Actually the policy conclusion is not correct, however without a sophisticated understanding of statistics it's easy to be misled.
It is true that a higher proportion of immigrants commit crimes in Sweden than natives. However, if you break down immigrants by socioeconomic status and educational attainment you don't see any difference between immigrants and natives. This is because of something called Simpson's Paradox.
What's happening here is that Sweden is a wealthy advanced country with a low birth rate, and it's been importing low-education poor workers to augment it's own dwindling underclass in filling low-paying jobs. Now poor, uneducated people commit many kinds of crimes at a higher rate that affluent, educated people. Whether they are native or immigrant makes no difference. So what the Swedish statistics actually tell us is that uneducated low-wage workers make up a higher proportion of immigrants than they do of natives, which should be no surprise because that's why the largest proportion of immigrants have been admitted.
This also raises another possibility: you can actually reduce crime rates with immigration, if you let in the right people. In fact there is evidence this is happening in Canada, which places a premium on education and language skills when deciding who to admit.