This paper addresses intrasentential code-switching in US American Danish spoken by 1st generation immigrants from Denmark, who migrated to the US around 1900 as part as the transatlantic European mass emigration and their descendants (2nd and 3rd generation). The analysis is based on the Corpus of American Danish, specifically a dataset of 173 speakers producing 46 hours of speech. In this dataset, we observe significantly different patterns of intrasentential code-switching in the speech of the immigrant speakers (1st generation) and US-born heritage speakers (2nd and 3rd generation): The code-switching patterns of the heritage speakers show a preference for English lexemes that are integrated morphologically into Danish or which are part of Danish-English bilingual compounds. In contrast, the immigrant speakers prefer non-integrated English words for code-switching. This result taken per se shows that code-switching patterns show variation across generations just as other linguistic variables. Taking the result further, we have connected it to a previous study of representations of linguistic proficiency in immigrant Danish in the Americas. By this, we are able to show that the morphologically or lexically integrated code-switching of the heritage speakers correlate with features representing fluent speech, while the non-integrated code-switching of the immigrant speakers rather seems to correlate with features showing a low activation of Danish in a situation of language shift. Thus, the heritage speakers seem to have developed a way of speaking US American Danish where English word stems are an integrated part of speaking fluently and lexically varied.