Based on changes in their drinking patterns, predictions of tail biting and diarrhea amongst growing pigs were generated and evaluated. A spatial, multivariate dynamic linear model was used to model simultaneously monitored drinking patterns across multiple pens in a section of weaners, and across multiple pens in a section of finishers. Both volume (litres/hour) and frequency (bouts/hour) were monitored and modelled. The predictive performances of the two sets of data were evaluated separately for both weaners and finishers using a two-sided tabular CUSUM. Time windows, including 24 hours and 48 hours before the day of the event, were applied and the predictive performances were reported as Area Under the ROC Curve (AUC). Generally, the predictive performances were high for both volume and frequency (Tail biting finishers: AUC = 0.79 - 0.88, Tail biting weaners: AUC = 0.89 - 0.92). For finishers (30-110 kg), frequency (bouts/hour) was a better predictor of outbreaks of both tail biting and diarrhea as well as either of the events given either time window. For weaners (7- 30 kg), however, the performances for predicting tail biting were equally high for volume and frequency given either length of the time window. For the prediction of diarrhea or either of the events, the performances were also close to equal between volume (litres/hour) and frequency (bouts/hour) but they differed between the two lengths of time windows. In conclusion, frequency tends to be a better predictor of events amongst finishers, whereas the differences in predictive performances between volume and frequency were less distinct amongst weaners.