Assuming you're using the standard output in Lighthouse studio for HB analysis, the lower and upper bounds of your confidence interval do this. If the upper to lower range includes zero, then your level's utility is not significantly different from zero. Ditto for your attribute importances.
Note that because of the fact that attributes are zero centered, it will not be at all unusual for attributes with more than two levels to have some levels that are not significantly different from zero, if you code your attributes as categorical variables - this is an artifact of the coding scheme and if you used dummy codes instead of zero-centering effects codes that are the default in the software, you might find different levels that are not significantly different from zero. It's generally not a good idea to get all hung up on which specific levels are or aren't significantly different from zero when the answer depends on the specific coding you're using.