Obtaining a representative estimate of discard mortality for population and ecosystem assessments is very challenging. This can only rarely be done directly by recovering tagged discarded individuals. Instead, semi-quantitative measures of individual fish vitality or physical condition, obtained by onboard observers prior to discarding, can be used. Such vitality measures can be a good indicator of discard mortality, and by virtue of the data collection method, should also reflect the condition of discards throughout the fishery. Furthermore, vitality can be predicted using covariates known to affect discard mortality, allowing for a more general assessment. We argue that a representative mortality rate can be estimated using the product of at least two probabilities: that of belonging to a particular vitality class, conditional on the factors experienced during capture and catch handling; and the probability of surviving the event, conditional on pre-release vitality. Here we estimate mortalities for five fish taxa captured in southern Gulf of St. Lawrence fisheries. The conditional survival probabilities were obtained using survival analysis of data from experiments in which fish were captured using commercial fishing methods and held to assess short-term mortality (2-3 days). The analysis included a mixture model with a fraction of unaffected individuals, which appears appropriate for data from bycatch mortality studies. Based on this study and the mechanistic interpretation of the mixture model, short-term monitoring of discard mortality may be sufficient to characterize longer term impacts in a number of taxa.