If the test is to be scored using a norm-referenced approach (such as T-Scores or an IQ scale) it needs to have norms - and these are based on the standardization sample. It is the responsibility of the test developer to engage in this work. Doing it well is expensive, of course. Jan makes good points here.

Translating a test into a new language is a project in its own right. Best practice is to have it translated by one bilingual individual, then have another such individual do a "back-translation" into the original language. This acts as a check on the accuracy of the initial translation. If discrepancies emerge, the process needs to be repeated until the back-translation is essentially identical with the original. If the test includes "culturally loaded" content, then work needs to be done to identify equivalent content in the new culture. For example, if an American test asked about the number or members in the Senate, some equivalent institution in the new country would be used.

The translated test would then need to be "normed" in the new cultural group. When the new country is a relatively small one (like Croatia) or a very poor one, this usually needs to be done in a "home grown" way. That is, no major test developer is going to invest pots of money in the project (it'll never pay!), so local clinicians and/or academics do the best they can with the resources they have available.

The BDI is not a norm-referenced test, though, so there isn't a standardization sample as such. Specific raw scores are identified as "cut offs" for various levels of depression. A single study (with two samples) might show whether this works well enough in Croatia. Get one sample of depressed people (folks seeking treatment) and one of non-treatment-seeking community dwellers to take the BDI. Analyze the results. The depressed people should mostly score above the cutoff; few of the community sample should.

Thank You! It was very helpful.