You could examine all three outcome variables in a single model using path analysis. Having a single model (rather than a separate model for each outcome) is often preferable. You just need to make sure that you properly account for the scale level/distributions of your dependent variables. Some (but not all) programs for path analysis and SEM (e.g., Mplus) allow you to run path analysis for non-metrical (e.g., binary, ordinal, count etc.) outcome variables.

Thank you very much dr. Christian Geiser.

In your case, it's important to understand that SEM and logistic/linear regression are different techniques and they are used to answer different questions. SEM is a multivariate approach, where you can estimate the relationships between multiple variables simultaneously. However, logistic/linear regression is univariate, where you can estimate the relationship between one dependent variable and one or more independent variables.

If you have found that some variables that were significant in the logistic/linear regression turned non-significant in the SEM, it could be due to the fact that SEM is estimating the relationships between multiple variables simultaneously, and this may lead to different results.

Thank you so much Prof. Dinesh Kumar