Efforts to quantify disturbances to aquatic systems often use landscape-level metrics, presumably linked to ecological integrity, but fewer studies have directly linked ecological integrity to instream habitat, and applied these results to unsampled stream reaches throughout a landscape. We developed a flexible, quantitative approach that characterizes stream impairment across a landscape and identifies least-disturbed stream reaches to serve as benchmarks for high quality physical habitat and ecological integrity. Fish and macroinvertebrate community characteristics, reach-level physical habitat and water quality metrics were summarized in 891 wadeable stream reaches across two ecoregions in Missouri, USA. The influence of reach and water-quality characteristics as well as landscape-level variables on 7 fish and 3 macroinvertebrate community biological indicator metrics was then modeled using boosted regression trees (BRTs). On average, reach-level models explained more variance (25 and 27% in the two ecoregions examined) in biotic metrics than landscape-level models (18% and 20%). Abiotic and biotic associations differed among ecoregions and stream sizes, however, reach-level habitat (e.g., bankfull width/depth ratio, channel incision height) and water quality (e.g., dissolved oxygen, total chlorophyll) were consistently top predictors. At the landscape scale, fish richness in the agriculturally dominated ecoregion increased with decreased fragmentation/flow modification, while invertebrate metrics responded to agricultural disturbances. Invertebrate metrics in the forested ecoregion showed community degradation apparent with crop coverage as low as 8-10% of the riparian zone, while urban impairment was best detected using invertebrate indicators of biotic integrity and measures of fish trophic ecology. Relationships among landscape-scale variables and reach characteristics identified as top predictors in BRTs also highlighted potential mechanistic relationships among landscape, habitat, and measures of ecological integrity. Using the results of the landscape-level models, estimates for overall ecological integrity were predicted for over 28,000 stream reaches throughout Missouri, and a total of 1,423 candidate reference reaches were identified. The objective approach to characterizing stream impairment developed in this study offers specific advantages, including a reach and landscape-level evaluation of human disturbance as well as an inductive, multi-metric determination of ecological integrity.