Should I use item analysis or DFA?


I am working a study that is looking at if items on a anxiety questionnaire that's most often used with adults is applicable to young children as well. I want to find out which questions I can remove from this survey to make it more applicable to kids. For example, one question would be if the individual has a job/how long they work every week. This wouldn't be applicable to children, so i'd have to take it out. I also want to understand if some questions apply to everyone regardless of age (these would be 'true' measures of anxiety). The items are measured on a likert scale of strongly agree to strongly disagree that I've assigned values from 0-4 to. just from looking at the data I can tell that some questions are more frequently endorsed or have higher scores in kids with anxiety than other items. I want to get rid of questions that elicit similar endorsements (like 90% of kids with anxiety endorse item A and so do 70% of kids without anxiety. This item wouldn't be the best item to help diagnose anxiety).

We have data for gender, age, and all the responses to the survey questions. We also have their anxiety diagnosis. We gave this to both kids (under 15) with and without anxiety and adults. The IV is anxiety diagnosis and the DV is responses to each item on the survey.

I'm wondering if an item analysis is the best way to go to figure out which questions best capture anxiety diagnosis in children or if there was a better way - maybe like a discriminant function analysis?

Thank you