We developed and evaluated a multimetric index of lake diatom condition (LDCI) based on surface-sediment samples for the National Lake Assessment (NLA) by the US Environmental Protection Agency. We selected final metrics in each of 5 categories for use in composite metrics of biological condition, which we combined in a hierarchical multimetric index. The final metrics selected for the LDCI had responses as predicted based on ecological principles, wide ranges, high signal-to-noise ratios, the ability to distinguish reference from disturbed lakes, and low intercorrelation. The final metrics were: Shannon diversity, taxon richness, % reference taxa, % tolerant taxa, % epiphytic individuals, % chain-forming individuals, % low-P taxa, % low-N taxa, % high-P taxa, % high-N taxa, % Achnanthidium individuals, % Cocconeis individuals, and % Cyclotella Stephanodiscus individuals. We developed and tested models for adjusting the expected reference LDCI at a site for natural variation among sites. The best model used lake morphology, watershed attributes, and climate variables to predict expected reference LDCI for each lake. The adjusted LDCI was the observed LDCI at a lake minus the expected reference LDCI modeled for that lake. The adjusted LDCI varied less among reference sites than the LDCI, indicating the adjusted LDCI better accounted for natural variability among sites than the LDCI. However, it had less power than the LDCI to separate reference and disturbed lakes, a result attributed to lower values of expected reference LDCI in landscapes with high agricultural land use. The adjusted LDCI is the first multimetric diatom index developed for lakes, applied at a scale as large and diverse as the USA, and corrected for natural variation among lakes. According to the adjusted LDCI, 47.1% of USA lakes were in reference condition, 27.0% were in fair condition, and 23.2% were in poor condition.