2014.ISDS.Abstracts.Final.pdf


ISDS Annual Conference Proceedings 2014. This is an Open Access article distributed under the terms of the Creative Commons Attribution-
Noncommercial 3.0 Unported License (http://creativecommons.org/licenses/by-nc/3.0/), permitting all non-commercial use, distribution, 
and reproduction in any medium, provided the original work is properly cited.

ISDS 2014 Conference Abstracts

Assessment of Several Algorithms for Outbreak 
Detection using Bovine Meat Inspection Data for 
Syndromic Surveillance: A Pilot Study on Whole 
Carcass Condemnation Rate
Céline Dupuy*1, 3, Eric Morignat1, Fernanda C. Dórea2, Christian Ducrot3, Didier Calavas1 
and Emilie Gay1

1ANSES, Lyon, France; 2Swedish Zoonosis Centre, Uppsala, Sweden; 3INRA, Saint Genès Champanelle, France

Objective
The objective of the work was to assess the performance of several 

algorithms for outbreak detection based on weekly proportions of 
whole carcass condemnation

Introduction
The majority of farmed animals are sent to slaughterhouses, 

making them a focal point for potential collection of health data. 
However, these data are not always available to health officials, and 
remain under-used for cattle health monitoring. Meat inspection 
data are mainly non-diagnostic (condemned portion and reasons for 
condemnation) and cover a large population. These characteristics 
make them a good candidate for syndromic surveillance. Whole 
carcass condemnation rate is linked to acute infections which reduces 
the dilution bias due to the variable period of time between cattle 
infection and the detection of lesions at the slaughterhouse.

Methods
Data from 177,098 cattle slaughtered in one French slaughterhouse 

from 2005 to 2009 were used (proportion of whole carcass 
condemnations: 0.97%). The method for outbreak detection covered 
three steps as previously explored by Dórea et al. on laboratory test 
data [1]: i) preparation of an outbreak-free historical baseline over 
five years, ii) simulation of over 100 years of baseline time series with 
injection of artificial outbreak signals with several shapes, durations 
and magnitudes and iii) assessment of the performance (sensitivity, 
specificity and precocity) of several algorithms to detect these artificial 
outbreak signals. The tested algorithms were the Shewart p-chart, one-
sided confidence interval of a negative binomial regression model, 
and EWMA and CUSUM control charts on residuals of a negative 
binomial model. Age and sex were taken into account because of their 
known effect on whole carcass condemnation [2].

Results
The highest sensitivity was obtained using negative binomial 

regression and the highest specificity using CUSUM or EWMA 
(Table 1). EWMA sensitivity was too low to select this algorithm 
for efficient outbreak detection. CUSUM showed complementary 
performance to negative binomial regression.

Conclusions
The use of whole carcass condemnation data for syndromic 

surveillance is more complex than monitoring counts because we need 
to take into account the denominator (number of cattle slaughtered) 
as well as age and sex. The recent deployment of a national meat 
inspection database in France will enable prospective investigation 
of this indicator on real data. The Shewart control chart could be 
used as a first step considering its high sensitivity and simplicity 
of implementation followed by the negative binomial model and 

CUSUM on residuals of the negative binomial model when historical 
data becomes available.

Summary statistical values of performance indicators for all age-sex categories

For each indicator the median (Minimum-Maximum) values for each age sex 
category, each outbreak duration (2, 4 and 8 weeks) and magnitude (1 to 4) 
are presented by outbreak shape and outbreak detection algorithm. Parameters 
for each algorithm were: for Shewart: K=1.3; for CUSUM: H=2; for EWMA: 
Lambda=0.4 and L=1.3; for negative binomial regression: 80% confidence 
interval.

Keywords
Syndromic surveillance; Animal health surveillance; Early outbreak 
detection; Time series

References

1.Dórea FC, et al. Syndromic surveillance using veterinary laboratory
data: data pre-processing and algorithm performance evaluation. J R 
Soc Interface 2013; 10(83).

2.Dupuy C, et al. Factors associated with offal, partial and whole carcass
condemnation in ten French cattle slaughterhouses. Meat Sci 2014; 
97(2): 262-269

*Céline Dupuy
E-mail: celine.dupuy@agriculture.gouv.fr    

Online Journal of Public Health Informatics * ISSN 1947-2579 * http://ojphi.org * (1):e125, 201