Skip to main content
Conference paper, 2009

Which cows are veterinary treated for mastitis, and when? – a multilevel analysis

Jansson, mörk Marie; Egenvall, Agneta; Lindberg, Ann; Vågsholm, Ivar


The objective of this study was to evaluate what herd- and cow characteristics affect whether a cow receive veterinary treatment for clinical mastitis. Mastitis events were reported by farmers during January, April, July and October 2004 and only data from herds with ≥4 mastitis events were considered (746 events from 87 herds). We fitted a two-level logistic model (herd and cow) with the dependent variable “veterinary treatment” (yes/no) using log likelihood estimation in Stata. The final model was also estimated using the second-order penalized quasi-likelihood (restricted iterative generalised square algorithm) and the Markov-chain Monte Carlo procedures in MLwiN. The median number of events per herd was 9 (range 4-50) and the median proportion of veterinary treated events per herd was 0.9 (range 0-1). The estimates and standard errors were reasonably similar in all estimation procedures. There were higher odds for veterinary treatment early in lactation and if there was another animal with a disease event on the same day. Events in January had higher odds for veterinary treatment than events in July and October. Also, the odds for veterinary treatment increased with increasing herd-average milk yield. The herd-level variance accounted for 63% of the total residual variance. In conclusion, the differences found could potentially affect the internal validity in studies based on farmers’ reports or veterinary records


dairy cow; disease monitoring system; hierarchical models; misclassification bias

Published in


12th International Symposium on Veterinary Epidemiology and Economics