Statistical modelling for clinical mastitis in the dairy cow: problems and solutions

Area of Science:

• Veterinary Epidemiology
• Statistical Modelling
• Dairy Cattle Health

Background:

• Clinical mastitis is a significant multifactorial disease in dairy cows.
• Accurate statistical modeling is crucial for understanding its occurrence and risk factors.
• Model validity depends on statistical unit independence and avoiding overdispersion.

Purpose of the Study:

• To identify sources of overdispersion in clinical mastitis risk models at various study levels (herd, lactation, animal, udder, quarter).
• To discuss solutions for controlling overdispersion across different modeling approaches.
• To evaluate the contribution of explanatory models versus generalist exploratory models in improving accuracy and relevance.

Main Methods:

• Identification of overdispersion sources at different levels of mastitis risk perception.
• Discussion of methods to control overdispersion at each study level.
• Comparison of generalist exploratory models with explanatory models, including state-space and survival models.

Main Results:

• Overdispersion is a key challenge in mastitis modeling, with sources varying by study level.
• Explanatory models, particularly state-space and survival models, offer improved accuracy and relevance.
• Methodological choices significantly impact study result comparability.

Conclusions:

• Addressing overdispersion is critical for robust clinical mastitis modeling.
• Explanatory modeling approaches enhance prediction accuracy for udder infections.
• Future research should focus on integrating these methods for a global predictive model.