Please could you provide additional information. For instance, information on the test type, lag length and whether you included only constant or constant and trend.

However, if you have considered all the possible options and some of the variables are non-stationary at the levels, it is necessary to do the first difference for those variable.

Moneim Khaled It is mainly depends upon your objectives. If only wish to check level of integration then it is okay. If you wish to go further and examine relationships in between variables then read about Panel ARDL (Or PMG) model. But additional details like N and T are necessary to determine model selection.