All statistical analyses were conducted using Stata 12 (Stata Corp., College Station,
Texas, USA).
RESULTS
Descriptive Statistics
There were a total of 211 provincial abattoirs slaughtering a total of 1,155,535
cattle from 2001-2007. However, there were only 98 abattoirs that remained in operation
for the entire study period and were used to represent the full set of abattoirs in this study.
This number is consistent with the current number of abattoirs processing cattle in
Ontario in 2014 [14]. During the study period there were 36,883 lungs condemned
representing approximately 9% of all portion condemnation and among the most
frequently reported condemnation in cattle. Further discussion of these data can be found
in a previous study by Alton et al.[69].
Three approaches for sentinel abattoir selection were compared to pneumonic
lung condemnation rates for each animal class from the full set of abattoirs (Table 5.1).
128
Abattoirs for each sentinel selection approach were chosen from a total of 98 abattoirs
processing cattle for the entire study period. The number of abattoirs selected for each
sentinel selection approach varied from 7 to 45. In assessing the percentage of overlap of
the selected abattoirs among the sentinel site selection approaches, the percentage of
shared abattoirs ranged from approximately 7% to 71%, with abattoirs open all weeks
and multi-selection criteria approaches having the highest percentage of abattoirs in
common (Table 5.2). All selection approaches led to surveillance systems based on
abattoirs representing all census agricultural regions across Ontario with the exception of
using the large abattoir criterion, which did not include abattoirs from eastern and
northern Ontario (Table 5.3).
The temporal condemnation rate graphs for all data indicate a gradual decrease in
pneumonic lung condemnation rates over time in calves from approximately 100
condemnations per 1000 slaughtered calves in 2001 to approximately 20 condemnations
per 1000 slaughtered calves in 2007 (Figure 5.1). In comparison, pneumonic lung
condemnation rates in cows, heifers and steers remained much more stable over the study
period (Figure 5.1). Boxplots of the monthly condemnation rates for the full set of
abattoirs compared to the 3 sentinel site selection approaches for each animal class had
similar distributions for calves, heifers and steers (Figure 5.2), with the exception of the
large abattoir selection approach, which tended to be have more variability when
compared to the other selection approaches and full set of abattoirs.
When pneumonic lung condemnation rates from the 3 sentinel site selection
approaches were graphically compared to condemnation rates from the full set of data,
some variation can be observed between the approaches, although similar overall trends
129
were present (Figures 5.3 – 5.6). Based on descriptive plots for calves (Figure 5.3), the
multi-criteria selection approach did not show a good fit with the full dataset, with
alternating trends of either overestimating or underestimating the condemnation rate of
the full dataset. However, the other two alternative selection approaches tended to
approximate the overall secular trends of the full dataset, but generally overestimated the
condemnation rates of the full dataset throughout the study period. Based on the
descriptive plots for cows, heifers and steers (Figures 5.4 – 5.6), all three design
approaches tended to overestimate the condemnation rate relative to the full set of data,
however, the large abattoir selection approach had the largest overestimation in all animal
classes.
Negative binomial models
We used a negative binomial regression model to compare the monthly
pneumonic lung condemnation rates from the 3 sentinel surveillance system selection
approaches for each animal class to the full dataset. The model involving calves (Table
5.4a) found a difference between the large abattoir selection approach and the full set of
data. In comparison, negative binomial modelling of cow data (Table 5.4b) found, that all
3 design approaches overestimated the condemnation rate by a factor of 2 to 4 times,
when compared to the full dataset. Similarly, relative to the full dataset, all the sentinel
selection approaches overestimated the condemnation rates for heifers (Table 5.4c) and
steers (Table 5.4d).
There were similar seasonal and temporal trends noted among all sentinel site
selection approaches and the full set of abattoirs (Tables 5.5a – d). However, the abattoirs
open all weeks sentinel selection approach tended to have coefficients and trends which
130
best resembled the seasonal and temporal trends of the full dataset. Abattoirs open all
weeks also tended to have the closest approximation of coefficients when compared to
the full dataset for the animal class variable.
DISCUSSION
Three sentinel site selection approaches for a sentinel syndromic surveillance
system using abattoir condemnation data were proposed and compared to the full set of
provincially inspected abattoirs in Ontario from 2001 - 2007. While all selection
approaches tended to overestimate the condemnation rates of the full dataset to some
degree, the abattoirs open all weeks selection approach appeared to best capture the
overall seasonal and temporal trends of the full dataset and would be the most suitable
approach for sentinel abattoir selection. This selection approach utilizes data from 45
abattoirs. It may be advantageous to conduct enhanced surveillance at carefully selected
sentinel abattoirs rather than at the approximately 100 abattoirs slaughtering cattle in
Ontario [14]. This allows for more intensive and specialized training of inspectors for
syndromic surveillance. Furthermore, if sentinel abattoirs were selected properly, sentinel
abattoirs could help to reduce the cost of surveillance while still being representative of
Ontario provincial abattoirs.
While there was not one design approach that perfectly fit the full dataset in all
analyses, the results of the descriptive and quantitative analyses found a fairly good fit
between abattoirs open all weeks and the full dataset. This sentinel selection approach
utilizes less than half of the abattoirs from the full set of abattoirs. By collecting data
from fewer abattoirs, the cost of data collection and analysis is reduced. This approach
also allows for the use of targeted training of inspectors to reduce any presence of
131
inspector bias on the data, which appeared to be a possible issue with these data based on
previous studies [24, 71]. The multi-criteria sentinel selection approach also utilizes data
from approximately the same number of abattoirs as the abattoirs open all weeks,
however, this approach had larger amount of overestimation for heifers and steers than
abattoirs open all weeks and did not approximate the overall seasonal, secular and animal
class condemnation trends of the full dataset as well as the abattoirs open all weeks. This
suggests that the number of abattoirs selected is not necessarily as important as the
criteria used to select the abattoirs. While the abattoirs open all weeks sentinel selection
approach drastically reduced the number of abattoirs needed to conduct syndromic
surveillance involving provincial abattoir condemnation data compared to the full dataset,
it is uncertain whether this reduction is sufficient and manageable to conduct intensive
and targeted surveillance. Further research is needed to determine if there are other
sentinel selection approaches that could reduce this number of abattoirs even further.
Geographical representativeness is also an important consideration when
establishing sentinel selection criteria. While the sentinel selection approaches based on
abattoirs that are open all weeks and multi-criteria are geographically representative, the
large abattoir selection approach is geographically over-representative of abattoirs in
central and southern Ontario and has no representation of abattoirs in northern and
eastern Ontario. This could pose an issue for syndromic surveillance, as abattoirs
generally receive animals from relatively local farms [24], and a lack of representation
from abattoirs in these regions could lead to the inability to identify emerging health
issues in these under-represented areas. The distribution of abattoirs among regions is
almost identical for abattoirs open all weeks selection approach and the full dataset. In
132
addition, region has been shown to be an important variable associated with partial and
whole carcass condemnations [24, 69] and excluding certain regions could bias the
results of spatio-temporal cluster detection methods for syndromic surveillance involving
these data.
This study found bovine provincial abattoir condemnation data to be suitable for
sentinel surveillance assuming it is a cost-effective option. However, this study did not
investigate outbreak detection, as there were no documented animal health outbreaks in
cattle during the study period. Consequently, further research is needed to assess the
effectiveness of sentinel selection approaches for detecting emerging health issues using
documented or simulated outbreak data of both respiratory diseases as well as other
disease syndromes important to both animal health and food safety.
CONCLUSIONS
Sentinel abattoirs show promise for integration into a food animal syndromic
surveillance system using Ontario provincial abattoir condemnation data. The selection of
the sentinel abattoirs is extremely important in order to maintain representativeness of the
disease trends in the entire population. The information extracted from surveillance
systems varied by its respective selection approach and for each animal class. While all
selection approaches tended to overestimate the condemnation rates of the full dataset to
some degree, the abattoirs open all weeks selection approach appeared to best capture the
overall seasonal and temporal respiratory condemnation trends of the full dataset and
would be most suitable approach for sentinel abattoir selection involving these data.
Further studies should examine the performance of the proposed sentinel selection
approaches using simulated data that include disease outbreaks for both respiratory
133
diseases as well as other condemnation syndromes relevant to animal health and food
safety.
ACKNOWLEDGEMENTS
The authors would also like to acknowledge the following organizations for their
support for infrastructure, data retrieval and funding for this project: Canada Foundation
for Innovation (CFI), the Ontario Research Fund, the Ontario Graduate Scholarship, and
Ontario Ministry of Agriculture & Food (OMAF).
134
REFERENCES
1. Salman MD. Animal Disease Surveillance and Survey Systems: /methods and
Applications: Ames, Iowa: Iowa State Press; 2003.
2. Lee LM, Teutsch SM, Thacker SB, St. Louis ME. Principles and Practice of Public
Health Surveillance: 3rd
Ed.New York: Oxford University Press; 2010.
3. Lawson AB, Kleinman K: Spatial and Syndromic Surveillance for Public Health:
Chichester, West Sussex ; J. Wiley; 2005.
4. Vangeel I, De Leeuw I, Meroc E, Vandenbussche F, Riocreux F, Hooyberghs J,
Raemaekers M, Houdart P, Van dS, De Clercq K: Bluetongue sentinel surveillance
program and cross-sectional serological survey in cattle in Belgium in 2010-2011.
Prev Vet Med 106:235-243.
6. Amezcua M, Pearl D, Friendship R, McNab W: Evaluation of a veterinary-based
syndromic surveillance system implemented for swine. CJVR 2010, 74:241-250.
7. Dórea F, Sanchez J, Revie CW: Veterinary syndromic surveillance: current
initiatives and potential for development. Prev Vet Med 2011, 101:1-17.
8. Alton GD, Pearl DL, Bateman KG, McNab WB, Berke O: Suitability of portion
condemnations at Ontario provincially-inspected abbatoirs for food animal
syndromic surveillance. BMC Vet Res 2012, 8:88.
9. Alton GD, Pearl DL, Bateman KG, McNab WB, Berke O: Factors associated with
whole carcass condemnation rates in provincially-inspected abattoirs in Ontario
2001-2007: implications for food animal syndromic surveillance. BMC Vet Res 2010,
6:42.
10. Thomas-Bachli AL, Pearl DL, Friendship RM, Berke O. Suitability and limitations
of portion-specific abattoir data as part of an early warning system for emerging
diseases of swine in Ontario. BMC Vet Res 2012, 6:3.
11. Weber WD: Development of an animal monitoring system based on slaughter
condemnation data. Miami: Proceedings of the Eighth International Society for Disease
Surveillance Conference; 2009. 3-4 December.
12. O'Sullivan T, Friendship RM, Pearl DL, McEwen B, Dewey CE: Identifying an
outbreak of a novel swine disease using test requests for porcine reproductive and
respiratory syndrome as a syndromic surveillance tool. BMC Vet Res 2012, 8:192.
135
13. Palmer S, Sully M, Fozdar F: Farmers, Animal Disease Reporting And The Effect
Of Trust: A Study Of West Australian Sheep And Cattle Farmers. Rural Society
2009, 19:32-48.
14. Provincially Licensed Meat Plants
[http://www.omafra.gov.on.ca/english/food/inspection/meatinsp/licenced_operators_list.h
tm]
136
Table 5.1 – Summary of sentinel abattoirs selection approaches for Ontario provincial abattoirs (2001 – 2007)
Sentinel site
selection approach
Selection description Number of abattoirs
Full dataset Abattoirs processing at least 1 bovine carcass each year of study period
(2001 – 2007)
98
Abattoirs open all
weeks
Abattoirs processing at least 1 bovine carcass 52 weeks per year 45
Multi-criteria Selection of abattoirs meeting the following criteria:
processed at least 499 cattle per year, processed at least 1 bovine carcass
44 weeks or more per year, processed cattle representing all animal
classes (calves, cows, heifers and steers)
44
Large abattoirs Abattoirs processing ≥ 6500 cattle per year 7
137
Table 5.2 – Percentage of shared abattoirs between sentinel selection approaches1
Selection
approaches
Full dataset2
(N=98)
Abattoirs open
all weeks3
(N=45)
Selection
criteria4
(N=44)
Large
abattoirs5
(N=7)
Full dataset2
(N=98)
100% (98) 45.9% (45) 44.9% (44) 7.1% (7)
Abattoirs open
all
weeks3(N=45)
45.9% (45) 100% (45) 71.1% (32) 15.6% (7)
Selection
criteria4 (N=44)
44.9% (44) 71.1% (32) 100% (44) 6.8% (3)
Large
abattoirs5
(N=7)
7.1% (7) 15.6% (7) 6.8% (3) 100% (7)
1To calculate the percentage of shared abattoirs, the largest number of abattoirs was used as the
denominator 2Abattoirs processing at least 1 bovine carcass each year of study period (2001 – 2007)
3Abattoirs processing at least 1 bovine carcass 52 weeks per year
4Selection of abattoirs meeting the following criteria: processed at least 499 cattle per year,
processed at least 1 bovine carcass 44 weeks or more per year, processed cattle representing all
animal classes (calves, cows, heifers and steers) 5Abattoirs processing ≥ 6500 cattle per year
138
Table 5.3 – Distribution of provincial abattoirs 2001 – 2007 among census agricultural
regions in Ontario based on three sentinel selection approaches
Selection approach Number of
abattoirs
Central
Ontario
Eastern
Ontario
Northern
Ontario
Southern
Ontario
Western
Ontario
Full dataset1 98 18% (18) 15% (15) 7% (7) 36% (35) 23% (23)
Abattoirs open all
weeks2
45 18% (8) 13% (6) 7% (3) 38% (17) 24% (11)
Multi-criteria3 44 9% (4) 18% (8) 5% (2) 39% (17) 30% (13)
Large abattoirs4 7 29% (2) 0% (0) 0% (0) 43% (3) 29% (2)
1Abattoirs processing at least 1 bovine carcass each year of study period (2001 – 2007)
2 Abattoirs processing at least 1 bovine carcass 52 weeks per year
3 Selection of abattoirs meeting the following criteria: processed at least 499 cattle per year,
processed at least 1 bovine carcass 44 weeks or more per year, processed cattle representing all
animal classes (calves, cows, heifers and steers) 4 Abattoirs processing ≥ 6500 cattle per year
139
Table 5.4 – Negative binomial regression models comparing monthly pneumonic lung
condemnation rates for all abattoirs for each animal class with three methods of sentinel
abattoir selection
Sentinel selection method IRR 95% CI P-value
a) Calves
Full dataset1 --- --- ---
Abattoirs open all weeks2 1.12 0.95 – 1.32 0.16
Multi-criteria3 0.90 0.76 – 1.06 0.21
Large abattoirs4 1.25 1.06 – 1.47 0.01
b) Cows
Full dataset1 --- --- ---
Abattoirs open all weeks2 2.08 1.49 – 2.91 < 0.01
Multi-criteria3 2.06 1.47 – 2.89 < 0.01
Large abattoirs4 3.83 2.73 – 5.38 < 0.01
c) Heifers
Full dataset1 --- --- ---
Abattoirs open all weeks2 1.61 1.35 – 1.94 < 0.01
Multi-criteria3 1.70 1.42 – 2.04 < 0.01
Large abattoirs4 2.86 2.38 – 3.43 < 0.01
d) Steers
Full dataset1 --- --- ---
Abattoirs open all weeks2 1.39 1.20 – 1.61 < 0.01
Multi-criteria3 1.48 1.28 – 1.72 < 0.01
Large abattoirs4 2.02 1.74 – 2.34 < 0.01
1Abattoirs processing at least 1 bovine carcass each year of study period (2001 – 2007)
2 Abattoirs processing at least 1 bovine carcass 52 weeks per year
3 Selection of abattoirs meeting the following criteria: processed at least 499 cattle per year,
processed at least 1 bovine carcass 44 weeks or more per year, processed cattle representing all
animal classes (calves, cows, heifers and steers) 4 Abattoirs processing ≥ 6500 cattle per year
140
Table 5.5 – Multivariable multilevel negative binomial regression model examining seasonal
and annual variability in monthly pneumonic lung condemnations for three sentinel
selection approaches
Model IRR 95% CI P-value
a) Full dataset1
Winter --- ---- ---
Spring 0.97 0.85 – 1.11 0.61
Summer 0.90 0.78 – 1.03 0.12
Fall 0.89 0.78 – 1.03 0.11
2001 --- --- ---
2002 0.93 0.79 – 1.08 0.34
2003 0.99 0.85 – 1.16 0.93
2004 0.53 0.44 – 0.63 < 0.01
2005 0.45 0.38 – 0.54 < 0.01
2006 0.39 0.32 – 0.47 < 0.01
2007 0.32 0.26 – 0.40 < 0.01
Calves --- --- ---
Cows 1.05 0.87 – 1.27 0.60
Heifers 0.68 0.57 – 0.80 < 0.01
Steers 0.75 0.66 – 0.85 < 0.01
b) Abattoirs open all weeks2
Winter --- --- ---
Spring 1.01 0.88 – 1.16 0.93
Summer 0.92 0.80 – 1.05 0.22
Fall 0.91 0.79 – 1.05 0.22
2001 --- --- ---
2002 0.94 0.80 – 1.11 0.50
2003 1.03 0.88 – 1.21 0.68
2004 0.57 0.48 – 0.68 < 0.01
2005 0.46 0.38 – 0.55 < 0.01
2006 0.40 0.38 – 0.49 < 0.01
2007 0.34 0.28 – 0.42 < 0.01
Calves --- --- ---
Cows 1.03 0.84 – 1.26 0.75
Heifers 0.73 0.62 – 0.87 < 0.01
Steers 0.75 0.66 – 0.86 < 0.01
c) Multi-criteria3
Winter --- --- ---
Spring 1.08 0.92 – 1.28 0.33
Summer 0.95 0.80 – 1.12 0.51
Fall 0.92 0.78- 1.09 0.35
2001 --- --- ---
2002 1.03 0.84 – 1.27 0.76
2003 1.30 1.11 – 1.63 < 0.01
2004 1.04 0.85 – 1.28 0.72
2005 0.67 0.54 – 0.84 < 0.01
2006 0.67 0.49 – 0.79 < 0.01
141
2007 0.62 0.62 – 0.96 < 0.01
Calves --- --- ---
Cows 0.77 0.62 – 0.96 0.02
Heifers 0.62 0.51 – 0.75 < 0.01
Steers 0.71 0.61 – 0.82 < 0.01
d) Large abattoirs4
Winter --- --- ---
Spring 0.97 0.82 – 1.15 0.74
Summer 0.92 0.77 – 1.09 0.32
Fall 0.99 0.84 – 1.18 0.94
2001 --- --- ---
2002 1.38 1.11 – 1.72 < 0.01
2003 1.56 1.26 – 1.93 < 0.01
2004 0.97 0.77 – 1.22 0.78
2005 0.75 0.59 – 0.94 0.01
2006 0.64 0.50 – 0.82 < 0.01
2007 0.64 0.50 – 0.82 < 0.01
Calves --- --- ---
Cows 1.22 0.92 – 1.62 0.16
Heifers 1.02 0.83 – 1.25 0.86
Steers 0.87 0.74 – 1.03 0.10 1Abattoirs processing at least 1 bovine carcass each year of study period (2001 – 2007)
2 Abattoirs processing at least 1 bovine carcass 52 weeks per year
3 Selection of abattoirs meeting the following criteria: processed at least 499 cattle per year,
processed at least 1 bovine carcass 44 weeks or more per year, processed cattle representing all
animal classes (calves, cows, heifers and steers) 4 Abattoirs processing ≥ 6500 cattle per year
142
Figure 5.1 – Pneumonic lung condemnation rates per 1000 slaughtered cattle for each animal class for all abattoirs processing cattle
throughout the study period
05
01
00
150
05
01
00
150
2000
/07
2001
/07
2002
/07
2003
/07
2004
/07
2005
/07
2006
/07
2007
/07
2000
/07
2001
/07
2002
/07
2003
/07
2004
/07
2005
/07
2006
/07
2007
/07
CALVES COWS
HEIFERS STEERS
Co
nd
em
na
tio
n r
ate
per
100
0 s
laug
hte
red
ca
ttle
Year/month
143
Figure 5.2 – Boxplots comparing the pneumonic lung condemnation rates per 1000 slaughtered cattle for each sentinel abattoir selection
approach
0 50 100 150 200 0 50 100 150 200
CALVES COWS
HEIFERS STEERS
Full dataset (N=98) Abattoirs open all weeks (N=45)
Multi-criteria (N=44) Large abattoirs (N=7)
144
Figure 5.3 – Comparison of pneumonic lung condemnation rates in calves for sentinel site selection approaches and full dataset 2001 -
2007
05
01
00
150
Co
nd
em
na
tio
n r
ate
per
100
0 s
laug
hte
red
ca
ttle
2001
/01
2002
/01
2003
/01
2004
/01
2005
/01
2006
/01
2007
/01
2008
/01
Year/month
Full dataset (N=98) Abattoirs open all weeks (N=45)
Multi-criteria (N=44) Large abattoirs (N=7)
145
Figure 5.4 - Comparison of pneumonic lung condemnation rates in cows for sentinel site selection approaches and full dataset 2001 - 2007
05
01
00
150
200
Co
nd
em
na
tio
n r
ate
per
100
0 s
laug
hte
red
ca
ttle
2001
/01
2002
/01
2003
/01
2004
/01
2005
/01
2006
/01
2007
/01
2008
/01
Year/month
Full dataset (N=98) Abattoirs open all weeks (N=45)
Multi-criteria (N=44) Large abattoirs (N=7)
146
Figure 5.5 - Comparison of pneumonic lung condemnation rates in heifers for sentinel site selection approaches and full
dataset 2001 - 2007
05
01
00
150
Co
nd
em
na
tio
n r
ate
per
100
0 s
laug
hte
red
ca
ttle
2001
/01
2002
/01
2003
/01
2004
/01
2005
/01
2006
/01
2007
/01
2008
/01
Year/month
Full dataset (N=98) Abattoirs open all weeks (N=45)
Multi-criteria (N=44) Large abattoirs (N=7)
147
Figure 5.6 - Comparison of pneumonic lung condemnation rates in steers for sentinel site selection approaches and full dataset 2001 - 2007
05
01
00
150
Co
nd
em
na
tio
n r
ate
per
100
0 s
laug
hte
red
ca
ttle
2001
/01
2002
/01
2003
/01
2004
/01
2005
/01
2006
/01
2007
/01
2008
/01
Year/month
Full dataset (N=98) Abattoirs open all weeks (N=45)
Multi-criteria (N=44) Large abattoirs (N=7)
148
CHAPTER SIX:
SUMMARY, DISCUSSION, CONCLUSIONS AND RECOMMENDATIONS
Syndromic surveillance is currently being utilized in human health settings to
address the issue of timely identification of disease outbreaks [2, 73, 74]. Research in the
area of syndromic surveillance has recently emerged for application in animal health
surveillance [7-9, 50, 54, 75]. However, there has been little investigation into the
suitability of bovine abattoir condemnation data in Ontario for use in a food animal
syndromic surveillance system and the quantitative methods necessary for this type of
system.
In this thesis, I investigated factors that influence condemnation rates, the
performance of different cluster detection methods, and various approaches for selecting
abattoirs for sentinel surveillance. Overall, it was found bovine condemnation data from
provincially inspected abattoirs to be useful for food animal syndromic surveillance,
since they provided a more regionally detailed picture of emerging diseases in Ontario.
Specifically, it was found most cattle were shipped less than 100 km for slaughter at
provincial abattoirs [24]. This validated anecdotal evidence suggesting that cattle shipped
to provincial abattoirs originate from relatively local farms. Furthermore, disease-related
factors (e.g., age of cattle and seasonal trends) and non-disease factors (e.g., sales price of
cattle, abattoir audit rating) were shown to be associated with both whole carcass and
portion condemnations from cattle [24, 76]. Non-disease factors such as abattoir audit
rating and sales price were shown to have an impact on condemnation rates and improve
quantitative cluster detection methods by minimizing “noise” in the data, thus reducing
the probability for false-alarms. The effect this “noise” can have on quantitative cluster
149
detection methods was then demonstrated by comparing the results of various space-time
scan statistics with varying abilities to control for covariates [77]. The results from this
study suggested that model-adjusted approaches for controlling for covariates in scan
statistics appeared to perform best in terms of the ability to include all important
covariates and suitability for use with bovine abattoir condemnation data. Lastly,
methods to improve the efficiency of syndromic surveillance were investigated by
comparing the use of various sentinel abattoir selection approaches to select sentinel
abattoirs to reduce the number of sample sites, while still maintaining the trends of the
full dataset. The most effective sentinel selection approach was utilizing data from
abattoirs processing cattle all weeks of the year, and this approach shows promise for
integration into a food animal sentinel syndromic surveillance system.
While these findings suggest that bovine abattoir condemnation data would be
suitable for integration in a food animal syndromic surveillance system, there are some
limitations in the data which have become apparent through this thesis including data
quality issues and limitations with current methodological approaches. During this study,
data quality emerged as a potential barrier for the use of these data for syndromic
surveillance. Provincially inspected abattoirs in Ontario use lay meat inspectors. These
inspectors have the ability to partially condemn carcasses in provincial abattoirs without
direct veterinary oversight [78]. A similar study investigating the suitability and
limitations of provincially inspected porcine abattoir condemnation data found that while
provincial meat inspectors undergo rigorous classroom and field training programs, and
efforts are underway to improve standardization of inspection procedures across the
province, there may be great variability in classification of lesions by non-veterinary
150
inspectors leading to an increased recording bias between inspectors and individual
abattoirs [8]. This unique regulation, along with the subjective nature of meat inspection
makes observational and operator bias a potential concern when using abattoir
condemnation data for syndromic surveillance [8]. This study also found similar concerns
during the investigation of bovine condemnation data. Condemnation classifications used
by inspectors are subjective and can represent multiple pathologies causing
misclassification bias within these data. For example, “parasitic liver” is a condemnation
category which can represent a truly parasitic infection, as well as non-parasitic
conditions such as necrosis, fibrosis, cirrhosis, atrophy, telangiectasia and adhesions. This
could result in misclassification of condemnations and bias the results of cluster detection
methods involving these data. However, data are not currently available in Ontario to
identify and account for variation at the inspector level. To prepare for a future food
animal syndromic surveillance system, further research regarding specialized training of
inspectors and standardization of condemnation classification is essential.
Disease-related (e.g., older cattle and seasonal trends), and non-disease factors
(e.g., sales price of cattle, abattoir audit rating) were found to be important predictable
factors associated with both whole carcass and portion bovine condemnations in Ontario
[24, 76]. By controlling for these variables in quantitative methods for outbreak
surveillance, there would be a reduction of “background noise” in the data and reduce the
frequency of false-alarms. This reduction in data “noise” has been previously
demonstrated in a human health application for syndromic surveillance of influenza [48].
In the study, a model-adjusted version of the space-time scan statistic was applied for
syndromic surveillance of lower respiratory complaints in a human health care setting.
151
The study controlled for non-disease factors such as day of the week, month and holidays
and found that the number of false alarms could be reduced by removing the “noise” of
predictable covariates [48]. I also demonstrated the effect these covariates have on the
results of spatio-temporal scan statistics using various approaches for controlling for
confounding variables [77]. While it was suspected that the model-adjusted and model
residual adjustment approaches would be the most suitable for quantitative cluster
detection involving provincial abattoir condemnation data, there were some limitations
with these methods when used in practical applications. These methods are the most
versatile of the adjustment approaches, as they are able to account for both categorical
and continuous covariates, however, when the event is rare, as in the case of abattoir
condemnations, the precision of the residuals using the normal model is unstable and may
give inaccurate results for the model residual adjustment approach [77]. Thus, this
approach is not appropriate for the application of provincial abattoir condemnation data
unless counts are aggregated in space and/or time. The model-adjusted approach using
the ratio of observed versus expected condemnations under the Poisson model would be
more appropriate when utilizing relatively rare events, as in the case of abattoir
condemnation data. However, this approach, like the residual approach, also involves
more complex analyses than some of the other methodological approaches. In addition,
the multi-level model included temporal variables, such as year and season, which are
more conducive to retrospective analyses and would be difficult to account for
prospectively.
While bovine abattoir condemnation data were shown here to be suitable and
useful for food animal syndromic surveillance, a limitation of this thesis is the lack of a
152
documented bovine outbreak or simulated outbreak data to validate the results of our
work. It is recommended that further research and validation are conducted before the
methods suggested in this thesis are applied in practice. In addition, rather than using the
full complement of approximately 100 – 150 abattoirs, it may be advantageous to conduct
enhanced surveillance at carefully selected sentinel abattoirs. Abattoir selection methods
investigated in this thesis found that pneumonic lung condemnation rates from 45
abattoirs open all weeks of the year provided a good approximation of the condemnation
rates of all 98 abattoirs open throughout the study period. While the abattoirs open all
weeks sentinel selection approach strongly reduced the number of abattoirs needed to
conduct syndromic surveillance with these data, it is uncertain whether this reduction is
sufficient and manageable to conduct intensive and targeted surveillance. It is
recommended that further research is conducted to determine if there are other sentinel
selection approaches that could reduce this number of abattoirs even further.
In conclusion, the use of abattoir condemnation data is promising for integration
into a food animal syndromic surveillance system in Ontario. Research regarding this
under-utilized data resource continues to emerge worldwide using data from various
species [8, 33, 53, 54]. A study by Thomas-Bachli et al. investigated the suitability and
limitations of using abattoir condemnation data for syndromic surveillance of emerging
diseases of swine in Ontario [8]. This study found several clusters of high condemnation
rates for kidneys with nephritis in time and space-time which preceded the time frame
during which case clusters of PCV-2 were detected using traditional laboratory
surveillance. In addition, the trends found in kidney condemnation rates related to
nephritis lesions in Eastern Ontario were consistent with documented disease outbreaks
153
during that time period [8]. These findings further support the investigation of using
abattoir condemnation data for food animal syndromic surveillance. In addition, studies
in the United States have also utilized abattoir data for syndromic surveillance. Engle
[53] used condemnation data from the electronic Animal Disposition Reporting System
(eADRS) and found that a swine erysipelas outbreak in Iowa and Minnesota during July
2001 could have been identified approximately 10 months earlier if an automated
surveillance system had been in place. A similar study in the United States evaluated the
use of condemnation data from swine to set up an animal health monitoring system [53].
In 2014, Vial et al. [33] evaluated the use of Swiss condemnation data for cattle, pigs and
small ruminants to determine the suitability of these data for integration in a syndromic
surveillance system. This study investigated the patterns in the number of animals
slaughtered and condemned, reasons for whole carcass condemnations and reporting
biases and regional effects of condemnation rates. While the study found these data
provided simultaneous coverage of cattle, pigs and small ruminants for the entire country,
as well as traceability of each condemnation to its farm of origin, there were issues with
the lack of timeliness of reporting (30-60 day delays between condemnation and
notification); a delay which currently limits the use of these data for early detection in
Switzerland [33]. In contrast, a study in France utilizing slaughterhouse data from cattle
concluded that it was possible to define groups of underlying reasons for condemnation
of portions and encouraged the increased use of meat inspection data for syndromic
surveillance [54]. The recent emergence of research studies, including this thesis,
investigating the use of abattoir condemnation data for syndromic surveillance reinforces
154
the potential importance these data may have for food animal disease surveillance and the
need for continued research in this area.
While the findings from this thesis support the use of bovine abattoir
condemnation data for integration into a food animal syndromic surveillance system, the
Ontario Ministry of Agriculture and Food (OMAF) should consider supporting further
research involving these data before formalizing a food animal syndromic surveillance
system in Ontario. Specifically, I would recommend improving the standardization of
abattoir condemnation classifications to improve the quality of these data. In addition, it
is recommended that model-adjusted space-time scan statistics be applied in a
prospective manner without the use of retrospective variables such as year to evaluate its
performance. Supplementary investigation of sentinel selection criteria to determine if the
number of sentinel abattoirs could be further reduced is also encouraged. In addition,
validation of methods discussed in this thesis using documented outbreak or simulated
data are strongly recommended before a formal food animal syndromic surveillance
system is established. Finally, it should be remembered that a valuable contribution of
surveillance beyond rapid detection of disease, is the value of health in support of trade.
Research investigating the use of advanced methods for disease surveillance in Ontario
and Canada demonstrates and reassures Canada’s trade partners how closely animal
health is monitored from various sources and Canada’s potential leadership in advancing
food animal surveillance.
155
REFERENCES
1. Edge VL, Pollari F, Ng LK, Michel P, McEwen SA, Wilson JB, Jerrett M, Sockett PN,
Martin SW: Syndromic surveillance of Norovirus using over-the-counter sales of
medications related to gastrointestinal illness. Can J Infect Dis Med Microbiol 2006,
17:235-241.
2. Edge VL, Pollari F, Lim G, Aramini J: Syndromic surveillance of gastrointestinal
illness using pharmacy over-the-counter sales: A retrospective study of waterborne
Ouutbreaks in Saskatchewan and Ontario. Can J Public Health 2004, 95:446-446-50.
3. Steiner-Sichel L, Greenko J, Heffernan R, Layton M, Weiss D: Field Investigations of
Emergency Department Syndromic Surveillance Signals - New York City. MMWR
2004 :184-184-9.
4. O'Sullivan T, Friendship R, Pearl D, McEwen B, Dewey C: Identifying an outbreak
of a novel swine disease using test requests for porcine reproductive and respiratory
syndrome as a syndromic surveillance tool. BMC Vet Res 2012, 8:122.
5. Dupuy C, Morignat E, Maugey X, Vinard J-L, Hendrikx P, Ducrot C, Calavas D, Gay
E: Defining syndromes using cattle meat inspection data for syndromic surveillance
purposes: a statistical approach with the 2005-2010 data from ten French
slaughterhouses. BMC Vet Res 2013, 9:88.
6. Metre DC, Barkey DQ, Salman MD, Morley PS: Development of a syndromic
surveillance system for detection of disease among livestock entering an auction
market. J Am Vet Med Assoc 2009, 234:658-664.
7. Amezcua M, Pearl D, Friendship R, McNab W: Evaluation of a veterinary-based
syndromic surveillance system implemented for swine. CJVR 2010, 74: 241-251.
8. Dórea FC, Muckle CA, Kelton D, Mcclure JT, Mcewen BJ, Mcnab WB, Sanchez J,
Revie CW: Exploratory analysis of methods for automated classification of
laboratory test orders into syndromic groups in veterinary medicine. PloS One
8(3):e57334.
9. Thomas-Bachli AL, Pearl DL, Friendship RM, Berke O: Suitability and limitations
of portion-specific abattoir data as part of an early warning system for emerging
diseases of swine in Ontario. BMC Vet Res 2012, 8:3.
10. Alton GD, Pearl DL, Bateman KG, McNab WB, Berke O: Factors associated with
whole carcass condemnation rates in provincially-inspected abattoirs in Ontario
2001-2007: implications for food animal syndromic surveillance. BMC Vet Res 2010,
6:42.
156
11. Alton G, P., Alton GD, Pearl DL, Bateman KG, McNab WB, Berke O: Suitability of
bovine portion condemnations at provincially-inspected abattoirs in Ontario
Canada for food animal syndromic surveillance. BMC Vet Res 8:88.
12. Alton GD, Pearl DL, Bateman KG, Mcnab B, Berke O, Alton GD, Pearl DL,
Bateman KG, McNab B, Berke O: Comparison of covariate adjustment methods
using space-time scan statistics for food animal syndromic surveillance. BMC Vet Res
9:231.
13. Food Safety and Qality Act, 2001 [http://www.e-
laws.gov.on.ca/html/statutes/english/elaws_statutes_01f20_e.htm]
14. Kleinman K, Abrams A, Kulldorff M, Platt R: A model-adjusted space-time scan
statistic with an application to syndromic surveillance. Epidemiol Infect 2005,
133:409-419.
15. Vial F, Reist M: Evaluation of Swiss slaughterhouse data for integration in a
syndromic surveillance system. BMC Vet Res 10:33.
16. Weber WD: Development of an animal monitoring system based on slaughter
condemnation data. Miami: Proceedings of the Eighth International Society for Disease
Surveillance Conference; 2009. 3-4 December.