Description and factors of variation of the overall health score in French dairy cattle herds using the Welfare Quality® assessment protocol

Page 1

Dsa

MLa

b

Nc

d

ARRA

KODFW

T

0h

Preventive Veterinary Medicine 112 (2013) 296– 308

Contents lists available at ScienceDirect

Preventive Veterinary Medicine

j ourna l ho me pag e: ww w.elsev ier .com/ locate /prevetmed

escription and factors of variation of the overall healthcore in French dairy cattle herds using the Welfare Quality®

ssessment protocol

. Coignarda,b,∗, R. Guatteob,a, I. Veissierd,c, A. de Boyer des Rochesc,d,. Mounierc,d, A. Lehébela, N. Bareilleb,a

INRA, UMR 1300 Biology, Epidemiology and Risk Analysis in Animal Health, CS 40706, F-44307 Nantes, FranceLUNAM Université, Oniris, Nantes-Atlantic College of Veterinary Medicine, Food Sciences and Engineering, UMR BioEpAR, F-44307antes, FranceUniversité de Lyon, VetAgro Sup, UMR 1213 Herbivores, F-69280 Marcy L’Etoile, FranceINRA, UMR 1213 Herbivores, F-63122 Saint-Genès-Champanelle, France

a r t i c l e i n f o

rticle history:eceived 8 February 2013eceived in revised form 30 July 2013ccepted 31 July 2013

eywords:verall health scoreairy cattle herdactor of variationelfare assessment

a b s t r a c t

Extensive information is available in the literature on the specific risk factors of the mainhealth disorders afflicting dairy cattle herds. However, it remains difficult to manage aherd’s overall health because measures to control one risk factor can exacerbate the riskof another disease. To achieve and maintain good overall herd health, livestock systemsand management practices need to simultaneously take into account all of the main healthdisorders. We aimed to identify the characteristics of systems and practices conducive togood herd health using the Welfare Quality® assessment protocol for cattle. This protocolallows an assessment of the level of health and welfare at the herd level according to theopinion of a selected group of 13 experts from animal sciences. Our objectives were to (i)describe the distribution of dairy herds’ health scores in a representative sample of Frenchdairy cattle herds, and (ii) to investigate systems (housing system, milking system, herd size,breed, farm location) and management practices associated with variations of the overallhealth score of dairy herds. This protocol was carried out on 130 farms between December2010 and March 2011. A multivariable analysis of variance (ANOVA) was performed toinvestigate the factors of variation of the overall health score at the herd level.

The overall health scores of the farms in the sample were classified as moderate forthe vast majority of farms (95.4%) (mainly due to subclinical mastitis, dystocia and paininduced by disbudding/dehorning) and varied little between farms. Some livestock sys-tems were associated with a higher overall health score: straw yards and milking parlors(P < 0.0001), highland vs. lowland locations (P = 0.013), Montbeliarde rather than Holstein

breeds (P = 0.006). Some management practices also were associated with a higher level ofhealth: medium herd average parity (P = 0.03), low proportion of dirty cows (P = 0.002) andlow proportion of cows with abnormal body condition (P = 0.04). These results suggest thatsome systems contribute to better health and that improvement of health can be obtainedin the short term by the modification of routine management practices.

∗ Corresponding author at: INRA, UMR 1300 Biology, Epidemiology and Risk Ael.: +33 0240 68 76 74; fax: +33 0240 68 77 68.

E-mail address: [email protected] (M. Coignard).

167-5877/$ – see front matter © 2013 Elsevier B.V. All rights reserved.ttp://dx.doi.org/10.1016/j.prevetmed.2013.07.018

© 2013 Elsevier B.V. All rights reserved.

nalysis in Animal Health, CS 40706, F-44307 Nantes, France.

Page 2

eterinary

M. Coignard et al. / Preventive V

1. Introduction

The welfare of farm animals has become a growing con-cern in recent years. According to the 2005 Eurobarometerstudy on farm animal welfare, 78.3% of European citizensbelieve that more should be done to improve the welfareand protection of farm animals within the European Union(European Commission, 2005). For their part, farmers havealways been concerned about the condition of their animals(Von Keyserlingk et al., 2009) and they must adapt theirmanagement practices in order to improve and optimizethe welfare of their herd.

There is currently a lack of knowledge about theactual level of animal welfare on dairy cattle farms. How-ever, scientists and livestock professionals consider healthimpairments as major potential threats to animal welfare.Indeed, in considering the five welfare freedoms (FarmAnimal Welfare Council, 1992), some of the most impor-tant consequences of poor welfare in dairy cattle are theoccurrence of diseases, in particular foot and leg disordersand mastitis (EFSA Reports, 2009).

In the literature, extensive information is available onthe specific risk factors for the main health disorders(e.g. mastitis, lameness, and metabolic disorders) afflict-ing dairy cows. However, it remains difficult to manage aherd’s overall health because measures to control one riskfactor can exacerbate the risk of other diseases. When afarmer tries to solve a specific health disorder in his herdby making changes in his system or practices, the risk ofanother disorder occurring consequently can increase. Asan example, we can consider the two main health disordersobserved on dairy herds in terms of prevalence, economiclosses and pain induced: lameness (Whay et al., 1997;Green et al., 2002) and udder health disorders (Fourichonet al., 2001a; Seegers et al., 2003; Ghavi Hossein-Zadeh andArdalan, 2011). Cubicles are well known to increase therisk of the occurrence of lameness (Haskell et al., 2006;EFSA Reports, 2009) while they reduce the risk for mas-titis (Fregonesi and Leaver, 2001; Bareille et al., 1998). Theinverse has been observed for straw yards. It is thereforeimpossible to choose one housing system over the otherbased on the occurrence of these disorders. Such a decisionmust be based on an overall assessment of the health sta-tus of the dairy herd that takes into account the occurrenceand severity of all of the major disorders afflicting the herd.

Recently, a European research project named WelfareQuality® aimed at elaborating, for several livestock species(cattle, pigs and poultry), a system for the assessmentof welfare at the herd level based on experts’ opinion.These experts proposed to take into account the con-comitant measurement of the five freedoms through themeasurement of four principles (health, feeding, housingand behavior) which are finally combined into a global wel-fare scoring (Welfare Quality®, 2009). Within the WelfareQuality® protocol, the assessment of each principle is basedon the evaluation of two to four criteria. For instance, the‘health’ principle includes the assessment of (i) absence of

injuries (ii) absence of diseases and (iii) absence of paininduced by management procedures. To the best of ourknowledge, this is currently the only method allowingan overall assessment of farm animals’ health. The farm’s

Medicine 112 (2013) 296– 308 297

compliance with each criterion is checked in regards tothe experts thresholds thanks to measures collected fromobserving animals or resources on the farm and from ques-tioning the farmer on his management practices. For dairycows, a total of 33 measures, all selected for their validity,reliability and feasibility (Knierim and Winckler, 2009), areperformed.

The objectives of our study were to (i) describe thedistribution of dairy herds’ health scores in a repre-sentative sample of French dairy cattle herds and (ii)investigate systems (housing system, milking system, herdsize, breed, farm location) and management practices asso-ciated with a variation of the overall health score ofdairy herds to promote those related to a better healthstatus. We used the Welfare Quality® assessment proto-col to produce an overall assessment of the dairy herds’health.

2. Materials and methods

2.1. Study sample

Five stratification criteria were selected to represent themain components of the diversity of current dairy herds inFrance: herd location (lowland regions (Brittany and Paysde la Loire, western France) vs. highland regions (Auvergne,central France and Rhône-Alpes, eastern France)), breed(Holstein vs. Montbeliarde, the two main dairy breedsin France), milking system (milking parlor vs. automaticmilking system (AMS)), housing system (straw yard vs.cubicles), and finally herd size (‘small herd’ with less than50 lactating cows and ‘medium herd’ with at least 50 lac-tating cows). To be included in the study, herds also hadto fulfill the following criteria: (i) the farmer agreed toparticipate in a one-day cross-sectional survey, (ii) thefarm was enrolled in a milk recording scheme, and (iii)the dairy cows were being kept indoors when the farmwas visited. The combination of the five stratification crite-ria gave a total of 24 strata (strata composed of AMS andstraw yard were not retained in this stratification becausethese combinations are rare in France). The nine AnimalHealth Services (‘Groupement de Défense Sanitaire’) of thetargeted areas each provided an exhaustive list of dairyfarmers in their region. These lists contained no informa-tion regarding the characteristics of the farms. One hundredfarms were then randomly selected from each list usingthe R®2.10.1 software (R Development Core Team, 2009).We decided to visit about 120–130 herds due to the lim-ited number of observers and the limited study period inwhich cows were kept indoors. The goal was to recruit asimilar number of farms per stratum, leading to an aver-age of five to six farms per stratum since 24 strata wereinitially expected. Farmers were contacted by phone. Dur-ing each call, the project and the practical constraintsof the visit were presented. If the farmer agreed to par-ticipate, information relative to the stratification criteria

was then obtained. The strata were thus filled in pro-gressively. The study population was therefore deemedrepresentative of the diversity of dairy farms in Francetoday.

Page 3

2 eterinary

2

Doicafmta

lQucacshQwthoo

QhtsmQre

2

wct‘mtptsit(rfpto‘ime

98 M. Coignard et al. / Preventive V

.2. Data collection

The visits were conducted by five observers betweenecember 2010 and March 2011 (each observer carried outne farm visit per day). The observers had been trainedn advance on how to use the Welfare Quality® proto-ol. Training consisted of classroom exercises using photosnd videos, observations of animals, and test runs on fourarms. Each of the five observers applied the entire assess-

ent protocol (Welfare Quality®, 2009) on the test farmso ensure that they all applied the protocol in a consistentnd coherent manner.

Among the 33 measures defined by experts and col-ected through the Welfare Quality® protocol (Welfareuality®, 2009) to assess dairy herd welfare, the 14 meas-res strictly related to health aspects cover three differentriteria: injuries, diseases and pain induced by man-gement procedures (Table 1). These 14 measures wereollected directly from observing the animals (N = 8), con-ulting farm records (N = 4), and questioning the farmer onis management practices (N = 2). As defined in the Welfareuality® protocol, in each herd, the animal-based measuresere collected from a sample of cows chosen at random

o assess between 45% and 100% of the whole herd. Inerds containing 30 lactating cows or less, all animals werebserved; the proportion assessed decreased when the sizef the herd was larger.

Five other welfare aspects measured by the Welfareuality® protocol are also believed to influence dairy cowealth. These measures were therefore collected in addi-ion to the 14 health measures mentioned above. For theame reason, four other factors related to herd manage-ent practices that are not considered in the Welfareuality® protocol were also collected by consulting farm

ecords and questioning farmers (an exhaustive list ofxplanatory variables is provided in Table 4).

.3. Calculation of the ‘overall health score’

Once all of the health measures were available, theyere combined at the herd level to calculate health

riterion-scores defined by experts ranging from 0 (worst)o 100 (excellent), respectively for ‘Absence of injuries’,Absence of diseases’ and ‘Absence of pain induced by

anagement procedures’ (Fig. 1 Step.1). At each step ofhe calculation of the welfare scores (from measures torinciples), thresholds and scores values ranking farms inerms of welfare were chosen according to the opinion of aelected group of 13 experts from animal sciences (veter-narians, ethologists, biologists and researchers) in ordero mix expertise and potentially also value judgementsBotreau, 2008; Welfare Quality®, 2009). For the crite-ion ‘Absence of injuries’, after calculating partial scoresor lameness and integument alterations according to therevalence and severity of these injuries in the herd, thesewo scores were combined using a Choquet integral tobtain the final score for that criterion. For the criterion

Absence of diseases’, the final score was calculated by tak-ng the sum of warnings and alarms obtained for each

easure. The alarm threshold was defined for each dis-ase as the minimum value above which a health action

Medicine 112 (2013) 296– 308

plan is required according to the experts’ opinion; the war-ning threshold is half of the alarm threshold. Lastly, afterattributing a partial score to disbudding/dehorning and taildocking, reflecting the appropriateness of procedures, thefinal score for the criterion ‘Absence of pain induced bymanagement procedures’ corresponded to the worst par-tial score for these procedures.

In a final step, these three health criterion-scores werecombined again using a Choquet integral to calculate theoverall health score, which also ranged from 0 to 100(Fig. 1 Step.2). Indeed, experts made the choice of not toallow compensation between scores because they con-sidered that a good health cannot be reached if one ofthe health measures is clearly impaired. Thus, the Cho-quet integral was used since it is a specific mathematicaloperator which allows to take into consideration both theinteraction between criteria and the relative weight of eachone to calculate an overall score (Labreuche and Grabich,2003). Therefore, this scoring system implies that herdsobtaining a same overall health score can be affected byvery different health disorders.

For more details, the full description of the methodis detailed in the Welfare Quality® protocol available athttp://www.welfarequality.net/network/45848/7/0/40.

2.4. Strategy of analysis

In a first step, we described the distribution of the over-all health scores, criterion-scores (for injuries, diseasesand pain) and health measures using the survey analysisprocedures (PROC SURVEYMEANS) of SAS (SAS®, version9.2) that takes into account the sampling proportions. Thescores for the overall health score, the three health crite-ria, and the four partial scores for lameness, integumentalterations, dehorning and tail docking were then inter-preted according to the three categories used in the WelfareQuality® assessment protocol to assess the overall healthscore: Good (score > 55), Moderate (55 ≥ score > 20) andPoor (score ≤ 20). As described above, measures relatedto the criterion ‘Absence of diseases’ are not interpretedaccording to scores in the Welfare Quality® assessmentprotocol, but with warnings and alarms. However, a simi-lar interpretation of these health measures is proposed inthe present study and was used in our analysis: Good: nowarning or alarm for the measure in the herd; Moderate:at least one warning and no alarm; and Poor: at least onealarm.

In a second step, we investigated the factors of variationof the overall health score at the herd level through uni-variable and multivariable analyses of variance (ANOVA)using the PROC SURVEYREG of SAS. Firstly, the normal dis-tribution of the dependant variable ‘overall health score’was checked. The association between the ‘overall healthscore’ and a total of 13 explanatory variables was thentested through a univariable analysis of variance. Thesevariables were initially selected because they are knownor very likely to be associated with dairy cow health.

Similar univariable analysis (PROC SURVEYREG of SAS)or chi-squared test for rare events (PROC SURVEYFREQof SAS) were performed to identify the factors of varia-tion of health measures in order to illustrate the relative

Page 4

M. Coignard et al. / Preventive Veterinary Medicine 112 (2013) 296– 308 299

Table 1Data collected for the assessment of the overall health score of dairy cows in a herd using the Welfare Quality® assessment protocol.

Health measures Type of raw data Frequency calculation Description of the method for collecting data

Lameness Animal-based measure Prevalence of lamenessobserved on a sample of cowsthe day of the visit

Cows are observed when walking on a surface on which theynormally walk. The cows gait score is assessed using this scale:0: Not lame: timing of steps and weight-bearing equal on all fourfeet1: Moderately lame: imperfect temporal rhythm in stride creatinga limp2: Severely lame: Strong reluctance to bear weight on one limb, ormore than one limb affected

Integument alterations Animal-based measure Prevalence of integumentalterations observed on asample of cows the day of thevisit

Five body regions of cows (neck/shoulder/back, hindquarter,tarsus, flank/side/udder, carpus) are observed on one side ofthe animal. On each region, the number of hairless patches andlesions/swellings of a minimum diameter of 2 cm are recorded

Coughing Animal-based measure Mean number of coughsexpressed per cow per 15 min

Recording using continuous behavior sampling of sudden andnoisy expulsion of air from the lungs of cows during a totalperiod of 120 min

Nasal discharge Animal-based measure Prevalence of nasal dischargeobserved on a sample of cowsthe day of the visit

Animal is observed without being touched. Assessed using thisscale:0: No evidence1:Evidence of clearly visible flow/discharge from the nostrils;transparent to yellow/green and often of thick consistency

Ocular discharge Animal-based measure Prevalence of ocular dischargeobserved on a sample of cowsthe day of the visit

Animal is observed without being touched. Assessed using thisscale:0: No evidence1: Evidence of clearly visible flow/discharge (wet or dry) from theeye, at least 3 cm long

Hampered respiration Animal-based measure Prevalence of hamperedrespiration observed on asample of cows the day of thevisit

Animal is observed without being touched. Assessed using thisscale:0: No evidence1: Evidence of deep and labored respiration. Expiration mostlyaccompanied by pronounced sound

Diarrhea Animal-based measure Prevalence of diarrheaobserved on a sample of cowsthe day of the visit

Animal is observed without being touched. Assessed using thisscale:0: No evidence1: Evidence of loose watery manure on both sides of the tail. Areaaffected at least the size of a hand

Vulvar discharge Animal-based measure Prevalence of vulvar dischargeobserved on a sample of cowsthe day of the visit

Animal is observed without being touched. Assessed using thisscale:0: No evidence1: Evidence of purulent effluent from the vulva or on the bottomside of the tail

Milk somatic cell count Farm records Prevalence of cows withsubclinical mastitis within thelast 3 months

Cow milk somatic cell counts are obtained from individualmilk records and assessed using this scale:0: Somatic cell count below 400,000 cell/mL for the last 3 months1: Somatic cell count of 400,000 cell/mL or above at least oncewithin the last 3 months

Mortality Farm records Annual cumulative incidenceof mortality

Defined as the number of dairy cows which died or wereeuthanized due to disease or accidents or were emergencyslaughtered during the last 12 months. This number is dividedby the yearly average number of dairy cows in the herd

Dystocia Farm records Annual incidence of dystocia Defined as the number of calvings where major assistance wasrequired in the herd during the last 12 months. This number isdivided by the total number of calving over the year. Majorassistance includes heavy traction, cesarean section andembryotomy according to the French system based on thefarmer appraisal

Downer cows Farm records Annual cumulative incidenceof downer cow syndrome

Defined as the number of cases of non-ambulatory cows in theherd during the last 12 months. This number is divided by theyearly average number of dairy cows

Disbudding/dehorning Questionnaire The farmer is asked about his or her management practices fordisbudding/dehorning (procedures, use ofanesthetics/analgesics)

Tail docking Questionnaire

influence of each one on the results from the overallhealth score. As none of the quantitative explanatoryvariables fulfilled the linearity of effect assumption withthe ‘overall health score’, they were each grouped into

The farmer is asked about his or her mutilation management(procedures, use of anesthetics/analgesics)

classes according to quartiles of the distribution. Only vari-ables associated with the ‘overall health score’ (P ≤ 0.25)in this initial screening step were included in the mul-tivariable analysis. The association between the retained

Page 5

300 M. Coignard et al. / Preventive Veterinary Medicine 112 (2013) 296– 308

Health measures of d airy

cows in herds*

Welfare Qual ity®’ s

data processing*

Lameness Pa r�al score

Integument altera �ons Pa r�al score

Coughing Warning and/or alar m

Nasal discharge Warning and/or alarm

Ocular dischargeW arning and/or alar m

Hampered respira�on Warning and/or alar m

DiarrheaW arning and/or alar m

Vulvar dischargeW arning and/or alar m

Milk so ma� c cell count Warning and/or alar m

Mortality Warning and/or alar m

Dystocia Warning and/or al arm

Downer cows Warning and/or alarm

Disbudd ing/ Dehorning Pa r�al score

Tail do cking Pa r�al score

Health criterion-score for ‘Absence

of injuries’ [0-100]*

Health criterion-score for ‘Absence

of disease’ [0-100]*

Health criterion-score for ‘Absence

of pain induce d by management

procedures’ [0-100]*

Overall health

score calculated

at herd level*

[0-100]

Sum of wa rning sand alarms

Combinat ion usingChoquet integral

Worst scoreretained

Combination usingChoquet integral

STEP 2STEP 1

F s at the fQ

es(arslisletbpiahbpsuraaicfk

bKomfic+

ig. 1. Methodology for the calculation of health criteria and overall scoreuality® assessment protocol.

xplanatory variables was also tested using the ‘chi-quared test’. If two variables were highly correlatedP < 0.05), the one with the lowest P-value in the univari-ble analysis or the one with the most biological plausibleelationship was retained. The exceptions were milkingystem and housing system, which were highly corre-ated (P < 0.0001). We therefore decided to turn themnto a single combined variable, “housing system–milkingystem”, because both were highly significant and bio-ogically interesting. Careful attention was paid to thepidemiological plausibility before investigating interac-ions between explanatory variables. Indeed, interactionsetween the two variables ‘Breed’ and ‘Herd average milkroduction adjusted for breed’ and the variable ‘Milk-

ng system–Housing system’ were investigated under thessumption that the ability of cows to adapt to a givenomogeneous system is related to their size and theirehavior which in turn depend on their breed and milkroduction. The manual backward elimination of non-ignificant variables and interactions (P > 0.10) then wassed in the multivariable analysis. A P-value <0.10 wasetained as indicator of statistical significance to take intoccount for the limited sample size and the exploratorypproach of the study. The presence of confounders wasnvestigated by checking that the estimates were nothanged by more than 20% when a variable was withdrawnrom the model. In the affirmative, the variable was thenept in the model.

The goodness of fit of the final model was assessedy graphical inspection of the residuals and by theolmogorov–Smirnov test to assess if the distributionf standardized residuals fitted with a standard nor-

al distribution. The Kolmogorov–Smirnov test of the

nal model (P = 0.15) was supplemented by a verifi-ation of skewness and kurtosis coefficients [−0.01;0.39].

arm level based on health measures collected on farms using the Welfare

3. Results

3.1. Description of the study sample

Of the 432 farmers contacted by phone, 131 (30%) werefinally included in the study. The three main exclusioncriteria were: (i) unavailability of the farmer during thestudy period, (ii) reluctance to a full day visit and (iii)incompatibility of the farm’s characteristics with theinclusion criteria (e.g. the farm had a tie-stall system,which were not included in the study). Complete datawere available on 130 farms. One farm was excluded fromthe study because the farmer did not remember that hiscows had to be in the barn at the time of the visit. The herdsize ranged from 23 to 120 lactating cows (mean ± S.D:52.2 ± 17.2). The 130 herds included a total of 6785 dairycows, of which 4265 (62.8%) were submitted to individualhealth measures. The distribution of the herds accordingto the stratification criteria is presented in Table 2. Asplanned, we visited on average 6 farms per stratum. Stratacomposed of AMS and small herds were not filled becausethis combination is very rare in France. Therefore, a total of21 strata were filled, confirming our initial assumption thatthe stratification criteria chosen before the recruitmentprocess represent the main components of the diversity ofdairy herds in France today.

3.2. Description of the distribution and qualitativeassessment of the overall health score, healthcriterion-scores and health measures

3.2.1. Quantitative resultsThe overall health score varied from 13.9 to 54.2 in

the sample (mean ± S.D: 33.2 ± 8.1) and half of the herdsobtained a score greater or equal to 33 (Table 3; Fig. 2).

Among the three criteria, ‘Absence of injuries’ obtainedthe best scores (median = 54.6) and was the only criterion

Page 6

M. Coignard et al. / Preventive Veterinary Medicine 112 (2013) 296– 308 301

Table 2Description of the number of French dairy farms included per stratum during the study period (December 2010–March 2011).

Lowlands (N = 65) Highlands (N = 65)

‘Small’ herds(n < 50)

‘Medium’ herds(n ≥ 50)

‘Small’ herds(n < 50)

‘Medium’ herds(n ≥ 50)

Straw yard (N = 54) Montbeliarde Milking parlor 11 6 10 6Holstein Milking parlor 7 5 5 5

Cubicles (N = 76) Montbeliarde Milking parlor 5 6 8 8Montbeliarde Automatic

milking system0 3 0 7

7

1

Holstein Milking parlor

Holstein Automaticmilking system

for which the maximum score of 100 was observed. Thecriteria ‘Absence of diseases’ and ‘Absence of pain inducedby management procedures’ obtained a median score lessthan 50, respectively equal to 33.3 and 28.0. Between thesetwo criteria, the highest variability was observed for thecriterion ‘Absence of diseases’ (mean ± S.D: 37.2 ± 15.4).‘Absence of pain induced by management procedures’was the criterion with the lowest value (minimumvalue = 2.0).

With regard to the 14 Welfare Quality® health meas-ures, more than half of the herds experienced no case ofsevere lameness, increased respiratory rates, ocular dis-charge, diarrhea or vulvar discharge. In contrast, cases of

Table 3Distribution and qualitative assessment of the overall health score (in bold), heaherds during the study period (December 2010-March 2011).

Variables Mean [S.D] Minimum Qua

Overall health score 33.4 [8.1] 13.9 28.3

Criterion-score ‘Absence of injuries’ 55.2 [21.5] 8.7 40.0Not lame cows (%) 85.4 [14.1] 26.9 79.3Moderately lame cows (%) 11.7 [10.9] 0.0 3.4Severely lame cows (%) 2.9 [5.0] 0.0 0.0Cows with no lesion (%) 41.3 [30.2] 0.0 11.1Cows with hairless patch but no lesion (%) 19.5 [16.4] 0.0 5.9Cows with at least one lesion (%) 39.2 [32.1] 0.0 10.0

Criterion-score ‘Absence of diseases’ 37.4 [15.5] 11.9 27.4Mortality (%) 3.3 [3.4] 0.0 1.0Subclinical mastitis (%) 20.6 [9.3] 2.0 14.0Vulvar discharge (%) 0.4 [1.4] 0.0 0.0Dystocia (%) 5.8 [5.6] 0.0 2.0Downer cows (%) 4.8 [4.5] 0.0 2.0Diarrhea (%) 11.2 [19.8] 0.0 0.0Frequency of coughing (/cow/15 min) 0.1 [0.1] 0.0 0.0Cows with increased respiratory rate (%) 0.0 [0.3] 0.0 0.0Cows with nasal discharge (%) 16.4 [15.5] 0.0 2.4Cows with ocular discharge (%) 2.8 [6.1] 0.0 0.0

Criterion-score ‘Absence of pain induced bymanagement procedures’

32.8 [12.6] 2.0 28.0

Partial score for disbudding/dehorning 32.8 [12.5] 2.0 28.0Partial score for tail-dockinge 99.1 [9.5] 3.0 –

* Good: score >55 or no warning or alarm for the measure; Medium: 55 ≤ scor≤20 or at least one alarm for the measure.

a Assessment for partial score for lameness.b Assessment for partial score for integument alterations.c Assessment for respiratory problems (coughing and increased respiratory ratd Assessment for ocular and respiratory area (nasal discharge and ocular dischae All herds obtained a partial score for tail-docking equal to 100 excepted one w

6 8 58 0 4

subclinical mastitis and painful procedures for dehorningwere noticed in all the herds.

3.2.2. Qualitative assessmentIn terms of the three categories defined to assess the

overall health score (Welfare Quality®, 2009), no herd wasclassified in the ‘Good’ category for the overall health score(score above 55), and the vast majority (95.4%) fell in theintermediate category, ‘Moderate’ (score between 21 and

55).

Among the three criteria considered, ‘Absence ofinjuries’ was the only one for which almost 50.0% of herdswere classified as ‘Good’. In contrast, only 15.0% of herds

lth criterion-scores and health measures (in italics) in 130 French dairy

rtile 1 Median Quartile 3 Maximum Qualitative assessment*

(% of herds)

Good Moderate Poor

33.0 38.3 54.2 0.0 95.4 4.6

54.6 69.9 100.0 49.2 46.9 3.8 89.7 93.9 100.0 66.9a 31.5a 1.5a

9.2 17.1 48.6 0.0 3.6 34.6 40.5 65.0 100.0 31.5b 31.5b 36.9b

15.9 29.0 62.5 33.3 62.2 100.0

33.3 44.8 86.0 15.4 70.0 14.6 2.0 5.0 17.0 54.7 19.2 26.1 19.5 25.0 46.0 6.2 34.6 59.2 0.0 0.0 11.4 91.5 5.4 3.1 5.0 8.0 25.0 33.1 23.8 43.1 4.0 7.0 20.0 41.6 26.1 32.3 0.0 16.1 84.6 62.3 7.7 30.0 0.1 0.2 0.9 100.0c 0.0c 0.0c

0.0 0.0 2.8 14.5 28.6 58.3 34.6d 7.7d 57.7d

0.0 3.0 45.7

28.0 28.0 75.0 2.3 83.8 13.8

28.0 28.0 75.0 2.3 83.8 13.8– – 100.0 99.2 0.0 0.8

e < 20 or at least one warning and no alarm for the measure; Poor: score

e measures).rge measures).hich obtained a partial score of 3.

Page 7

302 M. Coignard et al. / Preventive Veterinary

Fd

s2don

variables and two interactions were retained in the final

TUM

fl

ig. 2. Distribution of the overall health score in 130 French dairy herdsuring the study period (December 2010–March 2011).

cored ‘Good’ for the criterion ‘Absence of diseases’ and.0% for ‘Absence of pain induced by management proce-

ures’. Three measures mainly contributed to the loweringf the score for ‘Absence of diseases’: subclinical mastitis,asal discharge and dystocia. The score for ‘Absence of pain

able 4nivariable associations between explanatory variables and the overall health scarch 2011).

Explanatory variables (N = 13) Category

Stratification criteria (N = 4)Breed* Montbeliarde

Holstein

Farm location* Highlands

Lowlands

Milking system–Housing system* Milking parlor-straw yard

Milking parlor-cubicles

AMS-cubicles

Herd size* <50 cows

≥50 cows

Welfare Quality®measures collected through the Welfare Quality® protocol (N = 5)Average indoor housing durationper year*

≤105 days

105–135 days

136–165 days

>165 days

Herd score for absence ofprolonged thirst*

<20

20–99

100

Proportion of cows with abnormalbody condition score per herd*

≤35%

>35%

Proportion of dirty cows per herda,* ≤65%

>65%

Herd score for positive emotionalstate

≤34

35–49

50–67

>67

Measures related to herd management practices (N = 4)Herd average milk productionadjusted for breed* (kg/lactation)*

HOb: ≤10,200; MBb: ≤8000

HOb: >10,200; MBb: >8000

Herd average parity adjusted forbreed*

HOb:≤2; MBb: ≤2.5

HOb:]2–2.5]; MBb:]2.5–3]

HOb: >2.5; MBb: >3

Cumulative indoor housingduration at the day of visit*

≤165 days

>165 days

Proportion of cows per herd with days inmilk >200 at the day of visit*

≤38%

>38%

a Proportion of cows per herd with at least a dirty body part among the threeank/upper legs).b HO, Holstein; MB, Montbeliarde.* Variables selected for multivariable model (P ≤ 0.25).

Medicine 112 (2013) 296– 308

induced by management procedures’ was the lowest dueexclusively to inappropriate procedures for dehorning.

3.3. Factors of variation of the overall health score andits components

Among the 13 explanatory variables tested, 12 wereselected (P ≤ 0.25) with a variation of the overall healthscore during the univariable step (Table 4). Of these 12 vari-ables, 2 (average housing duration per year and cumulativehousing duration at the day of visit) were not includedin the multivariable analysis because they were highlycorrelated to ‘farm location’ and ‘milking system–housingsystem’. Finally, 10 variables and 2 interactions (‘Breed*Milking system–housing system’ and ‘Herd average milkproduction adjusted on breed*Milking system–Housingsystem’) were included in the multivariable analysis. Six

multivariable model (Table 5). The overall health score wasbetter in herds with the following criteria: milking par-lor and straw yard (P < 0.0001), a proportion of dirty cows

ore in 130 French dairy herds during the study period (December 2010-

N farms Average overall health score [S.D] P-value

70 35.0 [7.4] 0.01660 31.5 [8.6]65 34.5 [8.3] 0.05365 31.8 [7.6]54 36.6 [7.3] 0.000253 32.1 [7.9]23 28.9 [8.1]61 34.4 [7.3] 0.1669 32.4 [8.7]

24 31.1 [7.6] 0.0526 32.9 [6.7]45 35.8 [8.5]26 31.0 [8.7]33 35.0 [6.6] 0.1958 32.1 [8.7]39 33.7 [8.8]94 34.1 [8.2] 0.0936 31.4 [7.8]20 37.6 [7.9] 0.01

110 32.6 [8.0]32 34.2 [7.5] 0.6034 32.4 [9.9]34 32.4 [7.9]30 34.4 [7.3]

96 33.9 [8.4] 0.1432 31.6 [7.2]13 30.7 [9.4] 0.1175 34.6 [8.5]42 31.9 [6.7]96 34.4 [8.3] 0.01532 30.6 [7.1]32 36.2 [7.3] 0.01698 32.4 [8.2]

taken into account in the Welfare Quality® protocol (hind legs, udder,

Page 8

M. Coignard et al. / Preventive Veterinary Medicine 112 (2013) 296– 308 303

Table 5Final multivariable model for variables associated with a variation of the overall health score (R2 = 0.37) in 130 French dairy herds during the study period(December 2010–March 2011).

Variables Adjustedmeans ‘overallhealth score’

95% CI P-value

Breed 0.006Montbeliarde 34.4 [31.8–37.0]Holstein 31.3 [28.6–33.9]

Farm location 0.013Highlands 34.5 [31.5–37.4]Lowlands 31.2 [28.7–33.7]

Milking system–Housing system <0.0001Milking parlor-straw yard 38.4 [35.9–41.0]a

Milking parlor-cubicles 29.9 [26.6–33.2]b

AMS-cubicles 30.1 [26.8–33.5]b

Herd average parity adjusted onbreed

0.03

HO: ≤2; MB: ≤2.5 30.6 [26.4–34.8]a

HO:]2–2.5]; MB:]2.5–3] 35.4 [32.9–37.9]b

HO: >2.5; MB: >3 32.4 [29.3–35.6]a

Proportion of cows withabnormal body condition scoreper herd

0.04

≤35% 34.3 [32.0–36.7]>35% 31.4 [28.2–34.5]

Proportion of dirty cows per herd 0.002≤65% 36.1 [32.0–40.2]>65% 29.6 [27.7–31.4]

Herd average milk productionadjusted for breed(kg/lactation)

0.81

HO: ≤10,200; MB: ≤8000 32.7 [30.1–35.2]HO: >10,200; MB: >8000 33.0 [30.1–35.9]

Breed*Milking system–Housingsystem

0.03

Montbeliarde*Milkingparlor-straw yard

38.2 [35.0–41.3]a

Montbeliarde*Milkingparlor-cubicles

31.0 [27.3–34.8]b

Montbeliarde*AMS-cubicles 33.9 [29.9–38.0]ab

Holstein*Milking parlor-strawyard

38.7 [35.4–42.1]a

Holstein*Milking parlor-cubicles 28.8 [25.0–32.6]b′

Holstein*AMS-cubicles 26.3 [22.6–30.1]b’

Herd average milk productionadjusted for breed*Milkingsystem–Housing system

0.06

HO: ≤10,200; MB: ≤8000*Milkingparlor-straw yard

37.4 [34.5–40.2]a

HO: ≤10,200; MB: ≤8000*Milkingparlor-cubicles

32.0 [28.8–35.1]b

HO: ≤10,200; MB:≤8000*AMS-cubicles

28.6 [24.6–32.6]b

HO: >10,200; MB: >8000*Milkingparlor-straw yard

39.5 [35.9–43.1]a′

HO: >10,200; MB: >8000*Milkingparlor-cubicles

27.8 [22.8–32.8]b′

HO: >10,200; MB: 31.7 [27.8–35.6]b′

0); mod

>8000*AMS-cubicles

a,bFor each variable, modalities values with significant differences (P < 0.1

lower than or equal to 65% (P = 0.002), Montbeliarde breed(P = 0.006), an average parity between 2 and 2.5 for Holsteinand between 2.5 and 3 for Montbeliarde (P = 0.03), locationin highlands (P = 0.013) and a proportion of cows with an

alities values with no significant differences (P > 0.10).

abnormal body condition score (either too fat or too lean)lower than or equal to 35% (P = 0.04). The two interactionsretained showed that the overall health score was betterin herds with straw yard regardless their breed and their

Page 9

3 eterinary Medicine 112 (2013) 296– 308

acsoottlcbmorchwRita6

4

bQtiserOaifcbfistev(swmes

odiom(Qpci

ssoc

iati

ons

betw

een

fact

ors

of

vari

atio

n

of

the

over

all h

ealt

h

scor

e

and

the

freq

uen

cy

of

hea

lth

mea

sure

s

per

her

d

in

130

Fren

ch

dai

ry

her

ds

du

rin

g

the

stu

dy

per

iod

(Dec

embe

r

2010

–Mar

ch

2011

).

Lam

enes

s*In

tegu

men

tle

sion

s*M

orta

lity

Sucl

inic

alm

asti

tis

Vu

lvar

dis

char

geD

ysto

cia

Dow

ner

cow

sD

iarr

hea

Cou

ghin

g

Incr

ease

dre

spir

ator

y

rate

Nas

al

dis

char

ge

Ocu

lar

dis

char

ge

–aH

O

>

MB

HO

>

MB

HO

>

MB

–

–

HO

>

MB

–

–

–

–

HO

>

MB

nc

–

–

HI >

LO

–

–

– –

–

–

–

LO

>

HI

LO

>

HI

em–H

ousi

ng

MP-

C,

AM

S-C

>

MP-

SM

P-C

,A

MS-

C

>

MP-

SA

MS-

C

>

MP-

S

–

–

– –

AM

S-C

>

MP-

SA

MS-

C

>

MP-

C

–

AM

S-C

>

MP-

S

–

e

par

ity

adju

sted

on–

–

3

>

2

3,2

>

1

–

–

–

–

–

–

–

–

f cow

s

wit

h

bod

y

con

dit

ion

scor

eH

>

L

H

>

L

H

>

L

–

– –

–

–

H

>

L

–

–

H

>

L

f dir

ty

cow

s

per

–

–

–

H

>

L –

–

–

H

>

L

–

–

H

>

L

H

>

L

e

of

lam

e

cow

s

incl

ud

ing

mod

erat

ely

and

seve

rely

lam

e;

Perc

enta

ge

of

cow

s

wit

h

lesi

on

incl

ud

ing

hai

rles

s

pat

ches

and

lesi

ons.

fica

nt

(P

>

0.10

).tb

elia

rde;

HO

, Hol

stei

n.

nd

s;

LO, l

owla

nd

s.g

par

lor-

stra

w

yard

;

2:

mil

kin

g

par

lor-

cubi

cles

;

3:

auto

mat

ic

mil

kin

g

syst

em-c

ubi

cles

.in

≤

2;

Mon

tbel

iard

e

≤

2.5;

2:

Hol

stei

n]2

–2.5

],

Mon

tbel

iard

e]2.

5–3]

;

3:

Hol

stei

n

>2.5

, Mon

tbel

iard

e

>

3.35

%);

H:

hig

h

(>35

%).

65%

);

H:

hig

h

(>65

%).

04 M. Coignard et al. / Preventive V

verage milk production. However, the presence of cubi-les was associated with a decrease of the overall healthcore, especially for higher-yielding herds (P = 0.06). More-ver, the presence of AMS was associated with the lowestverall health score in Holstein herds (P = 0.03). Some ofhese variables were also associated with the frequency ofhe health measures (Table 6). Indeed, the prevalence ofameness and integument lesions were higher in herds withubicles and with a proportion of cows with an abnormalody condition score higher than 35%. Moreover, cows hadore integument lesions in Holstein herds. The incidence

f mortality was higher in herds with the following crite-ia: Holstein breed, location in Highlands, with AMS andubicles, an average parity higher than 2.5 for Holstein andigher than 3 for Montbeliarde, and a proportion of cowsith an abnormal body condition score higher than 35%.egarding subclinical mastitis, the prevalence was higher

n Holstein herds, in herds with an average parity higherhan 2 for Holstein and higher than 2.5 for Montbeliardend in herds with a proportion of dirty cows higher than5%.

. Discussion

The first objective of our study was to describe the distri-ution of health scores that were derived from the Welfareuality® assessment protocol in French dairy herds. To

he best of our knowledge, this is the first study aim-ng to describe the overall health of dairy cows at such acale in a country. The farms were randomly sampled fromxhaustive lists and were classified into strata selected toepresent the main sustainable dairy systems in France.ur results are thus assumed to be representative of over-ll health in French dairy herds, but have to be confirmedn other areas or breeds. We did not investigate tie-stallarms, which are scarce in France. Further studies in otherountries where the tie-stall system is predominant woulde advisable. In order to avoid that the willingness of thearmers to participate in the study as well as their behav-or during the visit were affected by the objective of thetudy, the farmers were told that the study aimed at bet-er understanding cows ‘comfort and describing housingquipment when they were contacted by phone. Thus, theast majority of refusals were motivated by two reasons:i) the farmers could not receive us for a full day during thetudy period, (ii) we rejected the participation of farmershen their stratum was already filled. Despite our recruit-ent method based on a dummy objective, we cannot

xclude a putative selection bias as in any epidemiologicaltudy.

The Welfare Quality® assessment protocol allows anverall assessment of health based on a combination ofifferent measures considered by experts of welfare as

mpairing welfare. In humans, Health is defined as a statef complete physical, mental and social well-being, noterely negatively as the absence of diseases or infirmity

World Health Organization, 1946). Through the Welfare

uality® protocol, we can consider that the same com-onents in terms of both physical and mental painfulonditions are taken into account such as injuries, phys-cal painful diseases and also those inducing stressful Ta

ble

6U

niv

aria

ble

a

Var

iabl

es

Bre

edb

Farm

loca

tio

Mil

kin

g

syst

syst

emd

Her

d

aver

agbr

eed

e

Prop

orti

on

oab

nor

mal

per

her

df

Prop

orti

on

oh

erd

g

*Pe

rcen

tag

aN

ot

sign

ib

MB

, Mon

cH

I,

hig

hla

d1:

mil

kin

e1:

Hol

ste

fL:

low

(≤g

L:

low

(≤

Page 10

eterinary

M. Coignard et al. / Preventive V

conditions as downer cows and painful procedures suchas dehorning. Thus, the most frequent health disordersin dairy cows (mastitis, lameness, dystocia, downer cowsyndrome, diarrhea, metritis) are investigated (Fourichonet al., 2001b) through this protocol, but it also takesinto account uncommon health measures such as bovinerespiratory disorders in adults (cough, nasal dischargeand hampered respiration) which are usually consideredmainly in relation to calves. All of the health disordersconsidered are likely to produce pain, which makes themvery relevant when assessing animal welfare. For instance,while the classic threshold reported in the literature forsubclinical mastitis is 200,000 cell/mL (Schepers et al.,1997; Brolund, 1985), the choice was made by expertsto consider a higher threshold (400,000 cell/mL) in theWelfare Quality® assessment protocol to focus on sub-clinical mastitis supposed to induce pain (Eshraghi et al.,1999; Milne et al., 2003). In the same way, while lame-ness is usually considered as a disease, in this protocol,this health disorder is associated with the prevalence ofintegument alterations to indicate the presence of painrelated to injuries. Management procedures for disbud-ding/dehorning are also considered to assess the painrelated to these practices. Therefore, the implementation ofthe Welfare Quality® assessment protocol on farms duringa single one-day visit allows an accurate overall assessmentof health based on a large set of classical and unconven-tional measures of disorders likely to induce pain.

The prevalence that we found for most of the health dis-orders investigated were in the range of the data reportedin the literature. For example, our result for dystocia (5.7%)is close to that published in France (6.6%, Fourichon et al.,2001b) and in Ireland (2–7%, Mee, 2008). The annualcumulative incidence of mortality in this study (3.2%)corresponds to that reported in many countries (1–5%,Thomsen and Houe, 2006). The average percentage of lamecows per herd (14.7%) is lower than that published in theUK (22%, Whay et al., 2003). For other diseases, it is difficultto make further comparisons with the literature due to dif-ferent diagnostic methods. As an example, a threshold of400,000 cells/mL was retained to estimate the prevalenceof subclinical mastitis while a threshold of 200,000 cells/mLis classically found in the literature. However, in our sam-ple, an average of 20.6% of cow somatic cell count readingswere above or equal to 400,000 cells/mL which is consis-tent with the findings of Madouasse et al. (2010) in which17.1% of cow somatic cell count were above this threshold.Regarding vulvar discharge, our present findings cannot becompared with prevalence reported in the literature whichare focused on metritis and not only on vulvar discharge(Bruun et al., 2002). Finally, the scarcity of studies inves-tigating bovine respiratory disorders in adult cows doesnot allow a comparison with our results. However, amongthe possible comparisons, the similarities between the fre-quencies of health disorders found in the present studyand those of studies conducted elsewhere in Europe sup-port the validity of our results and measures. Despite these

similarities, a moderate overall health score was obtainedon 95.4% of the farms visited. Although greater variationwas observed among measures, the overall score varied lit-tle between farms (ranged 13.9–54.2) due to the way this

Medicine 112 (2013) 296– 308 305

score is calculated in the Welfare Quality® assessment pro-tocol. The calculations are also designed in such a way thatthe overall score is always closer to the minimum value ofthe lowest criterion-scores (Welfare Quality®, 2009), cor-responding to ‘Absence of pain induced by managementprocedures’ in our study.

When moderate and poor categories were combined,the worst assessments were obtained for pain inducedby disbudding/dehorning and some health disorders(subclinical mastitis, dystocia). In the Welfare Quality®

assessment protocol, two different approaches were usedby experts to produce the assessment of pain induced bymanagement procedures and the assessment of disease.The assessment of pain induced by management proce-dures was based on the recommendations of a panel ofexperts regarding the data to take into account whenassessing pain induced by human intervention. The dataobtained are not frequencies but qualitative answers toquestions such as the method used to disbud/dehorn orwhether anesthetics and/or analgesics are used. In our sam-ple, only 2.3% of the herds were classified as “good” for thiscriterion. This result is partly explained by the fact that cur-rent French regulations do not allow farmers to use localanesthesia while disbudding in calves is performed essen-tially by farmers (Kling-Eveillard et al., 2009). However, theuse of analgesics is not forbidden and the implementationof Nonsteroidal Anti-Inflammatory Drugs (NSAIDs) shouldbe promoted among farmers. The assessment of disease ismore difficult than the assessment of pain because it ishard to say what should be considered to be an accept-able number of diseased cows in a herd. The thresholds foralerts and alarms were defined in Welfare Quality® by apanel of experts which included veterinarians before theprotocol was implemented on a large sample. The thresh-old for alarms corresponds to the percentage of diseasedanimals that triggers the implementation of a health planon a farm. It could be that these Welfare Quality®’ thresh-olds are not really suitable to allow achievable goals ofimprovement. Indeed, in the present study, the imple-mentation of the Welfare Quality® assessment protocolprovides a description of the health disorders that seemsin accordance with previous reports in the literature, bothin France and elsewhere. However, over 50% of the herdssurpassed the threshold level (requiring intervention) fortwo measures (subclinical mastitis and nasal discharge),and 25% of the herds for six measures (subclinical mas-titis and nasal discharge plus mortality, dystocia, downercows and diarrhea). It would be interesting to compare ourresults with those from other European countries. Indeed,if these results are similar between countries, this wouldimply that a health control plan is required for a consider-able number of farms, which would be very expensive toimplement in practice. Moreover, at farm level, when weapply a health control plan which aims to reduce the per-centage of diseased cows until the low Welfare Quality®

alarm threshold under the assumption of its technical fea-sibility, we cannot be sure that the benefit provided by theimprovement would balance the health control plan cost in

all cases. Another approach would be to test the relevanceof these alarm thresholds by looking at the distributionof the frequency of diseases after the implementation of

Page 11

3 eterinary

tscestoeriepg2tiWb

tDtstfhedrwftcmicccWcfiy22atmtwitrsswcwie1H

higher risk for several diseases such as mastitis (Steeneveld

06 M. Coignard et al. / Preventive V

he Welfare Quality® protocol in a large sample to target amaller proportion of farms in which herd health programsould be effectively implemented (e.g. frequency of dis-ased cows above the 3rd quartile). A similar approach wasuggested by Fourichon et al. (2001b). Therefore, this pro-ocol is just one way of looking at the level of health basedn the opinion of a selected group of experts. It cannot bexcluded that the consultation of other experts may haveesulted in a different conception of health and thereforen different weighting of the measures, leading to differ-nt health scores. However, within the Welfare Quality®

rotocol, there was a low variability between the scoresiven by experts from a specific data set (Veissier et al.,010). Unfortunately, sensitivity analysis to examine howhe experts’ values change in response to different scenar-os in farms and to possibly adjust the parameters of the

elfare Quality®’s integration model were not conductedy the developers of the method.

The second objective of the study was to investigatehe putative factors of variation in overall health score.espite the low variability of the overall health score in

he sample resulting from the non-compensatory principle,everal factors of variation were identified. As expected,he combination, ‘housing system–milking system’ was theactor that showed the greatest variation in the overallealth score in the sample. The score was better in herdsquipped with straw yards and milking parlors indepen-ently of the herd average milk production. However, theeduction of the overall health score in herds equippedith cubicles especially in higher-yielding herds and there-

ore potentially in herds with the larger cows suggest thathe larger a cow is, the more the available space in theubicle is limited and the more the adaptation of the ani-al in this housing system is difficult. These results are

n accordance with previous studies reporting that cubi-le dimensions and design influence the health of dairyows and therefore should be considered together withow size to prevent health disorders (Zurbrigg et al., 2005).

hile contradictory results have been obtained for cubi-les and straw yards in the literature, with some studiesnding less lameness and integument alterations in strawards (Haskell et al., 2006; Webster, 2002; Livesey et al.,002), and less mastitis in cubicles (Fregonesi and Leaver,001; Bareille et al., 1998), our results suggest that over-ll health is significantly better in straw yards mainly dueo a lower frequency of lameness, integument alterations,

ortality, diarrhea and nasal discharge. Further investiga-ions should be carried out to determine if scores of otherelfare component (linked to housing, feeding, and behav-

or) are also better in straw yards compared with cubicles. Ifhis is the case, the important development of cubicles cur-ently taking place in France in response to increasing herdizes should be reconsidered. Furthermore, overall healthcores varied within the same housing system (cubicles),ith a trend for lower scores obtained in herds with AMS

ompared with milking parlors. Several factors associatedith AMS could explain this result: a reduction of the graz-

ng period, known for reducing the level of health (Regula

t al., 2004) by increasing the risk for lameness (Smits et al.,992; Gitau et al., 1996) and mastitis (Bendixen et al., 1986).owever, a farmer using such an automatic system could be

Medicine 112 (2013) 296– 308

able to detect earlier diseases by detecting changes withinthe individual cow (Jacobs and Siegford, 2012), and par-ticularly mastitis. However, this milking system is oftenadopted in large farms in which the time devoted to animalobservation may be reduced (Fahey et al., 2002). Furtherinvestigations should be undertaken to confirm our result.

Interestingly, Holstein cows were less healthy thanMontbeliardes in herds equipped with AMS. The Holsteinbreed was also found to be associated with a lower scorefor overall health. This result and the results at the healthmeasure level in this study are in agreement with the liter-ature in which Hoslteins are reported to be at higher risk formany diseases investigated through the Welfare Quality®

assessment protocol, such as subclinical mastitis (Busatoet al., 2000), integument alterations (Potterton et al., 2011)and mortality (Thomsen et al., 2006). The higher risk forother diseases in Holsteins such as dystocia (Heins et al.,2006) and lameness (Faye and Barnouin, 1988) have beenreported by previous authors but have not been identifiedin this study. The Holstein breed is the most specialized inmilk production whereas Montbeliarde breed has a dualpurpose (Walsh et al., 2008). Our results therefore supportthe relationship advanced by previous studies showing thathigh genetic merit for milk is associated with higher diseasesusceptibility (Pryce and Veerkamp, 2001; Dematawewaand Berger, 1998).

Some management practices were also found to be asso-ciated with different overall health scores. As expected,a lower overall health score was observed in herds witha higher proportion of dirty cows. Indeed, our results atthe health measure level are in accordance with previousstudies reporting that dirtiness is a risk factor for mastitis(Schreiner and Ruegg, 2003; Reneau et al., 2005; Sant’Annaand Paranhos da Costa, 2011). The higher risk for lame-ness has also been reported in others studies (Cook andNordlund, 2009; Relun et al., 2012). In addition, lowerscores for overall health were observed in herds with ahigher proportion of cows with an abnormal conditionscore (either too fat or too lean). At the measure level,our results support previous studies indicating that cowswhich are too fat or too lean are more likely to be lame(Gearhart et al., 1990; Bicalho et al., 2009). Moreover, otherpublications have reported that such an abnormal condi-tion score is also associated with dystocia (Zaborski et al.,2009), elevated somatic cell counts (Breen et al., 2009),and that lean cows are more at risk for metritis (Heueret al., 1999; Hoedemaker et al., 2009). Standard recom-mendations focused on hygiene and good feeding practicestherefore could contribute to an overall improvement ofhealth.

Interestingly, the overall health score was lower in‘young herds’, i.e. herds with an average parity below 2.5for Montbeliardes and below 2 for Holsteins. The oppositeresult could have been expected given the fact that prim-iparous cows are at a higher risk only for dystocia (Heinset al., 2006) and metritis (Bruun et al., 2002; Ghavi Hossein-Zadeh and Ardalan, 2011), while multiparous cows are at a

et al., 2008; Breen et al., 2009) as shown our results,lameness (Groehn et al., 1992), milk fever (Erb et al., 1985)and in addition mortality as demonstrated in this study.

Page 12

eterinary

M. Coignard et al. / Preventive V

Furthermore, a farmer may decide to increase the cullingrate to only keep healthy cows. Given our findings, wecan assume that herds with a high proportion of prim-iparous cows face a high culling rate, particularly theanticipated culling of primiparous or second lactationcows. This could be the sign of herds with a high preva-lence of several diseases, even in primiparous cows.The average parity of a herd therefore might be usedas an indicator for the detection of farms with healthproblems.

The percentage of variability explained by our model(R2 = 0.37) suggested that some factors, such as factors spe-cific to the farm itself, play an important role in the overallhealth score of a herd. Within the same system, differentfarmers will manage routine practices differently, leadingto heterogeneous results in terms of health and productionperformances (Faye, 1991). This suggests that the farmeris at least as important as the system itself (Regula et al.,2004; Dawkins et al., 2004).

5. Conclusion

This study shows that the overall health of dairy cows,defined by an assessment based on experts’ opinion, wasmoderate but ranged with the farming system. Indeed,the health score was higher in herds housed in a strawyard, milked in a parlor, located in highlands, or in whichthe dominant breed is Montbeliard, compared respectivelyto herds housed in cubicles, milked by a AMS, located inlowlands, or in which the dominant breed is Holstein. Inaddition, within farming systems, a high early culling rate,a high proportion of dirty cows or of cows which wereeither too fat or too lean, were factors associated with loweroverall health scores. Due to a high variability of the healthscore within a given system, these findings suggest that theimprovement of health in the short term can be obtainedthrough the modification of routine management practicesin all types of farming systems. Over the longer term, suchstudy results could be used to guide the choice of dairyproduction systems by promoting those associated with abetter health status.

Acknowledgments

We thank Danone Reseach for their financial support ofthis work. We are also very grateful to farmers involved inthis study and to 9 Animal Health Services for providinglist of farms. Eric Delval (INRA, Saint-Genès Champanelle,France), Christophe Mallet (INRA, Saint-Genès Champan-elle, France) and Rémi Debauchez (ISARA, Lyon, France)are gratefully acknowledged for their help in data collec-tion, as Jean-Yves Audiart (ONIRIS, Nantes, France) andDidier Billon (ONIRIS, Nantes, France) for their contribu-tion with data entry. Finally, we thank Anne Lamadon,Yoan Gaudron and Pascal Champciaux (INRA, Saint-Genès

Champanelle, France) for their contribution in this studyregarding the calculation of health scores and AurélienMadouasse (ONIRIS, Nantes, France) for reviewing theEnglish of this manuscript.

Medicine 112 (2013) 296– 308 307

References

Bareille, N., Seegers, H., Fourichon, C., Beaudeau, F., Malher, X., 1998.Survenue et expression des mammites cliniques et subcliniques entroupeaux bovins laitiers: facteurs de risque liés à la conception et àl’utilisation du bâtiment (Housing as risk factor for clinical and sub-clinical mastitis in dairy herds). Renc. Rech. Rum. 5, 297–300.

Bendixen, P.H., Vilson, B., Ekesbo, I., Astrand, D.B., 1986. Disease frequen-cies of tied zero-grazing dairy cows and of dairy cows on pastureduring summer and tied during winter. Prev. Vet. Med. 4, 291–306.

Bicalho, R.C., Machado, V.S., Caixeta, L.S., 2009. Lameness in dairy cattle: adebilitating disease or a disease of debilitated cattle? A cross-sectionalstudy of lameness prevalence and thickness of the digital cushion. J.Dairy Sci. 92, 3175–3184.

Breen, J.E., Bradley, A.J., Green, M.J., 2009. Quarter and cow risk factorsassociated with a somatic cell count greater than 199,000 cells permilliliter in United Kingdom dairy cows. J. Dairy Sci. 92, 3106–3115.

Botreau, R., 2008. Evaluation multicritère du bien-être animal. Exemplesdes vaches laitières en ferme. Institut des Sciences et Industries duVivant et de l’Environnement (Agro Paris Tech), pp. 436 (Ph.D. Thesis).

Brolund, L., 1985. Cell counts in bovine milk: causes of variation andapplicability for diagnosis of subclinical mastitis. Acta. Vet. Scand. 80,1–123.

Bruun, J., Ersboll, A.K., Alban, L., 2002. Risk factors for metritis in Danishdairy cows. Prev. Vet. Med. 54, 179–190.

Busato, A., Trachsel, P., Schällibaum, M., Blum, J.W., 2000. Udder healthand risk factors for subclinical mastitis in organic dairy farms inSwitzerland. Prev. Vet. Med. 44, 205–220.

Cook, N.B., Nordlund, K.V., 2009. The influence of the environment on dairycow behavior, claw health and herd lameness dynamics. Vet. J. 179,360–369.

Dawkins, M.S., Donnelly, C.A., Jones, T.A., 2004. Chicken welfare is influ-enced more by housing conditions than by stocking density. Nature427, 342–344.

Dematawewa, C.M., Berger, P.J., 1998. Genetic and phenotypic parametersfor 305-days yield, fertility, and survival in Holsteins. J. Dairy Sci. 81,2700–2709.

EFSA Reports, 2009. Scientific Opinion of the Panel on Animal Health andWelfare on a request from European Commission on welfare of dairycows. Effects of farming systems on dairy cow welfare and disease.EFSA J. 1143, 1–38.

Erb, H.N., Smith, R.D., Oltenacu, P.A., Guard, C.L., Hillman, R.B., Powers,P.A., Smith, M.C., White, M.E., 1985. Path model of reproductive disor-ders and performance, milk fever, mastitis, milk yield, and culling inHolstein cows. J. Dairy Sci. 68, 3337–3349.

Eshraghi, H.R., Zeitlin, I.J., Fitzpatrick, J.L., Ternent, H., Logue, D.,1999. The release of bradykinin in bovine mastitis. Life Sci. 64,1675–1687.

European Commission, 2005. Attitudes of consumers towards the wel-fare of farmed animals. Spec. Eur. 229, http://ec.europa.eu/food/animal/welfare/euro barometer en.pdf

Fahey, J., O’Sullivan, K., Crilly, J., Mee, J.F., 2002. The effect of feeding andmanagement practices on calving rate in dairy herds. Anim. Reprod.Sci. 74, 133–150.

Farm Animal welfare Council, 1992. FAWC updates the five freedoms. Vet.Rec. 17, 357.

Faye, B., 1991. Interrelationships between health status and farm man-agement system in French dairy herds. Prev. Vet. Med. 12, 133–152.

Faye, B., Barnouin, J., 1988. Les boiteries chez la vache laitière. Synthèse desrésultats de l’enquête écopathologique continue. INRA Prod. Anim. 1,227–234.

Fourichon, C., Seegers, H., Beaudeau, F., Verfaille, L., Bareille, N., 2001a.Health-control costs in dairy farming systems in western France.Livest. Prod. Sci 68, 141–156.

Fourichon, C., Beaudeau, F., Bareille, N., Seegers, H., 2001b. Incidence ofhealth disorders in dairy farming systems in western France. Livest.Prod. Sci. 68, 157–170.

Fregonesi, J.A., Leaver, J.D., 2001. Behaviour, performance and healthindicators of welfare for dairy cows housed in strawyard or cubiclesystems. Livest. Prod. Sci. 68, 205–216.

Gearhart, M.A., Curtis, C.R., Erb, H.N., Smith, R.D., Sniffen, C.J., Chase, L.E.,Cooper, M.D., 1990. Relationship of changes in condition score to cowhealth in Holsteins. J. Dairy Sci. 73, 3132–3140.

Ghavi Hossein-Zadeh, N., Ardalan, M., 2011. Cow-specific risk factors for

retained placenta, metritis and clinical mastitis in Holstein cows. Vet.Res. Commun 35, 345–354.

Gitau, T., McDermott, J.J., Mbiuki, S.M., 1996. Prevalence, incidence andrisk factors for lameness in dairy cattle in small scale farms in kikuyuDivision, Kenya. Prev. Vet. Med. 28, 101–115.

Page 13

3 eterinary

G

G

H

H

H

H

J

K

K

L

L

M

M

M

P

P

R

R

R

08 M. Coignard et al. / Preventive V

reen, L.E., Hedges, V.H., Schukken, Y.H., Blowey, R.W., Packington, A.J.,2002. The impact of clinical lameness on the milk yield of dairy cows.J. Dairy Sci. 85, 2250–2256.

roehn, J.A., Kaneene, J.B., Foster, D., 1992. Risk factors associated withlameness in lactating dairy cattle in Michigan. Prev. Vet. Med. 14,77–85.

askell, M.J., Rennie, L.J., Bowell, V.A., Bell, M.J., Lawrence, A.B., 2006. Hous-ing system, milk production, and zero-grazing effects on lameness andleg injury in dairy cows. J. Dairy Sci. 89, 4259–4266.

eins, B.J., Hansen, L.B., Seykora, A.J., 2006. Calving difficulty and still-births of pure Holsteins versus crossbreds of Holstein with Normande,Montbeliarde, and Scandinavian Red. J. Dairy Sci. 89, 2805–2810.

euer, C., Schukken, Y.H., Dobbelaar, P., 1999. Postpartum body condi-tion score and results from first test day milk as predictors of disease,fertility, yield and culling in commercial dairy herds. J. Dairy Sci. 82,295–304.

oedemaker, M., Prange, D., Gundelach, Y., 2009. Body condition changeante- and postpartum, health and reproductive performance in Ger-man Holstein cows. Reprod. Domestic. Anim. 44, 167–173.

acobs, J.A., Siegford, J.M., 2012. Invited review: The impact of automaticmilking systems on dairy cow management, behavior, health, andwelfare. J. Dairy Sci. 95, 2227–2247.

ling-Eveillard, F., Dockès, A.C., Ribaud, D., Mirabito, L., 2009. Cattledehorning in France: farmers’ practices and attitudes. In: RencontreRecherche Ruminants symposium n◦16, Paris, France, pp. 249–252.

nierim, U., Winckler, C., 2009. On-farm welfare assessment in cattle:validity, reliability and feasibility issues and future perspectives withspecial regard to the Welfare Quality® approach. Anim. Welfare 18,451–458.

abreuche, C., Grabich, M., 2003. The Choquet integral for the aggregationof interval scales in multicriteria decision making. Fuzzy. Set. Syst.137, 11–26.

ivesey, C.T., Marsh, C., Metcalf, J.A., Laven, R.A., 2002. Hock injuries incattle kept in straw yards or cubicles and rubber mats or mattresses.Vet. Rec. 150, 677–679.

adouasse, A., Huxley, J.N., Browne, W.J., Bradley, A.J., Green, M.J., 2010.Somatic cell counts dynamics in a large sample of dairy herds inEngland and Wales. Prev. Vet. Med. 96, 56–64.

ee, J.F., 2008. Prevalence and risk factors for dystocia in dairy cattle. Areview. Vet. J. 176, 93–101.

ilne, M.H., Nolan, A.M., Cripps, P.J., Fitzpatrick, J.L., 2003. Assessment andalleviation of pain in dairy cows with clinical mastitis. Cattle Pract. 11,289–294.

otterton, S.L., Green, M.J., Harris, J., Millar, K.M., Whay, H.R., Huxley,J.N., 2011. Risk factors associated with hair loss, ulceration, andswelling at the hock in freestall-housed UK dairy herds. J. Dairy Sci.94, 2952–2963.

ryce, J.E., Veerkamp, R.F., 2001. The incorporation of fertility indicesin genetic improvement programmes. Br. Soc. Anim. Sci. 26,237–249.

Development Core Team, 2009. R Version 2.10.1.: A Language andEnvironment for Statistical Computing. R Foundation for Statisti-cal Computing, Vienna, Austria, Accessible on the network website:http://www.R-project.Org

egula, G., Danuser, B., Spycher, B., Wechler, B., 2004. Health and welfareof dairy cows in different husbandry systems in Switzerland. Prev. Vet.Med. 66, 247–264.

elun, A., Lehébel, A., Bareille, N., Guatteo, R., 2012. EstimationUsing Survival Analysis on the Relative Impact of Treatment and

Medicine 112 (2013) 296– 308

Management Factors on the Occurrence of Digital Dermatitis inDairy Cattle. The Society for Veterinary Epidemiology and PreventiveMedicine (SVEPM), Glasgow.

Reneau, J.K., Seykora, A.J., Heins, B.H., Endres, M.I., Farnsworth, R.J., Bey,R.F., 2005. Association between hygiene scores and somatic cell scoresin dairy cattle. J. Am. Vet. Med. Assoc. 227, 1297–1301.

Sant’Anna, A.C., Paranhos da Costa, M.J.R., 2011. The relationship betweendairy cow hygiene and somatic cell count in milk. J. Dairy Sci. 94,3835–3844.

Schepers, A.J., Lam, T.J.G.M., Schukken, Y.H., Wilmink, J.B.M., Hanekamp,W.J.A., 1997. Estimation of variance components for somatic cellcounts to determine thresholds for uninfected quarters. J. Dairy Sci.80, 1833–1840.

Schreiner, D.A., Ruegg, P.L., 2003. Relationship between udder and leghygiene scores and subclinical mastitis. J. Dairy Sci. 86, 3460–3465.

Seegers, H., Fourichon, C., Beaudeau, F., 2003. Production effects relatedto mastitis and mastitis economics in dairy cattle herds. Vet. Res. 34,475–491.

Smits, M.C.J., Frankena, K., Metz, J.H.M., Noordhuizen, J.P.T.M., 1992. Preva-lence of digital disorders in zero-grazing dairy farms. Livest. Prod. Sci.32, 231–244.

Steeneveld, W., Hogeveen, H., Barkema, H.W., Van den Broek, J., Huirne,R.B.M., 2008. The influence of cow factors on the incidence of clinicalmastitis in dairy cows. J. Dairy Sci. 91, 1391–1402.

Thomsen, P.T., Houe, H., 2006. Dairy cow mortality. a review. Vet. Q. 28,122–129.

Thomsen, P.T., Kjeldsen, A.M., Sorensen, J.T., Houe, H., Ersboll, A.K., 2006.Herd-level risk factors for the mortality of cows in Danish dairy herds.Vet. Rec. 158, 622–626.

Veissier, I., Botreau, R., Perny, P., 2010. Evaluation multicritère appliquéeau bien-être des animaux en ferme ou à l’abattoir: difficultés et solu-tions du projet Welfare Quality® . INRA Prod. Anim. 3, 269–284.

Von Keyserlingk, M.A.G., Rushen, J., de Passillé, A.M., Weary, D.M., 2009.Invited review: The welfare of dairy cattle-Key concepts and the roleof science. J. Dairy Sci. 92, 4101–4111.

Walsh, S., Buckley, F., Pierce, K., Byrne, N., Patton, J., Dillon, P., 2008. Effectsof breed and feeding system on milk production, body weight, bodycondition score, reproductive performance, and postpartum ovarianfunction. J. Dairy Sci. 91, 4401–4413.

Webster, A.J.F., 2002. Effect of housing practices on the development offood lesions in dairy heifers in early lactation. Vet. Rec. 151, 9–12.

Welfare Quality, 2009. Welfare Quality® Assessment Protocol for Cat-tle. Lelystad, The Netherlands, Accessible on the network website:http://www.welfarequality.net/network/45848/7/0/40

Whay, H.R., Main, D.C.J., Green, L.E., Webster, A.J.F., 2003. Assessmentof the welfare of dairy cattle using animal-based measurements:Direct observations and investigation of farm records. Vet. Rec. 153,197–202.

Whay, H.R., Waterman, A.E., Webster, A.J.F., 1997. Associations betweenlocomotion, claw lesions and nociceptive threshold in dairy heifersduring the peri-partum period. Vet. J. 154, 155–161.

World Health Organization, 1946. The Preamble of the Constitution of theWorld Health Organization. Bul. World Health Organ. 80, 981–982.

Zaborski, D., Grzesiak, W., Szatkowska, I., Dybus, A., Muszynska, M., Jedrze-

jczak, M., 2009. Review article: Factors affecting dystocia in cattle.Reprod. Domestic. Anim. 44, 540–551.

Zurbrigg, K., Kelton, D., Anderson, N., Millman, S., 2005. Tie-stall designand its relationship to lameness, injury, and cleanliness on 317 Ontariodairy farms. J. Dairy Sci. 88, 3201–3210.