There are several elements to this. First, whether the question stem has the same meaning in all the countries, secondly whether the response categories for capturing the answer are equivalent and thirdly whether the question was asked to the same population. While the dataset will contain variables that are either simply direct responses to a question posed by the interviewer or are constructed from responses to a series of questions it is good practice to always check back to the source question to satisfy yourself that the question has been transformed into the variable in a standard way and that you fully understand the concept captured in the variable. For countries that share a language this is more straightforward than where the question has been translated or adapted but even where countries share a language this may not always mean that there are comparable. Consider for example comparisons between the US and the UK. The question will rarely be exactly the same as each country will reflect their own use of English and differences in vocabulary. Have the questions been adapted adequately to reflect the use of language in both countries?