Outline The “PigLame” 1.The case problem Model 2.Modeling methods in general An example of an Object- Oriented Bayesian network model 3.Qualitative structure of the model Leonardo de Knegt Tina Birk Jensen 4.Elicitation of probabilities 5.Use of the model Leg disorders in finishers Leg disorders in finishers An economical problem for farmers due to: • Increased work load • Cost of treatments • Reduced productivity • Risk of condemnations • Leg disorders: Any lesion or dysfunction of the leg or • A negative impact on animal welfare claw that might give rise to lameness • Lameness: Deterioration in the gait and/or posture Causes of leg disorders Control strategies 1. Infectious Control strategies against leg disorders will depend on the cause category Mycoplasma hyosynoviae, Erysipelothrix rhusiopathiae, Haemophilus parasuis, Streptococcus suis <2% • Infectious leg disorders Antibiotics 2. Physical • Physical leg disorders Reconstructing the pen Fracture, lesion to the claw wall, lesion to the claw sole • Inherited leg disorders Boar semen <1% 50-80% Weight gain 3. Inherited Osteochondrosis manifesta, osteochondrosis dissecans 70% 30% 1
Useful information Useful information Herd level Pig level • Herd size (number of pigs delivered) • Observe pigs from outside the pen Cheap • Stocking density (high/low) Cheap • Clinical investigation • Floor type in pens (slatted/concrete) • Supply of straw in pens (deep/sparse/no) • Bacteriological investigation Expensive • Purchase policy (own piglets/1/>1) • Pathological investigation Expensive • Production type (sectioned/continuous) To make a herd diagnosis of leg disorders The ”PigLame” model Purpose of the model Challenges • To estimate probability distributions of different • What information to use manageable causes of leg disorders in finisher herds • How much information to use • How to collect the information Qualitative structure of the model Herd Pig Characteristics Info Info • Based on information from the literature • All nodes are discrete Leg disorder • Each cause-category defined as a risk index I on an arbitrary scale from 0 to 9 Strategy Strategy Strategy 1 2 3 2
Qualitative structure of the model Object-oriented Bayesian network Characteristics The Pig class • Individual pig information Object-oriented structure • Ease the specification of the Bayesian network Gender Lean meat percentage Diagnostic test results Leg disorder • Hierarchical structure • Two classes: Herd class and pig class Object-oriented Bayesian network Feed strat The Herd class Breed Herd size Produc- Pen Herd Pur- Floor Straw Production Gain Den Size tion chase Purchase Pen density Floor type Straw Cause category Physical Physical Infectious Infectious Inherited Inherited Pig object Pig object Pig object Herd class Herd class Physical Infectious Inherited Floor Floor Pen Den Pen Den Gain Gain …… Gen der LMP Frac Claw Claw Hae Myco Strep Erysi OCM OCD ture Wall Sole mo Physical Infectious Inherited Frac Claw Claw Claw Frac Claw Pig Myco Myco Strep Strep Erysi Erysi Haemo Haemo OCM OCM OCD OCD Sole ture ture Wall Wall Sole Lame Obs C1 P1 C2 P2 C3 P3 C4 P4 B1 C5 P5 B2 C6 P6 B3 C7 P7 B4 C8 P8 C9 P9 Lame Pig class Pig class 3
Herd class Elicitation of probabilities The probabilities in the model are based on Floor Floor Pen Den Gain …… 1. Results from published literature • Conversion of odds ratios to conditional probabilities Physical Infectious Inherited • <40 conditional probabilities 2. Expert opinions (9 experts) • >150 conditional probabilities Frac Frac Claw Claw Myco Strep Erysi Haemo OCM OCD Sole ture ture Wall • Not randomly distributed • Average of individual elicitations Pig class Elicitation of probabilities Example: Always (almost) 100 P(Fracture|fully slatted floors) 85 Usually 75 R 1 R 2 R n Consider 100 pigs examined individually Often at a herd visit. The herd has fully slatted floors in the pens. As often as not 50 How often do you, during the examination I expect to find a pig with a fracture? Sometimes 25 Once in a while Pig class 15 D 1 D n D 2 (Almost) never 0 Elicitation of probabilities Elicitation of probabilities Cause-categories: R 1 R 1 R 2 R 2 R n R n • Defined as Risk Index I on an arbitrary scale • 0: Low risk I • 9: High risk Pig class D 1 D n D 2 4
Elicitation of probabilities Elicitation of probabilities Risk index based on a linear equation R 1 R 2 R n • I the resulting risk index • μ the intercept I • ρ k the systematic effect of risk factor k • ε random residuals • Assumptions: Pig class D 1 D n D 2 • No interactions between the risk factors • The effects are additive Elicitation of probabilities Elicitation of probabilities Leg disorder nodes: Modeled using a logistic regression • Parameter estimates for the Risk Index and leg disorder nodes found by: ( ( )) = α + β Logit P D I k k k • Using the probabilities elicited by experts or literature • Logit(P(D k )): Logistic transformation of the conditional probability • Fitting a logistic linear model of a pig to have the leg disorder k • Optimizing the fit • α: Intercept indicating the base prevalence of the leg disorder k • β: Slope indicating the sensitivity to changes in the risk level of the herd • I: Risk Index Use of the model Use of the model Two fictitious herds with same prevalence of lameness: 20% pigs are lame due to Mycoplasma hyosynoviae • Decide on the level of information needed in order to identify the most likely cause-category • Low risk herd: Deliver 2000 finishers annually Sectioned production • Is it necessary to investigate individual pigs in a herd? Produce own piglets Low pen densities • Which diagnostic test(s) should be performed? Solid floors No supply of straw • How should pigs for diagnostic examination be selected? • High risk herd: Deliver 6000 finishers annually • How many pigs should be selected? Continuous production Purchase from several herds High pen densities Partially slatted floors Sparse supply of straw 5
Use of the model Different scenarios investigated: Risk index Risk index 1. Herd evidence 2. Herd evidence and observing 50 randomly selected pigs for lameness Risk index Risk index 3. Herd evidence and performing diagnostic ex. of lame pigs 4. Herd evidence and performing diagnostic ex. of all pigs Risk index Risk index Risk index Risk index Use of the model Conclusion • ”PigLame” model is an OOBN model • Low risk herd Ease the specification of the model • Necessary to perform diagnostic examination of pigs • Suitable method of combining information from two • • High risk herd different levels • Information regarding the herd characteristics is sufficient A similar approach can be used for other problems at • herd level More economic benefit in performing • Probabilities mainly from experts diagnostic examination of individual pigs in the low risk herd Prone to subjectivity • 6
Recommend
More recommend
