Assuming you are using Postgres you can download the 8 Geofabrik regional Extracts and use Osmosis and Osm2Pgsql to load a data subset of each region except for the North America dataset, then load the complete North America dataset
I have read about, but not encountered issues relating to the same OSM ID existing in two different datasets (on the edges of the extract) and the database insert failing due to the unique primary key constraints in Postgres
Alternatively, you can load the planet file then delete the data not needed. Osm2pgsql ends up with 4 tables planet_osm_line which contains all the ways), planet_osm_road which is a road only subset of planet_osm_line (which you could drop), planet_osm_points (pois which you could just drop) and plamet_osm_polygon which has all the polygon data. You could delete delete all entries in lines and polygons and only keep what you needed based on the tags in the database and a bounding box/poly outside North America. (Use PostGis ST_Contains, ie, is Not Contained in the bounding box)