The quality of geodata on OpenStreetMap (OSM) is characterized by heterogeneity. Results of conventional approaches of quality assessment based on the comparison with reference data can thus not be applied to the entire dataset. Hence, the completeness of OSM data is usually analysed using intrinsic methods. Here, we propose a novel approach that consists in the generation of reference data via a regression model. Using demographic and socio-economic raster datasets and sporadically available GTFS data, the number of local bus routes is predicted on the cells of a global hexagon grid. The results are compared with the OSM dataset.