# Best statistical approach...?

I am currently thinking about different statistical methods and which is the most suitable for my question. I have a questionnaire with 12 dichotomous items (yes/no) that ask about stressful situations that could hypothetically lead to anxiety symptoms. In addition, I also assess the extent of anxiety symptoms using a validated questionnaire, leading to a quantitative outcome (sum score). Age is also asked as a possible covariate. All but two of the binary variables (stressful situations) correlate significantly positively with the anxiety outcome.

If I now calculate a hierarchical regression with anxiety as the dependent variable, in which I include age as a covariate in the first step and all binary variables in the second step, I would be able to see which of the situations explain particularly much variance in the dependent variable as soon as they become significant (please correct me if I am wrong here), but I would be trapped in the framework of my measurement instrument. The latter would be because the inclusion of the situations is predetermined by the measurement instrument and if the results fluctuate, as long as fewer variables are included in the regression analysis, there could even be interactions between the situations - right? And is it even feasible to run a hierarchical regression with so many binary predictors (N = 150, VIF/ Tolerance values seem all to be OK)?

Another idea would be latent class analyses, where at least all situations are considered and people are classified according to their response behavior. Here one could see how these persons are related to the outcome and other descriptive variables. The problem is, however, that such an analysis with my data yields exactly two classes: People who have experienced a lot of stressful situations vs. people who have experienced less stressful situations.

Does anyone have any ideas about the most sensible way to proceed here?