Calderón Díaz et al. Porcine Health Management (2020) 6:40 https://doi.org/10.1186/s40813-020-00176-w RESEARCH Open Access A bio-economic simulation study on the association between key performance indicators and pluck lesions in Irish farrow- to-finish pig farms Julia Adriana Calderón Díaz1* , Maria Rodrigues da Costa1,2,3, Laurence Shalloo4, Jarkko K. Niemi5, Finola Catherine Leonard2, Daniel Crespo-Piazuelo1, Josep Gasa6 and Edgar García Manzanilla1,2 Abstract Background: Pluck lesions are associated with decreased performance in grower-finisher pigs, but their economic impact needs to be further investigated. This study aimed to identify the main pluck lesions and the cut-off value for their prevalence, associated with changes in average daily gain (ADG) during the wean-to-finish period, to simulate their effects on economic performance of farrow-to-finish farms. Pigs (n = 162 ± 51.9 per farm) from 56 farrow-to-finish farms were inspected at slaughter and the prevalence of enzootic pneumonia-like lesions, pleurisy, lung scars, abscesses, pericarditis, and liver milk spots was estimated. For each farm, annual performance indicators were obtained. Regression trees analysis (RTA) was used to identify pluck lesions and to estimate cut-off values for their prevalence associated with changes in ADG. Different scenarios were simulated as per RTA results and economic and risk analyses were performed using the Teagasc Pig Production Model. Risk analysis was performed by Monte Carlo sampling using the Microsoft Excel add-in @Risk with 10,000 iterations. Results: Pleurisy and lung scars were the main lesions associated with changes in ADG. Three scenarios were simulated based on RTA results: a 728 sow farrow-to-finish farm with prevalence of i) pleurisy < 25% and lung scars < 8% (LPLSC; ADG = 760 g); ii) pleurisy < 25% and lung scar ≥8% (LPHSC; ADG = 725 g) and iii) pleurisy ≥25% (HP; ADG = 671 g). The economic analysis showed increased feed and dead animals for disposal costs, and lower sales in the HP and LPHSC scenarios than in the LPLSC scenario; thereby reducing gross margin and net profit. Results from the risk analysis showed lower probability of reaching any given level of profit in the HP scenario compared with the LPHSC and LPLSC scenarios. Conclusion: Under the conditions of this study, higher prevalence of pleurisy and lung scars were associated with decreased ADG during the grower-finisher period and with lower economic return in the simulated farms. These results highlight the economic benefits and importance of preventing and/or controlling respiratory disease. Keywords: Economic modelling, Lung scars, Pig production systems, Pleurisy, Regression trees, Stochastic budgeting * Correspondence: Julia.CalderonDiaz@teagasc.ie; ja.calderondiaz@gmail.com 1 Pig Development Department, Teagasc Animal and Grassland Research and Innovation Centre, Moorepark, Fermoy, Co. Cork, Ireland Full list of author information is available at the end of the article © The Author(s). 2020 Open Access This article is licensed under a Creative Commons Attribution 4.0 International License, which permits use, sharing, adaptation, distribution and reproduction in any medium or format, as long as you give appropriate credit to the original author(s) and the source, provide a link to the Creative Commons licence, and indicate if changes were made. The images or other third party material in this article are included in the article's Creative Commons licence, unless indicated otherwise in a credit line to the material. If material is not included in the article's Creative Commons licence and your intended use is not permitted by statutory regulation or exceeds the permitted use, you will need to obtain permission directly from the copyright holder. To view a copy of this licence, visit http://creativecommons.org/licenses/by/4.0/. The Creative Commons Public Domain Dedication waiver (http://creativecommons.org/publicdomain/zero/1.0/) applies to the data made available in this article, unless otherwise stated in a credit line to the data. Calderón Díaz et al. Porcine Health Management (2020) 6:40 Page 2 of 15 Background economic impact of pleurisy in British pig herds and es- Assessment of pluck lesions can be used as a tool for timated a cost of £5 (approximately €5.8) per pig pro- on-farm disease surveillance as these lesions reflect re- duced. Their calculations were based on industry spiratory disease affecting pigs [1] and a high prevalence standard costs associated with 6% increase in mortality of pluck lesions could be indicative of farms with higher during the grower-finisher period, a prevalence of pleur- proportion of diseased animals [2]. Lesions in the lungs, isy of 20%, reduction in ADG of 50 g and reduction of heart, pleura and liver are the most important ones ob- feed conversion efficiency of 0.1. Ferraz et al. [6] esti- served during post-mortem examinations at the abattoir mated a cost of US$6.55 (approximately €5.5) per pig [3]. They represent lesions caused by respiratory patho- produced considering a reduction in ADG of 27 g associ- gens and parasite infestations that contribute to reduced ated with lesions (pneumonia and pleurisy) in ≥15% of performance [4–7] during the grower-finisher period. the lungs compared with pigs with no lesions. Doing Pneumonia and pleurisy are the most commonly ob- these calculations is not easy as there are other variables served lung lesions at the abattoir [8–10] and they are such as market prices fluctuations and the complex in- associated with an increased likelihood of observing terrelationships between different aspects of production other lung lesions such as abscesses and scars [3, 11]. to consider when using real-world farm-level data. Using Further, infestation by Ascaris suum, as the larvae mi- bio-economic stochastic modelling approaches can pro- grate through the liver and lungs, may cause both direct vide opportunities to consider all factors influencing the damage to the lungs and affect the immune response to economic performance and be flexible when analysing pathogens such as Mycoplasma hyopneumoniae [12], their impact. Indeed, bio-economic models describe the which could potentially increase prevalence of pneumo- associations between different biological and financial nia lesions. Lesions observed at the abattoir are those components of farming systems [22]. Also, by incorpor- developed closer to slaughter, except for scars [13]. Le- ating stochasticity, bio-economic models allow to ac- sions occurring earlier in life can resolve [13, 14] and count for the uncertain impact of pathogens and thus, pluck lesions could underestimate the true preva- parasites on animal performance and of market condi- lence of lung, heart and liver lesions during the produc- tions on production costs and farm profitability. This tion cycle. However, it is possible that even when lesions study aimed to 1) identify the main pluck lesions associ- resolved, performance is reduced in affected pigs. For in- ated with changes in ADG during the wean-to-finish stance, pluck lesions are associated with higher mortality period, 2) estimate the cut-off values for the prevalence rates [2] and reduced average daily gain (ADG, [7, 15, of these lesions that maximize effects on ADG in the 16]) during the grower-finisher period and longer time studied population, and 3) simulate the economic per- to reach target slaughter weight [13] resulting in reduced formance of a farrow-to-finish farm by parameterising a feed conversion efficiency [7, 17]. Additionally, pluck le- previously constructed stochastic bio-economic model sions are associated with carcass yield losses due to con- with production effects associated with the cut-off values demnations or trimmings [18]. for the prevalence of the pluck lesions identified in ob- While respiratory disorders are known to cause sub- jective one. stantial economic losses to pig farms (e.g. [19, 20]), there is no information regarding hierarchy of the effects of Methods the different pluck lesions and cut-off values for their Farm selection, prevalence of pluck lesions and prevalence that could be used for decision making to performance indicators minimize adverse effects on animal performance, and This study is part of a larger project investigating re- farm profitability. Currently, farmers and veterinarians spiratory disease on Irish pig farms, associated risk fac- do not use objective quantitative methods to decide at tors, and the relationship with performance, welfare and which prevalence a disease becomes a problem. The antimicrobial use. The farms participating in this study hierarchy and cut-off values for the prevalence of differ- were part of the Teagasc e-Profit Monitor. Teagasc is ent lesions that producers and veterinarians see as crit- the Irish Authority for Agriculture and Food Develop- ical for intervention could be different than a cut-off ment, a semi-state organisation responsible for providing value obtained using evidence-based methods in a popu- integrated research, advisory and education/training ser- lation of farms. An objectively estimated hierarchy and vices for the agriculture and food industry in Ireland. cut-off values for the prevalence of pluck lesions for in- The Teagasc e-Profit Monitor is an online financial ana- terventions could aid producers and veterinary practi- lysis system for assessing farm profitability which con- tioners to maximise benefits on their farms. tains biological and economic records. Farms (n = 107) Although pluck lesions are associated with decreased providing data to the Teagasc e-Profit Monitor in the performance [7, 13, 15–17], only few studies have con- year 2017 were contacted directly by telephone or via ducted economic analyses. Jäger et al. [21] quantified the their Teagasc pig advisor specialist and invited to Calderón Díaz et al. Porcine Health Management (2020) 6:40 Page 3 of 15 participate in the study. A total of 56 farmers voluntarily included farrowing rate, litters per sow per year, average agreed to participate (52.3% participation rate). All 56 number of piglets born alive per litter, culling rate, mor- farms participating in the study were farrow-to-finish tality rates for different production stages, ADG, live farms with weekly farrowing batches which is the pre- weight at sale and dressing percentage. dominant pig production system in Ireland [23]. Upon written consent from the farmer, a visit to the Associations between pluck lesions and production abattoir was schedule to inspect pluck lesions in a batch parameters of pigs per farm. The absence of respiratory disease out- Regression tree [31] analysis was used to define the main breaks (i.e. an increase in mortality or clinical signs that lesions associated with ADG during the grow-finisher resulted in a major change in medication or vaccination) period and the cut-off value for their prevalence result- was confirmed during the call to schedule the abattoir ing in partitions showing maximum differences in ADG. visit. In total, 9254 pigs (162 ± 51.9 pigs; range 55 to Average daily gain was used as the main output variable 308) were inspected at slaughter for lesions in the lungs, based on previous reports in the scientific literature re- heart and liver by a single trained observer from garding decreased performance associated with presence November 2017 to April 2018. During the data collec- of pluck lesions (e.g. [4, 7]) and on previous results from tion period, the observer was blind to herd positive or our research group. Regression tree analysis is a non- negative status with regards to four of the most common parametric statistical technique based on recursive parti- respiratory pathogens (i.e. Actinobacillus pleuropeneu- tioning analysis [32] that selects those variables and their moniae, Mycoplasma hyopneumoniae, porcine repro- interactions that are most important in predicting the ductive and respiratory syndrome virus and swine outcome variable by calculation of their relative import- influenza virus) involved in the porcine respiratory dis- ance [33]. In other words, a regression tree is a hierarch- ease complex (PRDC, [24, 25]) and, that are associated ically organized structure [34]. There are several with the development of pluck lesions [26]. advantages to using regression tree analysis including Lung were removed from the carcass by abattoir that it does not require the data to be linear, and thus, personnel and assessed in the evisceration line for the accommodating multi-collinearity among predictors presence of enzootic pneumonia-like lesions, dorso- [33]; predictors can be continuous or categorical and it caudal pleurisy, lung scars (i.e. healing indicative of accounts for multiple interactions among predictors pneumonic lesions which developed earlier in the pig’s [35]. Additionally, the results are presented in a way that life) and abscesses. Pericarditis and liver milk spots (i.e. is easy to interpret by people not (very) familiar with presence of white spots in the liver suggestive of trans- statistical analyses. hepatic migration of the larvae of Ascaris suum) were Contrary to “classical” regression methods, regression also recorded. All pluck lesions were scored using the tree analysis does not develop a prediction equation but Ceva Lung Program app (CEVA Santé Animale, rather data are partitioned into subsets with homoge- Libourne, France) to facilitate data recording. Enzootic neous values of the dependent variable [36, 37]. In gen- pneumonia-like lesions were assessed according to the eral terms, the recursive partitioning process can be method described by Madec and Derrien [27] with the generalised to overall surface affected averaged accounting for lobe weights [28]. Pleurisy was scored in the dorso-caudal X n   ^f ðX Þ ¼ cm If X 1 ; X 2; ∈Rmg lobes using a modified version of the Slaughterhouse m¼1 Pleuritis Evaluation System (SPES, [29]) in a 4-point scale where 0 = no pleurisy; 2 = focal lesions in one lobe; indicating there is a continuous response variable Y and 3 = bilateral adhesions or monolateral lesions affecting inputs X1 and X2. cm are constants; I is an indicator more than 1/3 of the diaphragmatic lobe and 4 = exten- function returning 1 if its argument is true and 0 other- sive lesions affecting more than 1/3 of both diaphrag- wise and the recursive portioning resulting in Rm num- matic lobes. Lungs that remained attached to the chest ber of subsets. Each subset is created based on the wall and were not removed from the carcass were not regression tree identifying the predictor and split value scored. Pleurisy scores ≥2 were used to calculate the that partitions the data into two regions where the over- prevalence of pleurisy for each farm. Lung scars and ab- all sum of squares errors are minimised [37]. scesses, pericarditis and liver milk spots were recorded The recursive portioning process is done in a top- as present or absent. down fashion where the portioned done earlier in the For each participating farm, annual mean performance tree will not change based on later partitions [38]. A re- indicators required to parameterise the bio-economic gression tree starts with a root node containing all the model [30] were retrieved from the Teagasc e-Profit subjects; this is called a parent node. Every value of each monitor for the year 2017. Performance indicators predictor variable is considered as a potential split. The Calderón Díaz et al. Porcine Health Management (2020) 6:40 Page 4 of 15 optimal split is selected as the (cut-off) value of the pre- TPPM is a stochastic budgetary bio-economic simula- dictor variable that forms binary groups that are most tion model developed in Microsoft Excel. The TPPM is different with respect to the dependent variable [36]. representative of the predominant intensive farrow-to- This results in the parent node being split into two finish pig producing farms in Ireland with weekly child nodes (or branches) which in turn can be split farrowing batches, and several animal categories with into additional nodes [35]. The procedure continues different infrastructure and feeding practices for each through each node of the tree until a stopping rule is production stage [30]. The TPPM integrates biological, reached. At the point that no further split is made, a physical and technical parameters and economic ana- terminal node (or leaf) is created. The average value lysis. It allows the user to investigate the impact of of the dependent variable is estimated among the changes in pig production systems on farm performance subjects within each node. and profitability. First, the TPPM simulates weekly num- Regression tree analysis was done using the rpart bers of maiden gilts (24 to 32 weeks of age), gestating package [39] in R v3.4.2 [40]. It included ADG during sows (≥ 32 weeks of age), lactating sows (≥48 weeks of the grower-finisher period as outcome variable and the age), weaner pigs in stage 1 (c. 7 kg of body weight on prevalence of the six recorded pluck lesions as explana- transfer to this stage), weaner pigs in stage 2 (c. 19 kg of tory variables. The stopping criterion was a minimum of body weight on transfer to this stage), and finisher pigs 10 farms being required to create a branch and/or a leaf. (c. 38 kg of body weight on transfer to this stage) based This minimum of 10 farms was estimated based on a on biological inputs such as herd size, conception and minimum difference in ADG of at least 40 g between re- farrowing rate, number of litters per sow per year, num- gression tree partitions used to calculate sample size for ber of piglets born alive per litter and mortality rate for a t-test for independent samples in PROC POWER of each production stage. SAS v9.4 (SAS Inst. Inc., Cary, NC). Then, to simulate animal growth during the wean-to- After building the regression tree, it was pruned to finish period, the TPPM includes the Gompertz growth find an optimal size and avoid over-fitting of the data by function [41] using the formula BW = W0 exp[μ0(1 − using a cost complexity parameter (α) that penalises the e−Dt)/D]; where BW = body weight; W0 = the value of minimization of the sum of squares errors for the num- the growth function at age 0; μ0 = logarithm of the rela- ber of terminal nodes on the regression tree (T) tive growth rate at age 0 and D = slope of the logarithm of the relative growth rate. Dietary nutritional require- MinimisefSSE þ αjT jg ments (i.e. energy, amino acids and minerals) vary for each production stage and are estimated following the for a given value of α, the smallest pruned tree that has recommendations from the National Research Council the lower penalised error is found. If the cost of adding (NRC) Nutrient Requirements of Swine [42]. For each another variable to the regression tree from the current production stage, wheat-barley-soya-based diets were node is above the value of cost complexity parameter, formulated within the TPPM to meet or exceed NRC then tree building does not continue. The printcp() func- [42] requirements. All diets are representative of com- tion in rpart [39] was used to identify the cost complex- mon feeding practice in Irish pig farms. Daily energy ity parameter having the least cross-validated error and demand and feed intake are estimated following the use it to prune the regression tree. ANOVA test was NRC [42] equations for estimating nutrient requirements performed between each binary partition to determine for weaner-finisher pigs according to their estimated BW whether the groups created were statistically different obtained from the Gompertz growth curve (for more from each other using an alpha of 0.05. Following this, information please refer to Calderón Díaz et al. [30]). group averages were calculated for performance indica- Biological performance is then combined with man- tors required to parameterise a bio-economic model [30] agement practices including reproductive management, that was then used for the economic analysis of the labour, herd healthcare plan and infrastructure available different scenarios describe below. at the farm. Costs associated with management practices are also included in the TPPM. Other costs such as the Bio-economic simulation annual subscription to the Environmental Protection Economic analysis Agency, transport costs per pig to the abattoir, long term The Teagasc Pig Production model (TPPM; Calderón bank loans payments, and monthly feedstuff prices are Díaz et al. [30]) was used to simulate annual economic also included. In the TPPM, the only source of income performance of a farrow-to-finish pig farm using infor- is the sale of culled sows and finisher pigs. Cold carcass mation on the association between key performance in- weight is calculated by multiplying body weight at dicators and the cut-off values for the main pluck sale by the dressing percentage. Average monthly lesions identified during the regression tree analysis. The price per kg of meat is used to calculate income per Calderón Díaz et al. Porcine Health Management (2020) 6:40 Page 5 of 15 pig. For more information regarding the assumptions TPPM based on the mortality rates for each production of the bio-economic model please refer to Calderón stage and varied for each scenario. For all the scenarios, Díaz et al. [30]. all pigs were vaccinated for porcine circovirus type 2 Physical outputs from the TPPM include annual num- and Mycoplasma hyopneumoniae at weaning (i.e. 28 days ber of pig produced, annual number of kg of meat sold of age) with a single dose. Also, maiden gilts and and annual feed usage during the different production lactating sows were vaccinated for Erysipelothrix rhusio- stages. Financial outputs from the TPPM include vari- pathiae and porcine parvovirus as per normal practice able and fixed costs, gross income, net profit, an annual in Irish pig farms. Prices for the vaccines were obtained cash flow budget, annual profit and loss account, and from a major veterinary distributor in Ireland. In Ireland, annual balance sheet. Variable and fixed costs and sales only 1.8% of farms vaccinate for Actinobacillus pleup- are simulated based on current market costs and prices. neumoniae and thus, we did not include this vaccination The estimated annual gross income and net profit is pre- in the simulation. Four veterinarian visits per year at a sented on a total farm basis, as well as per pig produced cost of €300 each were considered for the economic ana- and per kg of meat sold. lysis as per usual practice. We simulated a 728 farrow-to-finish farm with weekly In all scenarios, pigs were weaned at 28 days of age farrowing batches for an entire year. This herd size cor- and 7 kg of body weight. Animal growth during the responds to the mean herd size in Ireland for the year wean-to-finish period was simulated to represent the 2017 [43]. Three different scenarios were simulated different ADG rates obtained from the regression tree based on results from the regression tree analysis. Infor- analysis. However, all pigs were slaughtered at 110.8 kg mation on mean farrowing rate, litters per sow per year, of body weight which was the average body weight at number of piglets born alive per litter, sow culling and sale in Irish farms in the year 2017 [43]. Carcass con- sow mortality rate, mortality rates during the weaner demnations associated with pluck lesions were not con- stage one, weaner stage two and finisher stage and dress- sidered in the simulation as no data were available; ing percentage were calculated for each scenario based however, a 1.5% condemnation rate was used for all sce- on farm record from the Teagasc e-profit monitor for narios. Monthly feedstuff and pork prices (per kg) for the year 2017 (Table 1). In the simulated farm, all animal the year 2017 were obtained from the Teagasc e-Profit categories were included (i.e., piglets, weaner and fin- monitor and did not vary between scenarios. In all sce- isher pigs, maiden gilts, pregnant and lactating sows and narios, diet formulation and feed intake was calculated boars for heat detection). The number of gilts, gestating as a function of body weight as previously explained. In- and lactating sows as well as number of piglets, weaners frastructure (e.g. number of pig spaces available in each and finisher pigs were calculated each week within the production stage, depreciation), management practices, Table 1 Biological parametersa obtained from the Teagasc e-Profit monitor used to parameterised the Teagasc Pig Production Model [30] to simulateb effects associated with different prevalence of pleurisy and lung scars on slaughter pigs on farm performance and profitability Parameter LPLSCc LPHSCd HPe Min Mean Max Min Mean Max Min Mean Max Farrowing rate, % 81 89.3 95.7 73 88.2 96.1 81.8 87.9 91.2 Litters per sow per year 2.09 2.30 2.43 2.11 2.28 2.50 2.24 2.33 2.46 Number born alive piglets per litter 12.4 13.7 15.2 11.6 13.6 15.3 12.8 13.4 14.0 Sow culling rate, % 34.3 46.9 58.0 37.8 51.1 63.9 39.0 47.0 54.9 Sow mortality, % 2.3 4.2 9.3 1.8 5.4 9.2 3.2 4.7 9.7 Piglet mortality, % 6.8 10.7 14.3 5.7 10.8 15.9 5.6 9.9 14.0 Weaner mortality, % 0.5 2.1 6.8 0.9 2.9 8.9 1.5 3.7 7.0 Finisher mortality, % 0.9 1.7 3.3 1.0 2.1 4.1 1.3 2.3 3.0 Dressing, % 75.1 76.2 78.1 74.8 76.2 77.4 76.0 76.8 77.8 a Mean values were used for each scenario during the economic analysis. The range of values was used to fit probability distributions for the stochastic simulation using Monte Carlo sampling in the Microsoft Excel add-in @Risk [44] b A 728 sow farrow-to-finish farm with weekly farrowing batches was simulated to represent three different scenarios c Scenario 1: a farrow-to-finish farm with prevalence of pleurisy < 25% and prevalence of lung scars < 8% with a wean-to-finish average daily gain (ADG) of 760 g and reaching target slaughter weight at 24 weeks of age d Scenario 2: a farrow-to-finish farm with prevalence of pleurisy < 25% and prevalence of lung scars ≥8% (LPHSC) with an ADG of 725 g and reaching target slaughter weight at 25 weeks of age e Scenario 3: a farrow-to-finish farm with prevalence of pleurisy ≥25% (HP) with and ADG of 671 g and reaching target slaughter weight at 26 weeks of age Calderón Díaz et al. Porcine Health Management (2020) 6:40 Page 6 of 15 labour requirements, capital investment did not differ feed prices and pork market conditions are among the between scenarios. A flow diagram describing data most important variables affecting profitability in pig sources and the bio-economic simulation process farms. To account for uncertainty and variation in bio- followed in this study is presented in Fig. 1. logical inputs, feed costs and carcass prices, they were included as stochastic input variables during the risk Risk analysis and risk assessment analysis. Estimates for feedstuff and pork prices (per kg) Risk analysis (i.e. the identification of possible outcomes) were generated based on data recorded on the Teagasc and risk assessment (i.e. estimation of probabilities and pig e-Profit monitor between the years 2013 to 2017. By the economic impacts that result from the correspond- including historical data, we provide the TPPM with a ing outcomes [45]) were implemented through Monte range of values that better reflect market conditions as Carlo sampling using the Microsoft Excel add-in @Risk feedstuff and pork prices can experience drastic changes [44]. Risk analysis and risk assessment provide informa- in relatively short periods of time. Estimates for bio- tion that can be used as the basis for evaluating different logical parameters were generated for each variable for scenarios allowing for better decision making under each scenario based on the value range obtained from uncertainty. Reproductive performance, mortality rates, the Teagasc e-profit monitor for the year 2017 (Table 1). Fig. 1 Flow diagram describing data sources and bio-economic modelling process followed to simulate effects associated with different prevalence of pleurisy and lung scars on slaughter pigs on farm performance and profitability. Three scenarios were simulated for a 728 sow farrow-to-finish farm Calderón Díaz et al. Porcine Health Management (2020) 6:40 Page 7 of 15 For each stochastic variable, the distribution fitting func- Results tion from @Risk was used to determine their appropriate Regression tree analysis and selection of simulated distribution for the Monte Carlo simulation. Spearman’s scenarios rank correlations were estimated in PROC CORR of The prevalence of each studied pluck lesion in 56 farms SAS v9.4 (SAS Inst. Inc., Cary, NC) to account for pos- included in the study is shown in Fig. 2 while results for sible co-variation between stochastic input variables, and the regression tree analysis are shown in Fig. 3. The they were included during the Monte Carlo simulation. main pluck lesion associated with changes in ADG was Gross margin and annual net profit were set as the sto- dorso-caudal pleurisy followed by lung scars. The associ- chastic output variables. A total of 10,000 iterations were ated cut-off values for these lesions were 25% prevalence run during the Monte Carlo simulation. At each iter- of dorso-caudal pleurisy and 8% prevalence of lung scars. ation, all stochastic input variables varied simultaneously Thus, three different scenarios representing different ex- by randomly sampling a new set of values for each vari- tents of lung lesions on the farms were defined based on able from their corresponding distributions. Addition- these cut-off values: ally, gross margins and net profit were calculated for each iteration. Finally, annual mean gross margin and  Scenario 1: A farrow-to-finish farm with prevalence net profit and the variation of results around the mean of pleurisy < 25% and prevalence of lung scars < 8% were reported. In addition, stochastic dominance ana- (LPLSC). Assuming a weaning weight of 7 kg at 4 lysis, which compared distributions of income between weeks of age and a slaughter weight of 110.8 kg, on scenarios, was carried out by performing pairwise com- average, pigs will reach target slaughter weight at 24 parisons of income distributions for different scenarios weeks of age. This was the average age at slaughter being considered by inspecting their cumulative distribu- of Irish pigs as per data from the e-Profit monitor. tion function (CDF) curve. The CDF describes the prob- This scenario was parameterised with biological in- ability that a variable X is less than or equal to x: puts originating from 17 farms with mean preva- lence of pleurisy 3.9% ± 4.94 and mean prevalence of F ðxÞ ¼ P ðX ≤ xÞ for all x lung scars 2.8% ± 3.32.  Scenario 2: A farrow-to-finish farm with prevalence Stochastic dominance is a partial order of random var- of pleurisy < 25% and prevalence of lung scars ≥8% iables. Scenarios with a CDF further to the right are pre- (LPHSC). Assuming a weaning weight of 7 kg at 4 ferred and thus, the income distribution that exceeds the weeks of age as and a slaughter weight of 110.8 kg, other, at any level, is stochastically dominant indicating on average, pigs will reach target slaughter weight at lower economic risk and allowing the identification of 25 weeks of age. This scenario was parameterised the preferred scenario [46]. When two alternatives A with biological inputs originating from 29 farms with and B, each with a probability distribution of outcomes mean prevalence of pleurisy 7.5% ± 6.17 and mean x [defined by the cumulative probability of FA(x) and prevalence of lung scars 18.7% ± 8.34. FB(x)] are compared, A first-order stochastically domi-  Scenario 3: A farrow-to-finish farm with prevalence nates B if FA(x) ≤ FB(x), for all x and there is a strong in- of pleurisy ≥25% (HP). Assuming a weaning weight equality in at least one point of the distribution. of 7 kg at 4 weeks of age and a slaughter weight of Graphically, the CDF of A must always lie below and to 110.8 kg, on average, pigs will reach target slaughter the right of the cumulative probability of B. Second- weight at 26 weeks of age. This scenario was para- order stochastic dominance is examined by comparing meterised with biological inputs originating from 10 the integrals of CDFs (i.e. area under CDF): A second- farms with mean prevalence of pleurisy 38.8% ± 7.55 order stochastically dominates B if and mean prevalence of lung scars 19.3% ± 12.1. Z x Z x Bio-economic simulation F A ðxÞdx ≤ F B ðxÞdx −∞ −∞ Economic analysis The farm simulated under the HP scenario had a 10.2 for all x with a strict inequality for some range of the and 6.7% increase in weaner and finisher feed usage, re- distribution. This implies that in a subset of distribution, spectively; produced 3.5% fewer pigs and sold 2.5% less the dominating alternative A may not lead to a better meat (Table 2) when compared with the LPLSC sce- outcome than the dominated alternative B. Graphically, nario farm. Higher feed usage translated into higher feed the CDF of the dominating scenario A is still further to costs which in turn increased variable costs (Table 3) in the right for the most part of the CDF and more predict- the HP farm. Also, variable costs increased in the HP able than the dominated alternative B, but not for the farm due to greater numbers of dead animals for dis- entire distribution. posal. Sales (i.e. finisher pigs and culled sows) were 2.4% Calderón Díaz et al. Porcine Health Management (2020) 6:40 Page 8 of 15 Fig. 2 Box and whisker plots for the prevalence (%) of pluck lesions (pleurisy, enzootic pneumonia-like lesions, lung scars and abscesses, pericarditis and liver milk spots) in finisher pigs from 56 farrow-to-finish farms. Each dot represents a farm lower in the HP farm than in the LPLSC farm. Addition- Increased variable costs translated into lower gross in- ally, gross margin and net profit decreased in the HP come (5%) and lower net profit (18.1%) in the HP farm by 14.3 and 41.4%, respectively compared with the farm compared with the LPHSC farm. LPLSC farm. When compared with the LPHSC farm, Finally, when compared with the LPLSC farm, the the HP farm had a 3.7% decrease in weaner feed usage, a LPHSC farm used 14.5% more weaner feed, 0.5% fewer 7.2% increase in finisher feed usage and sold 1.1% fewer finisher feed and sold 2.2% fewer pigs and 2.2% less meat pigs and 0.3% less meat (Table 2). Higher variable costs (Table 2). Higher variable costs were observed in the were observed in the HP farm due to higher finisher LPHSC farm due to higher weaner feed costs, higher feed costs and a greater number of dead animals for cost to produce replacement gilts and a greater number disposal. Additionally, sales were 0.5% lower in the of dead animals for disposal. Sales were 2% lower in the HP farm than in the LPHSC farm (Table 3). LPHSC farm than in the LPLSC farm. Gross margin and Fig. 3 Regression trees for average daily gain (ADG) where the prevalence of pleurisy, pneumonia, scars, pericarditis, abscesses, and liver milk spots were included as predictor variables. The cut-off value of the prevalence of pleurisy and lung scars showed the best division of farms with ADG (g). The percentage of farms in each node/leaf is given for the different groups Calderón Díaz et al. Porcine Health Management (2020) 6:40 Page 9 of 15 Table 2 Annual physical outputs obtained from the Teagasc Pig performance of different types of farms defined by Production Model [30] for the simulationa of effects associated their types and levels of pluck lesions. The selected with different prevalence of pleurisy and lung scars on slaughter types and levels of lesions may not be the same for pigs on farm performance and profitability different cohorts of farms, but the methodology and Output LPLSCb LPHSCc HPd the resulting economic consequences would be still of Feed usage, ton interest. Gestation feed 540.5 540.5 540.5 Pluck lesions are common findings in slaughtered pigs Lactation feed 367.8 367.8 367.8 [9, 10, 47] and multiple lesions are often seen simultan- eously [4, 11, 15]. Studying the association of different Creep feed 77.3 76.7 76.5 pluck lesions with key farm performance indicators is of Link feed 142.3 141.2 140.9 interest to infer economic costs. Using linear models to Weaner feed 1137.1 1301.7 1253.1 study such associations is one of the options and prob- Finisher feed 3164.9 3150.1 3377.6 ably the most classical approach. However, in practice, it Sales is often needed to define cut-off points to implement a No. finisher pigs sold 19,188 18,772 18,564 certain intervention (e.g. medication, vaccination, de- population). These decisions can be done based on the Meat sold, ton 1595.1 1560.6 1555.4 a veterinarian experience but there are several statistical A 728 sow farrow-to-finish farm with weekly farrowing batches was simulated to represent three different scenarios tools that can help us to objectively estimate cut-off b Scenario 1: a farrow-to-finish farm with prevalence of pleurisy < 25% and values. In the present study we propose the use of re- prevalence of lung scars < 8% with a wean-to-finish average daily gain (ADG) of 760 g and reaching target slaughter weight at 24 weeks of age gression trees because they provide results in the format c Scenario 2: a farrow-to-finish farm with prevalence of pleurisy < 25% and of a decision tree that are easily understood. With this prevalence of lung scars ≥8% (LPHSC) with an ADG of 725 g and reaching approach, it is possible to obtain not only a target slaughter weight at 25 weeks of age d Scenario 3: a farrow-to-finish farm with prevalence of pleurisy ≥25% (HP) with hierarchization of the different factors affecting the out- and ADG of 671 g and reaching target slaughter weight at 26 weeks of age put but also typologies of farms (i.e. scenarios) present within a cohort. net profit were 9.8 and 28.5% lower in the LPHSC farm From the six recorded pluck lesions, pleurisy and lung compared with the LPLSC farm. scars were the main lesions associated with changes in ADG. We selected ADG as outcome because previous Risk analysis and risk assessment analyses of our database. We observed that ADG was For the risk analysis and assessment 10,000 iterations were negatively associated with respiratory disease by a simulated. Results for the risk analysis are presented in greater extent than other key performance indicators Table 4 while results for the risk assessment are presented such as feed conversion ratio (R2 = 40% vs. R2 = 14%, re- in Fig. 4. First-order stochastic dominance was observed spectively for multivariable linear regression models with for LPLSC farms over LPHSC and HP farms. Second- ADG or feed conversion ratio as predicted variables and order stochastic dominance was observed for LPHSC over pluck lesions as predictor variables, unpublished data). HP farms. Therefore, HP farms were more likely to lose It is important to note that a causative relationship be- money than LPHSC farms but they also had less volatile tween these lesions and ADG is not suggested at any returns which implied that at some lower levels of profit point although there is evidence of direct effects of LPHSC farms were expected to obtain slightly higher pleurisy on performance [4, 7]. profits than HP farms (Fig. 4). Indeed, mean annual gross The levels of lesions were used in this study to classify income decreased 13.5% and 4.2% HP farms compared farms and study their bio-economic performance beyond with LPLSC and a LPHSC farms, respectively. A difference the direct effect in the grower-finisher stage. Both, pleur- in gross margin of 9.6% was observed between LPLSC and isy and scars are part of the PRDC. Pleurisy lesions in LPHSC farms. Similarly, mean annual net profit decreased the dorso-caudal lobes are suggestive of Actinobacillus by 42.1% and 17.1% in HP farms compared with LPLSC pleuropneumoniae [26, 48], and scars are related to and a LPHSC farms, respectively and mean annual net healed enzootic pneumonia caused by Mycoplasma profit was 30.2% lower in the LHPSC farms compared hyopneumonia. However, other pathogens are often in- with the LPLSC farms. volved in the development of such lesions [49, 50]. In fact, the PRDC includes pathogens that affect the herd Discussion in all the stages, like PRRS virus, and is also affected by This study describes the main associations between management practices which are interrelated between pluck lesions recorded at slaughter and ADG in the stages at a farm level. grower-finisher period using regression trees. Then, The estimated cut-off value for pleurisy in this study is these associations were used to simulate bio-economic lower than the alarm (55%; defined as the incidence Calderón Díaz et al. Porcine Health Management (2020) 6:40 Page 10 of 15 Table 3 Comparison of trade profit and loss accounts obtained from the Teagasc Pig Production Model [30] for the simulationa of effects associated with different prevalence of pleurisy and lung scars on slaughter pigs on farm performance and profitability Item €/year €/pig produced €/kg meat b c d LPLSC LPHSC HP LPLSC LPHSC HP LPLSC LPHSC HP Sales Finisher pigs 2,582,595 2,526,610 2,518,277 134.6 134.6 135.7 1.62 1.62 1.62 Culled sows 40,972 44,641 41,059 2.1 2.4 2.2 0.03 0.03 0.03 Total Sales 2,623,567 2,571,250 2,559,336 136.7 137.0 137.9 1.64 1.65 1.65 Variable costs Gestation feed 126,460 126,403 126,348 6.6 6.7 6.8 0.08 0.08 0.08 Lactation feed 94,750 94,711 94,674 4.9 5.0 5.1 0.06 0.06 0.06 Creep feed 71,235 70,689 70,506 3.7 3.8 3.8 0.04 0.05 0.05 Link feed 84,540 83,892 83,676 4.4 4.5 4.5 0.05 0.05 0.05 Weaner feed 296,067 338,451 326,804 15.4 18.0 17.6 0.19 0.22 0.21 Finisher feed 753,112 749,258 803,019 39.2 39.9 43.3 0.47 0.48 0.52 Replacement gilts 60,232 68,829 60,219 3.1 3.7 3.2 0.04 0.04 0.04 Dead animal Disposal 9170 11,274 12,213 0.5 0.6 0.7 0.01 0.01 0.01 Health care 41,961 41,689 41,547 2.2 2.2 2.2 0.03 0.03 0.03 Reproduction 37,309 37,309 37,308 1.9 2.0 2.0 0.02 0.02 0.02 Manure handling 16,093 15,866 15,706 0.8 0.8 0.8 0.01 0.01 0.01 Transport 18,183 17,845 17,603 0.9 1.0 0.9 0.01 0.01 0.01 Total variable costs 1,609,112 1,656,215 1,689,624 83.9 88.2 91.0 1.01 1.06 1.09 Gross margin 1,014,455 915,036 869,712 52.9 48.7 46.8 0.64 0.59 0.56 Fixed costs Admin and accounting 2500 2500 2500 0.1 0.1 0.1 0.00 0.00 0.00 Electricity, heating and light 81,614 81,614 81,614 4.3 4.3 4.4 0.05 0.05 0.05 Insurance 20,533 20,533 20,533 1.1 1.1 1.1 0.01 0.01 0.01 Repairs 20,533 20,533 20,533 1.1 1.1 1.1 0.01 0.01 0.01 Annual subscription to EPA 10,000 10,000 10,000 0.5 0.5 0.5 0.01 0.01 0.01 Labour 279,136 279,136 279,136 14.5 14.9 15.0 0.17 0.18 0.18 Loan repayments (interest) 75,780 75,780 5780 3.9 4.0 4.1 0.05 0.05 0.05 Depreciation 175,021 175,021 175,021 .1 9.3 9.4 0.11 0.11 0.11 Total fixed costs 665,117 665,117 665,117 34.7 35.4 35.8 0.42 0.43 0.43 Total costs 2,274,229 2,321,332 2,354,741 118.5 123.7 126.8 1.43 1.49 1.51 Net Profit 349,338 249,918 204,595 18.2 13.3 11.0 .22 0.16 0.13 a A 728 sow farrow-to-finish farm with weekly farrowing batches was simulated to represent three different scenarios b Scenario 1: a farrow-to-finish farm with prevalence of pleurisy < 25% and prevalence of lung scars < 8% with a wean-to-finish average daily gain (ADG) of 760 g and reaching target slaughter weight at 24 weeks of age c Scenario 2: a farrow-to-finish farm with prevalence of pleurisy < 25% and prevalence of lung scars ≥8% (LPHSC) with an ADG of 725 g and reaching target slaughter weight at 25 weeks of age d Scenario 3: a farrow-to-finish farm with prevalence of pleurisy ≥25% (HP) with and ADG of 671 g and reaching target slaughter weight at 26 weeks of age above which a health plan is required) and warning (28% The different pleurisy cut-off values found between stud- defined as the half value of an alarm value) cut-off values ies may be attributable to the different methods, sample for pleurisy recommended by the Welfare Quality® population and year when data was obtained for their es- Protocol for pigs [51] although these values were sub- timation. We used an objective method (i.e. regression jectively defined by consultation with experts from the tree analysis) to estimate the cut-off values while Jäger animal science field. Also, our estimated cut-off value et al. [2] subjectively chose the cut-off values. Jäger et al. for pleurisy is higher than the one selected by Jäger et al. [2] monitored farms participating in the BPHS abattoir [2] of < 5% (low prevalence) and > 10% (high prevalence). pathology monitoring scheme database and that had at Calderón Díaz et al. Porcine Health Management (2020) 6:40 Page 11 of 15 Table 4 Mean ± standard deviation (SD) and the associated 90% confidence interval (CI, 5 to 95%), minimum and maximum annual gross margin and net profit in farmsa with different prevalence of pleurisy and lung scars on slaughter pigs. Results were obtained during the risk analysis by performing stochastic simulation analysis by Monte Carlo sampling with 10,000 iterations (‘farms’) using the Microsoft Excel add-in @Risk [44] Variable € per year € per pig € per kg LPLSCb LPHSCc HPd LPLSC LPHSC LP LPLSC LPHSC HP Gross margin, € Mean 978,020 883,579 846,131 51.1 47.2 45.2 0.61 0.57 0.54 SD 73,587 89,923 52,997 1.8 2.1 1.7 0.02 0.02 0.02 5% CI 861,044 740,681 761,039 48.1 43.8 42.4 0.58 0.53 0.51 95% CI 1,101,716 1,035,199 935,183 54.1 50.4 48.1 0.65 0.61 0.57 Minimum 730,977 571,026 657,281 43.7 38.0 38.2 0.53 0.46 0.46 Maximum 1,267,410 1,191,485 1,069,961 58.5 55.3 52.0 0.70 0.67 0.62 Net profit, € Mean 312,902 218,461 181,014 16.2 11.4 9.6 0.20 0.14 0.11 SD 73,587 89,923 52,997 3.1 4.1 2.6 0.04 0.05 0.03 90% CI 195,926-436,598 75,563-370,082 95,921-270,066 10.1–21.2 4.5–17.9 5.4–13.8 0.13–0.26 0.05–0.21 0.06–0.16 Minimum 65,860 −94,091 − 7836 4.0 −6.3 −0.5 0.05 −0.08 −0.01 Maximum 602,293 526,367 404,843 27.1 24.4 19.6 0.33 0.29 0.23 a 728 sow farrow-to-finish farms with weekly farrowing batches were simulated to represent three different scenarios. A total of 10,000 iterations (i.e. farms) were simulated for each scenario. At each iteration, all stochastic input variables (i.e. biological inputs, feedstuff and pork prices) varied simultaneously by randomly sampling a new set of values for each variable from their corresponding distributions. Additionally, gross margins and net profit were calculated for each iteration b Scenario 1: a farrow-to-finish farm with prevalence of pleurisy < 25% and prevalence of lung scars < 8% with a wean-to-finish average daily gain (ADG) of 760 g and reaching target slaughter weight at 24 weeks of age c Scenario 2: a farrow-to-finish farm with prevalence of pleurisy < 25% and prevalence of lung scars ≥8% (LPHSC) with an ADG of 725 g and reaching target slaughter weight at 25 weeks of age d Scenario 3: a farrow-to-finish farm with prevalence of pleurisy ≥25% (HP) with and ADG of 671 g and reaching target slaughter weight at 26 weeks of age least 50 pigs assessed in at least three occasions in 24 thus, feed cost varied in the different production stages. months. Then they subjectively decided on cut-off values Therefore, assumptions for economic analysis were dif- based on examining the distribution of the full dataset ferent between the study by Jäger et al. [21] and the and on a sample size calculation to determine the mini- current study. mum number of farms needed in each classification. Under the conditions of this study, a farm with a high Jäger et al. [2] reported increased weaner mortality in prevalence of pleurisy (≥ 25%) will experience greater farms with higher prevalence of pleurisy and Jäger et al. economic loses than a farm with pleurisy at a lower [21] estimated a pleurisy cost of £5 (approximately €5.8) prevalence even when combined with high prevalence of based on industry standard values. This is lower than other lung lesions such as lung scars (> 8%). Under the the cost of €7.9 per pig produced estimated in this study HP scenario, a farm would experience higher variable when compared to pigs produced under a scenario con- costs, mainly due the associated lower ADG during the sidering a farm with lower prevalence of pleurisy and grower-finisher period which would result in increased lung scars. Direct comparison with our results is how- weaner feed usage as pigs would require an extra week ever difficult as Jäger et al. [21] based their calculation in the second weaner stage to reach adequate weight for on an increased mortality rate of 1% being worth 50p transfer (i.e. 38 kg of body weight) to the finisher stage. per pig and considering a rise in mortality from 2 to 8% Also, a farm under the HP scenario would use more fin- being worth £3 per pig. They also estimated that for a ished feed as pigs would also require more time in the batch of 100 pigs with a prevalence of pleurisy of 20% finisher stage to reach target slaughter weight (i.e. 110.8 and a reduction in ADG of 50 g, the cost will be £21 kg of body weight); this would increase finisher feed across the batch (21p per pig). To this they summed a costs. Furthermore, the HP scenario was associated with cost of £1.40 per pig for a change of 0.1 in feed conver- higher mean mortality rates throughout the production sion efficiency plus costs associated with medication and cycle which would increase carcass disposal costs. Also, labour for treatment. Furthermore, feed costs differed higher mortality rates in the HP scenario would decrease between studies; Jäger et al. [21] assumed feed costs of the number of pigs and kg of meat produced thereby re- £140/t for all kinds of feed provided during the grower- ducing farm income. When examining results from the finisher period while we used current market values and risk analysis, farms simulated under the HP scenario Calderón Díaz et al. Porcine Health Management (2020) 6:40 Page 12 of 15 Fig. 4 Cumulative distribution function showing the cumulative probability (Y-axis) of reaching a given level of profit (X-axis) for three scenarios simulating effects associated with different prevalence of pleurisy and lung scars on slaughter pigs on farm performance and profitability. LPLSC farms [farms with prevalence of pleurisy < 25% and prevalence of lung scars < 8% with a wean-to-finish average daily gain (ADG) of 760 g and reaching target slaughter weight at 24 weeks of age] were first-order stochastically dominant compared with LPHSC (farm with prevalence of pleurisy < 25% and prevalence of lung scars ≥8% (LPHSC) with an ADG of 725 g and reaching target slaughter weight at 25 weeks of age) and HP farms (farm with prevalence of pleurisy ≥25% (HP) with and ADG of 671 g and reaching target slaughter weight at 26 weeks of age). LPHSC farms were second-order stochastically dominant compared to HP farms. For all scenarios, 728 sow farrow-to-finish farms with weekly farrowing batches were simulated. A total of 10,000 iterations (i.e. farms) were simulated for each scenario by performing stochastic simulation analysis by Monte Carlo sampling using the Microsoft Excel add-in @Risk [44] were less profitable, at any level, than farms simulated directly attribute the higher sow replacement rate in the under the LPLSC or LPHSC scenarios. For example, HP LPHSC scenario farm to the higher prevalence of lung farms only had 10% probability of reaching an annual scars and thus, we are unable to provide an explanation net profit of €250,000 while the probability of reaching for this result. Sows are culled/die due to several causes the same annual net profit was 81% for LPLSC farm and and thus, this result should be viewed with caution. 36% for LPHSC farms. These findings confirm that bet- Some methodological aspects of the current study are ter lung health is economically advantageous and suggest worth considering when interpreting the results. For in- that emphasis should be placed on improving animal stance, farms were invited to participate in the study health as means to improve performance. based on the criteria of providing performance records Our results also suggest that a farm where the preva- to the Teagasc e-Profit Monitor system. Previous work lence of pleurisy is low, a higher prevalence of lung scars reported differences between farms recordkeeping in the would result in greater economic risk. In this study, pigs Teagasc e-Profit Monitor and those that do not partici- produced under the LPHSC scenario would require one pate in the Teagasc e-Profit monitor with regards to wel- extra week to be transferred from the weaner to the fin- fare indicators [52] and thus, it is possible that farms isher stage which would increase weaner feed usage and agreeing to participate in this study had better health costs. Other variables cost would also increase in a and performance when compared to farms not partici- LPHSC farm. For example, under the LPHSC scenario, a pating in the Teagasc e-Profit Monitor. However, the farm would have higher replacement gilt costs because 107 herds providing data to the Teagasc e-Profit Moni- sow replacement rate (i.e. sow culling and sow mortality tor during the year 2017 represented 53% of the national rates) was higher than in LPLSC scenario farm and it sow herd and a high participation rate (52%) was ob- would require to produce more replacement gilts per tained among the farmers contacted. Moreover, key per- week to maintain breeding herd numbers. We cannot formance indicators were similar between participating Calderón Díaz et al. Porcine Health Management (2020) 6:40 Page 13 of 15 and non-participating farms recordkeeping in the acknowledge that a longer grower-finish period as per Teagasc e-Profit Monitor [43]. Therefore, the 56 partici- the HP and LPHSC scenarios would require extra space pating farms provide a representative sample of all herds to accommodate the pigs and that we did not account in the Teagasc e-Profit Monitor, and constitute a large for it in the bio-economic simulation. However, in real- sample of 29.2% of the Irish national sow herd. ity farmers are unable to provide additional housing to The economic calculations presented in this study were accommodate pigs for longer because of higher preva- based on mean values for a cohort of farms representing lence of pluck lesions. Farmers are likely to focus on different prevalence of pleurisy and lung scars associated management strategies that could improve ADG and/or with changes in ADG. Thus, we also acknowledge that the that would lower prevalence of the lesions in the long biological assumptions used to parameterise the three sce- term, thereby reducing the time pigs require to reach narios investigated in this study were likely affected by target slaughter weight. Further expansion of the TPPM other diseases and many other management and environ- to include these additional aspects would be a natural mental issues present on the farms. Integration of other continuation for this study once the necessary informa- diseases and confounding effects is very complex and in- tion becomes available. This will further aid in determin- troduces more uncertainty around the contribution of dif- ing the cost implications of management strategies ferent diseases/environmental issues on key biological currently used, as well as the feasibility of implementing parameters affected [53]. However using performance data new practices to improve animal health and from real farms takes into account the multifactorial na- performance. ture of different pathogens/environmental issues and their associated lesions. Changes in the prevalence of a specific Conclusion pathogen or changes in environmental conditions (e.g. Under the conditions of this study, higher prevalence of animal husbandry practices) would result in changes in pleurisy and lung scars were associated with decreased the prevalence of a secondary pathogen as they interact ADG during the grower-finisher period and with lower among them. Thus, it is questionable if the impact of a economic return in the simulated farms. The regression single pathogen or environmental condition can be tree analysis illustrates an easy to understand objective truly isolated from each other [53]. Moreover, moni- method that could be used by pig producers and veterin- toring several batches of pigs at regular time intervals ary practitioners to identify cut-off points for the preva- will provide a more reliable measure of the prevalence lence of various health problems to decide when to of different pluck lesions on a given farm compared implement disease prevention/eradication strategies on with using the information of a single batch as in this their farms to minimise adverse disease effects on animal study; however, it is not unusual to determine the performance. A lower prevalence of lung lesions was as- prevalence of pluck lesions based on a single visit to sociated with lower economic risk at any given level of the abattoir in countries without computerized sys- profit confirm that better lung health is economically ad- tems for lesions recording (e.g. [11, 26]). vantageous. Thus, our results also form a foundation for In this study we decided to use the prevalence of le- future work which should consider additional variables sions, instead of the prevalence of different severity of le- related to control of pluck lesions on pig farms and lon- sions (in the case of pleurisy and pneumonia), as we gitudinal approaches. Additionally, our results highlight were interested in identifying which lesions, and their the economic benefits and importance of preventing or possible combinations, were more associated with controlling respiratory disease. changes in ADG. By using the prevalence of different Acknowledgements scores, it would be likely that the regression tree results The authors would like to thank the farmers participating in the cross- presented cut-off values for different severities for the sectional study from where the data on prevalence of pluck lesions were col- same lesions e.g. pleurisy since it was the main lesion as- lected, and the abattoirs for allowing the scoring of pluck lesions in the pig batches followed to slaughter. sociated with changes in ADG. Associations between ADG and prevalence of different severity scores for e.g. Authors’ contributions pleurisy and their bio-economic simulation would in- JACD developed the bio-economic model used for analysis, carried-out the bio-economic analysis and initial drafting of the manuscript. MRC – scored deed be interesting for a follow up study. Also, while we pluck lesions at the abattoir; retrieved production parameters from the Tea- accounted for differences in ADG, additional costs from gasc e-profit monitor and constructed the database used to identify the dif- e.g., differences in condemnation rates associated with a ferent scenarios used in this study. LS and JN- contributed to the development of the bio-economic model used for analysis and to outlining different prevalence of pleurisy and lung scars lesions, or the research focus of the article. FCL and JG - contributed to outlining the different management strategies, labour requirements, research focus of the article. DCP- carried out the regression tree analysis. and healthcare treatments to reduce these lesions were EGM - contributed to the development of the bio-economic model used for analysis and to outlining the research focus and the idea of the article. All not available for inclusion in the TPPM for the eco- authors contributed to preparing the article. The author (s) read and ap- nomic analysis due to a lack of data. We also proved the final manuscript. Calderón Díaz et al. Porcine Health Management (2020) 6:40 Page 14 of 15 Funding slaughter-aged pigs. Vet J 2010;184:326–333. https://doi.org/10.1016/j.tvjl. Julia Adriana Calderón Díaz and data collection used for this study were 2009.03.029. funded by the Department of Agriculture, Food and the Marine under the 9. Meyns T, Van Steelant J, Rolly E, Dewulf J, Haesebrouck F, Maes D. A cross- Research Stimulus Funded project “Investigation of respiratory disease on sectional study of risk factors associated with pulmonary lesions in pigs at Irish pig farms, associated risk factors, and the relationship with performance, slaughter. Vet J 2011;187:388–392. https://doi.org/10.1016/j.tvjl.2009.12.027. welfare and antimicrobial use” (PSP; grant no. 14/S/832). Maria Rodrigues da 10. Martínez J, Peris B, Gómez EA, Corpa JM. The relationship between Costa was supported by the Teagasc Walsh Fellowship Scheme grant-in-aid infectious and non-infectious herd factors with pneumonia at slaughter and project “Evaluation of current feeding and management practices in Irish pig productive parameters in fattening pigs. Vet J. 2009;179:240–6. production and future strategies for improvement” ref. PDPG-0231-6666. 11. Fablet C, Marois C, Dorenlor V, Eono F, Eveno E, Jolly JP, et al. Bacterial Daniel Crespo-Piazuelo was supported by the Health Research Board (grant pathogens associated with lung lesions in slaughter pigs from 125 herds. no. JPI-AMR-2017-1-A). Res Vet Sci 2012;93:627–630. https://doi.org/10.1016/j.rvsc.2011.11.002. 12. Steenhard NR, Jungersen G, Kokotovic B, Beshah E, Dawson HD, Urban JF Jr, Availability of data and materials et al. Ascaris suum infection negatively affects the response to a The R code used for the regression tree analysis and datasets for the results mycoplasma hyopneumoniae vaccination and subsequent challenge presented in this study are available from the corresponding author upon infection in pigs. Vaccine. 2009;27:5161–9. reasonable request. 13. Maes D, Verdonck M, Deluyker H, de Kruif A. Enzootic pneumonia in pigs. Vet Q. 1996;18:104–9. 14. Roepstorff A, Eriksen L, Slotved H, Nansen P. Experimental Ascaris suum Ethics approval and consent to participate infection in the pig: worm population kinetics following single inoculations All the farmers participating in this study consented to the use of the data with three doses of infective eggs. Parasitology. 1997;115:443–52. collected at the abattoir, and to the retrieval of their production data from 15. Brewster VR, Maiti HC, Tucker AW, Nevel A. Associations between EP-like the Teagasc e-Profit monitor in accordance with Teagasc internal data pro- lesions and pleuritis and post trimming carcass weights of finishing pigs in tection regulations. England. Livest Sci 2017;201:1–4. https://doi.org/10.1016/j.livsci.2017.04.012. 16. Vlaminck J, Düsseldorf S, Heres L, Geldhof P. Ascaris suum , lung pathogens Consent for publication and technical performance parameters. Vet Parasitol 2015;210:151–158. Not applicable. https://doi.org/10.1016/j.vetpar.2015.04.012. 17. Cleveland-Nielsen A, Nielsen EO, Ersboll AK. Chronic pleuritis in Danish Competing interests slaughter pig herds. Prev Vet Med. 2002;55:121–35. The authors declare that they have no competing interests. 18. Teixeira DL, Harley S, Hanlon A, Connell NEO, Dwyer CM. Study on the association between tail lesion score, cold carcass weight, and viscera Author details condemnations in slaughter pigs. Front Vet Sci. 2016;3:24. 1 Pig Development Department, Teagasc Animal and Grassland Research and 19. Niemi JK, Jones P, Tranter R, Heinola K. Cost of production diseases to pig Innovation Centre, Moorepark, Fermoy, Co. Cork, Ireland. 2School of farms. 24th Int Pig Vet Soc Congr. Dublin, Ireland; 2016. p. 302. Veterinary Medicine, University College Dublin, Dublin, Ireland. 20. Stygar AH, Niemi JK, Oliviero C, Laurila T, Heinonen M. Economic value 3 Epidemiology research unit, Scotland’s rural College (SRUC), Inverness, of mitigating Actinobacillus pleuropneumoniae infections in pig Scotland IV2 5NA, UK. 4Livestock Production Systems, Teagasc Animal and fattening herds. Agric Syst 2016;144:113–121. https://doi.org/10.1016/j. Grassland Research and Innovation Centre, Moorepark, Fermoy, Co. Cork, agsy.2016.02.005. Ireland. 5Natural Resources Institute Finland (LUKE), Kampusranta 9, 60320 21. Jäger HJ, Mckinley TJ, Pearce GP, Tucker AW, Wood JLN. Pleurisy in Pigs : Seinäjoki, Finland. 6Departament de Ciència Animal I dels Aliments, Facultat Associated risk factors and impact on health , welfare. BPEX Rep. 2009:1–94.. de Veterinaria, Universitat Autònoma de Barcelona, Barcelona, Spain. 22. Kragt ME. Bioeconomic modelling: Integrating economic and environmental systems? 2012 Int Congr environ model Softw Manag Resour a ltd planet, Received: 4 August 2020 Accepted: 24 November 2020 Sixth Bienn Meet 2012;9. http://www.iemss.org/sites/iemss2012/ proceedings/D8_0464_Kragt.pdf. 23. Marquer P, Rabade T, Forti R. Pig farming in the European Union: References considerable variations from one member state to another. 2014. 1. VanAlstine W. Respiratory system. In: Zimmerman J, Karriker L, Ramirez A, 24. Hansen MS, Pors SE, Jensen HE, Bille-Hansen V, Bisgaard M, Flachs EM, et al. Schwartz KJ, Stevenson G, editors. Dis Swine. 10th ed. Ames, Iowa, USA: An investigation of the pathology and pathogens associated with porcine Blackwell Publishing Professional; 2010. p. 348–62. respiratory disease complex in Denmark. J Comp Pathol. 2010;143:120–31. 2. Jäger HC, McKinley TJ, Wood JLN, Pearce GP, Williamson S, Strugnell B, et al. 25. Bochev I. Porcine respiratory disease complex ( Prdc ): a review. I. etiology , Factors associated with pleurisy in pigs: a case-control analysis of slaughter epidemiology , clinical forms and pathoanatomical features. Bulg J Vet Med. pig data for England and Wales. PLoS One. 2012;7:1–9. 2007;2:131–46. 3. Scollo A, Gottardo F, Contiero B, Mazzoni C, Leneveu P, Edwards SA. 26. Jirawattanapong P, Stockhofe-Zurwieden N, van Leengoed L, Wisselink H, Benchmarking of pluck lesions at slaughter as a health monitoring tool for Raymakers R, Cruijsen T, et al. Pleuritis in slaughter pigs: relations between pigs slaughtered at 170 kg (heavy pigs). Prev Vet Med. 2017;144:20–28. lung lesions and bacteriology in 10 herds with high pleuritis. Res Vet Sci https://doi.org/10.1016/j.prevetmed.2017.05.007. 2010;88:11–15. https://doi.org/10.1016/j.rvsc.2009.06.007. 4. Pagot E, Pommier P, Keïta A. Relationship between growth during the 27. Madec F, Derrien H. Frequence, intensite et localosation des lesion fattening period and lung lesions at slaughter in swine. Rev Med Vet pulmonaires chez le porc charcutier. 13es Journées la Rech Porc. Paris, (Toulouse). 2007;158:253–9. France; 1981. p. 231–6. 5. Greve J h. Internal parasites: Helminths. In: Zimmerman JJ, Karriker LA, 28. Christensen G, Soarensen V, Mousing J. Diseases of the respiratory system. In: Ramirez A, Schwartz KJ, Stevenson GW, editors. Dis Swine 10th ed. 10th Straw B, D’Allaire S, Taylor DJ WM, editors. Dis Swine. 8th ed. 1999. p. 913–40. editi. Oxford, UK: Wiley-Blackwell; 2012. p. 908–920. 29. Ostanello F, Dottori M, Gusmara C, Leotti G, Sala V. Pneumonia disease 6. Ferraz MES, Almeida HMS, Storino GY, Sonálio K, Souza MR, Moura CAA, assessment using a slaughterhouse lung-scoring method. J Vet Med Ser A et al. Lung consolidation caused by mycoplasma hyopneumoniae has a Physiol Pathol Clin Med. 2007;54:70–5. negative effect on productive performance and economic revenue in 30. Calderón Díaz JA, Shalloo L, Niemi JK, Kyriazakis I, Mckeon M, Mccutcheon G, finishing pigs. Prev Vet Med 2020;182:105091. https://doi.org/10.1016/j. et al. Description, evaluation, and validation of the teagasc pig production prevetmed.2020.105091. model. J Anim Sci. 2019;97:2867–85. https://doi.org/10.1093/jas/skz163. 7. Straw BE, Shin SJ, Yeager AE. Effect of pneumonia on growth rate and feed 31. Breiman L, Friedman J, Olshen R, Stone C. Classification and regression trees. efficiency of minimal disease pigs exposed to Actinobacillus pleuropneumoniae Boca Raton, FL, USA: Chapman & Hall; 1984. and mycoplasma hyopneumoniae. Prev Vet Med. 1990;9:287–94. 32. Strobl C, Malley J, Tutz G. An introduction to recursive partitioning: rationale, 8. Fraile L, Alegre A, López-jiménez R, Nofrarías M, Segalés J. Risk factors application and characteristics of classification and regression trees, bagging associated with pleuritis and cranio-ventral pulmonary consolidation in and random forests. Psychol Methods. 2009;14:323–48. Calderón Díaz et al. Porcine Health Management (2020) 6:40 Page 15 of 15 33. Speybroeck N, Berkvens D, Mfoukou-Ntsakala A, Aerts M, Hens N, Van Huylenbroeck G, et al. Classification trees versus multinomial models in the analysis of urban farming systems in Central Africa. Agric Syst. 2004;80:133–49. 34. Morgan J. Classification and regression trees. Tech. Rep. Boston; 2014. 35. Lewis RJ, Ph D, Street WC. An Introduction to Classification and Regression Tree (CART ) Analysis. 2000 Annu Meet Soc Acad Emerg Med. 2000. p. 14p. Available from: http://citeseerx.ist.psu.edu/viewdoc/download?doi=10.1.1. 95.4103&rep=rep1&type=pdf. 36. Lemon SC, Roy J, Clark MA, Friedmann PD, Rakowski W. Classification and regression tree analysis in public health: methodological review and comparison with logistic regression. Ann Behav Med. 2003;26:172–81. 37. Krzywinski M, Altman N. Classification and regression trees. Nat Methods. 2017;14:757–8. 38. Yang L, Liu S, Tsoka S, Papageorgiou LG. A regression tree approach using mathematical programming. Expert Syst Appl 2017;78:347–357. https://doi. org/10.1016/j.eswa.2017.02.013. 39. Therneau T, Atkinson B. rpart: Recursive Partitioning and Regression Trees. 2018. https://cran.r-project.org/package=rpart%0A. 40. R Core Team. R: a language and environment for statistical computing. Vienna, Austria: R Foundation for Statistical Computing; 2017. https://www.r- project.org/. 41. Wellock I, Emmans G, Kyriazakis I. Describing and predicting potential growth in the pig. Anim Sci. 2004;78:379–88. 42. National Research Council. Nutrient requirements of swine. Washington, D. C: National Academies Press; 2012. 43. Teagasc. National Pig Herd Performance Report - 2017. 2018. 44. Palisade. @Risk Risk Analysis Add-in for Microsoft Excel Versions 6.1.2. Newfield, New York, USA: Palisade Corporation; 2013. 45. Strong RA, Steiger NM, Wilson JR. Introduction to financial risk assessment using Monte Carlo simulation. Austin: Winter Simulation Conference; 2009. p. 99-118. 46. Hardaker JB, Lien G, Anderson JR, Huirne RBM. Coping with risk in agriculture: applied decision analysis. 3rd. Editi. Oxfordshire, UK: CABI; 2015. 47. Maes D, Chiers K, Haesebrouck F, Laevens H, Verdonck M, De Kruif A. Herd factors associated with the seroprevalences of Actinobacillus pleuropneumoniae serovars 2,3 and 9 in slaughter pigs from farrow-to- finish pig herds. Vet Res. 2001;32:409–19. 48. Merialdi G, Dottori M, Bonilauri P, Luppi A, Gozio S, Pozzi P, et al. Survey of pleuritis and pulmonary lesions in pigs at abattoir with a focus on the extent of the condition and herd risk factors. Vet J. 2012;193:234–9. 49. Villarreal I, Meyns T, Dewulf J, Vranckx K, Calus D, Pasmans F, et al. The effect of vaccination on the transmission of mycoplasma hyopneumoniae in pigs under field conditions. Vet J. 2011;188:48–52. 50. Holt HR, Alarcon P, Velasova M, Pfeiffer DU, Wieland B. BPEX pig health scheme: a useful monitoring system for respiratory disease control in pig farms? BMC Vet Res. 2011;7:82. 51. Welfare Quality Consortium. Welfare Quality® assessment protocol for pigs. Lelystad: Welfare Quality Consortium; 2009. 52. Van Staaveren N, Teixeira DL, Hanlon A, Boyle LA. Pig carcass tail lesions: the influence of record keeping through an advisory service and the relationship with farm performance parameters. Animal. 2017;11:140–6. 53. Nathues H, Alarcon P, Rushton J, Jolie R, Fiebig K, Jimenez M, et al. Cost of porcine reproductive and respiratory syndrome virus at individual farm level – an economic disease model. Prev Vet Med 2017;142:16–29. https://doi. org/10.1016/j.prevetmed.2017.04.006. Publisher’s Note Springer Nature remains neutral with regard to jurisdictional claims in published maps and institutional affiliations.