Random effect selection in generalised linear models: a practical application to slaughterhouse surveillance data in Denmark

Matt Denwood; Hans Houe; Björn Forkman; Søren Saxmose Nielsen

Abstract

We analysed abattoir recordings of meat inspection codes with possible relevance to onfarm animal welfare in cattle. Random effects logistic regression models were used to describe individual-level data obtained from 461,406 cattle slaughtered in Denmark. Our results demonstrate that the largest variance partition was at farm level for most codes, but there was substantial variation in reporting for some meat inspection codes between abattoirs.
There was also substantial agreement for the relative under or over-reporting of different slaughter codes within individual abattoirs. This indicates that the sensitivity of routine surveillance in Denmark is affected by differences in the working practices between abattoirs, resulting in biased prevalence estimates. Therefore, it is essential to correct for the variation in reporting between abattoirs before meaningful inference can be made from prevalence estimates based on data derived from meat inspection.

Original language	English
Publication date	26 Mar 2015
Number of pages	11
Publication status	Published - 26 Mar 2015
Event	Annual Meeting of the Society of Veterinary Epidemiology and Preventive medicine: SVEPM - Ghent, Belgium Duration: 25 Mar 2015 → 27 Mar 2015

Conference

Conference	Annual Meeting of the Society of Veterinary Epidemiology and Preventive medicine
Country/Territory	Belgium
City	Ghent
Period	25/03/2015 → 27/03/2015

Access to Document

Denwood_SVEPM Proceedings 2015Final published version, 207 KB

Cite this

Random effect selection in generalised linear models : a practical application to slaughterhouse surveillance data in Denmark. / Denwood, Matt ; Houe, Hans ; Forkman, Björn et al.

2015. Abstract from Annual Meeting of the Society of Veterinary Epidemiology and Preventive medicine, Ghent, Belgium.

Research output: Contribution to conference › Conference abstract for conference › Research › peer-review

@conference{7cda486bac2a41f89938242f32c78944,

title = "Random effect selection in generalised linear models: a practical application to slaughterhouse surveillance data in Denmark",

abstract = "We analysed abattoir recordings of meat inspection codes with possible relevance to onfarm animal welfare in cattle. Random effects logistic regression models were used to describe individual-level data obtained from 461,406 cattle slaughtered in Denmark. Our results demonstrate that the largest variance partition was at farm level for most codes, but there was substantial variation in reporting for some meat inspection codes between abattoirs.There was also substantial agreement for the relative under or over-reporting of different slaughter codes within individual abattoirs. This indicates that the sensitivity of routine surveillance in Denmark is affected by differences in the working practices between abattoirs, resulting in biased prevalence estimates. Therefore, it is essential to correct for the variation in reporting between abattoirs before meaningful inference can be made from prevalence estimates based on data derived from meat inspection.",

author = "Matt Denwood and Hans Houe and Bj{\"o}rn Forkman and Nielsen, {S{\o}ren Saxmose}",

year = "2015",

month = mar,

day = "26",

language = "English",

note = "Annual Meeting of the Society of Veterinary Epidemiology and Preventive medicine ; Conference date: 25-03-2015 Through 27-03-2015",

}

TY - ABST

T1 - Random effect selection in generalised linear models

T2 - Annual Meeting of the Society of Veterinary Epidemiology and Preventive medicine

AU - Denwood, Matt

AU - Houe, Hans

AU - Forkman, Björn

AU - Nielsen, Søren Saxmose

PY - 2015/3/26

Y1 - 2015/3/26

N2 - We analysed abattoir recordings of meat inspection codes with possible relevance to onfarm animal welfare in cattle. Random effects logistic regression models were used to describe individual-level data obtained from 461,406 cattle slaughtered in Denmark. Our results demonstrate that the largest variance partition was at farm level for most codes, but there was substantial variation in reporting for some meat inspection codes between abattoirs.There was also substantial agreement for the relative under or over-reporting of different slaughter codes within individual abattoirs. This indicates that the sensitivity of routine surveillance in Denmark is affected by differences in the working practices between abattoirs, resulting in biased prevalence estimates. Therefore, it is essential to correct for the variation in reporting between abattoirs before meaningful inference can be made from prevalence estimates based on data derived from meat inspection.

AB - We analysed abattoir recordings of meat inspection codes with possible relevance to onfarm animal welfare in cattle. Random effects logistic regression models were used to describe individual-level data obtained from 461,406 cattle slaughtered in Denmark. Our results demonstrate that the largest variance partition was at farm level for most codes, but there was substantial variation in reporting for some meat inspection codes between abattoirs.There was also substantial agreement for the relative under or over-reporting of different slaughter codes within individual abattoirs. This indicates that the sensitivity of routine surveillance in Denmark is affected by differences in the working practices between abattoirs, resulting in biased prevalence estimates. Therefore, it is essential to correct for the variation in reporting between abattoirs before meaningful inference can be made from prevalence estimates based on data derived from meat inspection.

M3 - Conference abstract for conference

Y2 - 25 March 2015 through 27 March 2015

ER -

Random effect selection in generalised linear models: a practical application to slaughterhouse surveillance data in Denmark

Abstract

Conference

Access to Document

Fingerprint

Cite this