Association Between Sensor-Based Prepartum Behaviour and Early Postpartum Health in Dairy Cows

The transition period in dairy cows is characterised by significant physiological changes, increasing their vulnerability to periparturient disease. This study examined whether prepartum behavioural sensor data could predict postpartum health outcomes. Nineteen Holstein-Friesian cows were monitored from 2 weeks before to 2 weeks after calving, using neck and leg accelerometers to record behaviour and location. For the 8 days preceding calving, we calculated the mean, variance, autocorrelation, and non-periodicity for the following behaviours: lying, standing, ruminating, eating, inactivity, step counts, and neck movements. Mean and variance of Earth Mover’s Distance values from day-to-day occupancy heatmaps were also derived. Postpartum health was evaluated three times weekly, and a total deficit score (TDS), as a measure for resilience, was calculated for the 2 weeks after calving. Behavioural variables associated with TDS (p < 0.25) were included in a general linear model adjusting for parity. In the final model, greater prepartum standing time and higher neck-movement rates were associated with lower TDS, indicating greater resilience (R2 = 0.72; p < 0.001). Parity was not significant. Although limited by a small sample size, these findings highlight the promise of AI-supported sensor analytics for early identification of cows at risk of postpartum health problems.