During the last three decades, explaining cause-effect relationships between natural and anthropogenic disturbances with measures of stream health have motivated the growing application of statistical, machine learning, and soft computing methods.The aim of this review is to provide insight into the most widely used methods for predicting biological variables based on macroinvertebrate and fish species in riverine ecosystems. Therefore, we describe several methods including multiple linear regression, generalized linear models, generalized additive models, boosted regression trees, random forests, artificial neural networks, fuzzy logic-based, and Bayesian belief networks along with recent applications of these. Moreover, issues regarding variable selection, model interpretability, ensemble modelling, and model evaluation and overfitting are discussed. Recent advances have suggested the need for integrated modelling systems to enhance the predictive ability and improve interpretability.However, trade-offs between model complexity and accuracy demand research efforts in uncertainty quantification/propagation in model ensembles. Additionally, models should be perceived as complementary tools that require further validation with field measurements. Therefore, a consensus regarding monitoring and modelling practices for stream health applications is recommended.