Individual piglet birth weight (BW0) and within-litter birth weight variability (BWvar) strongly affect preweaning survival. Piglet mortality in commercial pig operations poses significant economic and ethical concerns, as well as animal welfare implications. The objective of this study was to evaluate the effects of BW0, BWvar and other potential sow and environmental factors, including parity, sex, litter size, year, season and farm, on piglet survival from birth to weaning age using a logistic regression approach. The study determined the critical threshold values for BW0 and BWvar, both separately and for a combined index to predict preweaning survival. This was done through a receiver operating characteristic (ROC) curve analysis. Data consisted of 68 394 piglet records from 1 661 sows obtained from two research farms; the Research Institute for Farm Animal Biology in Germany and Agroscope in Switzerland. The BW0 and BWvar significantly influenced piglet survival at birth, but their influence changed with piglet age and management interventions such as cross-fostering. The BW0 exerted the greatest effect on survival, with the probability of survival increasing with increasing BW0 following a curvilinear trend. A significant observation was that BWvar was more important than litter size in determining piglet survival at birth. The ROC analysis revealed that piglets below a BW0 cut-off value of 1.18 kg (accuracy = 0.73) had a lower survival probability at birth compared to their heavier counterparts. With a cut-off value of 0.277 g (accuracy = 0.50), the BWvar amongst total piglets born predicted survival less accurately. Piglets with low BW0 and born in litters with high BWvar had the lowest survival probabilities. Consequently, two novel indexes were developed, namely the birth weight-to-variation ratio and the birth weight-to-variation composite index, to offer a comprehensive assessment of piglet viability. The results suggest that using the derived indexes for predicting piglet survival was more informative (accuracy = 0.89) than relying solely on BW0 or BWvar. This study demonstrates a robust methodology for the identification of low-viability piglets using fundamental and easy-to-measure birth weight traits.
