A machine learning approach using partitioning around medoids clustering and random forest classification to model groups of farms in regard to production parameters and bulk tank milk antibody status of two major internal parasites in dairy cows

Perinatal mortality in German dairy cattle: Unveiling the importance of cow-level risk factors and their interactions using a multifaceted modelling approach

Perinatal mortality (PM) is a common issue on dairy farms, leading to calf losses and increased farming costs. The current knowledge about PM in dairy cattle is, however, limited and previous studies lack comparability. The topic has also primarily been studied in Holstein-Friesian cows and closely related breeds, while other dairy breeds have been largely ignored. Different data collection techniques, definitions of PM, studied variables and statistical approaches further limit the comparability and interpretation of previous studies. This article aims to investigate the factors contributing to PM in two underexplored breeds, Simmental (SIM) and Brown Swiss (BS), while comparing them to German Holstein on German farms, and to employ various modelling techniques to enhance comparability to other studies, and to determine if different statistical methods yield consistent results. A total of 133,942 calving records from 131,657 cows on 721 German farms were analyzed. Amongst these, the proportion of PM (defined as stillbirth or death up to 48 hours of age) was 6.1%. Univariable and multivariable mixed-effects logistic regressions, random forest and multimodel inference via brute-force model selection approaches were used to evaluate risk factors on the individual animal level. Although the balanced random forest did not incorporate the random effect, it yielded results similar to those of the mixed-effect model. The brute-force approach surpassed the widely adopted backwards variable selection method and represented a combination of strengths: it accounted for the random effect similar to mixed-effects regression and generated a variable importance plot similar to random forest. The difficulty of calving, breed and parity of the cow were found to be the most important factors, followed by farm size and season. Additionally, four significant interactions amongst predictors were identified: breed-calving ease, breed-season, parity-season and calving ease-farm size. The combination of factors, such as secondiparous SIM breed on small farms and experiencing easy calving in summer, showed the lowest probability of PM. Conversely, primiparous GH cows on large farms with difficult calving in winter exhibited the highest probability of PM. In order to reduce PM, appropriate management of dystocia, optimal heifer management and a wider use of SIM in dairy production are possible ways forward. It is also important that future studies are conducted to identify farm-specific contributors to higher PM on large farms.

Random forest classification as a tool in epidemiological modelling: Identification of farm-specific characteristics relevant for the occurrence of Fasciola hepatica on German dairy farms

Fasciola hepatica is an internal parasite of both human and veterinary relevance. In order to control fasciolosis, a multitude of attempts to predict the risk of infection such as risk maps or forecasting models have been developed. These attempts mainly focused on the influence of geo-climatic and meteorological features. Predicting bovine fasciolosis on farm level taking into account farm-specific settings yet remains challenging. In the present study, a new methodology for this purpose, a data-driven machine learning approach using a random forest classification algorithm was applied to a cross-sectional data set of farm characteristics, management regimes, and farmer aspects within two structurally different dairying regions in Germany in order to identify factors relevant for the occurrence of F. hepatica that could predict farm-level bulk tank milk positivity. The resulting models identified farm-specific key aspects in regard to the presence of F. hepatica. In study region North, farm-level production parameters (farm-level milk yield, farm-level milk fat, farm-level milk protein), leg hygiene, body condition (prevalence of overconditioned and underconditioned cows, respectively) and pasture access were identified as features relevant in regard to farm-level F. hepatica positivity. In study region South, pasture access together with farm-level lameness prevalence, farm-level prevalence of hock lesions, herd size, parity, and farm-level milk fat appeared to be important covariates. The stratification of the analysis by study region allows for the extrapolation of the results to similar settings of dairy husbandry. The local, region-specific modelling of F. hepatica presence in this work contributes to the understanding of on-farm aspects of F. hepatica appearance. The applied technique represents a novel approach in this context to model epidemiological data on fasciolosis which allows for the identification of farms at risk and together with additional findings in regard to the epidemiology of fasciolosis, can facilitate risk assessment and deepen our understanding of on-farm drivers of the occurrence of F. hepatica.

Associations of production characteristics with the on-farm presence of Fasciola hepatica in dairy cows vary across production levels and indicate differences between breeds

Fasciola hepatica is one of the economically most important endoparasites in cattle production. The aim of the present work was to evaluate the relevance of production level on the associations of on-farm presence of F. hepatica with farm-level milk yield, milk fat, and milk protein in Holstein cows, a specialised dairy breed, and in Simmental cows, a dual purpose breed. Furthermore, we investigated whether differential associations were present depending on breed. Data from 560 dairy farms across Germany housing 93,672 cows were analysed. The presence of F. hepatica antibodies was determined via ELISA on bulk tank milk samples. Quantile regression was applied to model the median difference in milk yield, milk fat, and milk protein depending on the interaction of breed and fluke occurrence. Whereas a reduction in milk yield (-1,206 kg, p < 0.001), milk fat (-22.9 kg, p = 0.001), and milk protein (-41.6 kg, p <0.001) was evident on F. hepatica positive German Holstein farms, only milk fat (-33.8 kg, p = 0.01) and milk protein (-22.6 kg, p = 0.03) were affected on F. hepatica positive German Simmental farms. Subsequently, production traits were modelled within each of the two breeds for low, medium, and high producing farms in the presence of F. hepatica antibodies and of confounders. On Holstein farms, the presence of F. hepatica seropositivity was associated with lower production, while on German Simmental farms such an association was less evident. This work demonstrates that production level is relevant when assessing the associations between the exposure to F. hepatica with production characteristics. Moreover, both models indicate a breed dependence. This could point towards a differential F. hepatica resilience of specialised dairy breeds in comparison with dual purpose breeds.

Multinomial logistic regression based on neural networks reveals inherent differences among dairy farms depending on the differential exposure to Fasciola hepatica and Ostertagia ostertagi

Fasciola hepatica and Ostertagia ostertagi are cattle parasites with worldwide relevance for economic outcome as well as animal health and welfare. The on-farm exposure of cattle to both parasites is a function of host-associated, intrinsic, as well as environmental and farm-specific, extrinsic, factors. Even though knowledge on the biology of both parasites exists, sophisticated and innovative modelling approaches can help to deepen our understanding of key aspects fostering the exposure of dairy cows to these pathogens. In the present study, multiple multinomial logistic regression models were fitted via neural networks to describe the differences among farms where cattle were not exposed to either F. hepatica or O. ostertagi, to one parasite, or to both, respectively. Farm-specific production and management characteristics were used as covariates to portray these differences. This elucidated inherent farm characteristics associated with parasite exposure. In both studied regions, pasture access for cows, farm-level milk yield, and lameness prevalence were identified as relevant factors. In region 'South', adherence to organic farming principles was a further covariate of importance. In region 'North', the prevalence of cows with a low body condition score, herd size, hock lesion prevalence, farm-level somatic cell count, and study year appeared to be of relevance. The present study broadens our understanding of the complex epidemiological scenarios that could predict differential farm-level parasite status. The analyses have revealed the importance of awareness of dissimilarities between farms in regard to the differential exposure to F. hepatica and O. ostertagi. This provides solid evidence that dynamics and relevant factors differ depending on whether or not cows are exposed to F. hepatica, O. ostertagi, or to both.

Establishment and Validation of Fourier Transform Infrared Spectroscopy (FT–MIR) Methodology for the Detection of Linoleic Acid in Buffalo Milk

Buffalo milk is a dairy product that is considered to have a higher nutritional value compared to cow’s milk. Linoleic acid (LA) is an essential fatty acid that is important for human health. This study aimed to investigate and validate the use of Fourier transform mid-infrared spectroscopy (FT-MIR) for the quantification of the linoleic acid in buffalo milk. Three machine learning models were used to predict linoleic acid content, and random forest was employed to select the most important subset of spectra for improved model performance. The validity of the FT-MIR methods was evaluated in accordance with ICH Q2 (R1) guidelines using the accuracy profile method, and the precision, the accuracy, and the limit of quantification were determined. The results showed that Fourier transform infrared spectroscopy is a suitable technique for the analysis of linoleic acid, with a lower limit of quantification of 0.15 mg/mL milk. Our results showed that FT-MIR spectroscopy is a viable method for LA concentration analysis.