Anthropogenic disturbances particularly affect biodiversity in sensitive freshwater ecosystems by causing species loss. Thus, measuring the response of species to multiple disturbances is a key issue for conservation and environmental management. 2. As it is not practical to assess the response of every species in a community, we compared the performance of trait and taxonomic-based groupings of species for their abilities to predict species loss in a threatened freshwater fish assemblage. Specifically, we examined responses of a Fijian freshwater fish assemblage to deforestation, placement of anthropogenic barriers (overhanging culverts) and the presence of introduced cichlids. 3. Species grouped by traits showed more consistent responses to disturbances than taxonomic groups. In particular, species belonging to trait groups that were estuary associated favoured medium-to-hard substrate, while feeding specialists were highly likely to be absent in catchments with high deforestation and overhanging culverts. The presence of introduced cichlids (Oreochromis mossambicus and O. niloticus) had a smaller effect than deforestation and barriers, but was negatively associated with species richness of diadromous species with climbing ability and positively associated with presences of some piscivores. The trait groups also revealed that detritivores, species favouring soft substrate, and those with a broad dietary range were less sensitive to anthropogenic disturbances. 4. Our study indicates that using traits to predict species loss from disturbed environments can aid in detecting the responses of rare species to disturbance. In addition, we provided a method to estimate the consistency of species’ responses to disturbance. This study may ultimately help managers identify the most effective actions for conserving sensitive species that are seldom recorded in surveys.