Hi what you asked for is not really what you want. You really want multivariate ordinal logistic regression because your multiple DVs are ordinal. Stepwise regression doesn't work and I am attaching a paper on that with many references to show why that is true. I recommend instead adaptive lasso variable selection which does work and is available for multivariate ordinal logistic regression. I have attached many references that need to be read prior to doing your project correctly. I apologize for all the reading but in the end you will have a paper that is correct. My advice is that it's well doing this reading. Please ask any additional questions. Best wishes, David Booth

https://www.google.com/url?sa=t&source=web&rct=j&url=https://www.statisticssolutions.com/free-resources/directory-of-statistical-analyses/ordinal-regression-2/&ved=2ahUKEwjE0qHN1abxAhWI2BQKHaFMDn4QFjAAegQIAxAC&usg=AOvVaw0o4tcOIBQ44_rsgAARt3Ux

Article mvord : An R Package for Fitting Multivariate Ordinal Regres...

In my area of practice, structural equation modeling (SEM) is a preferred approach for multivariate ordinal outcomes. This is mainly because when you have multivariate dimensions become exponentially complex for interpretation. In SEM, you are summarizing multivariate outcomes into latent variables (make sure you use polychoric correlation matrix for confirmatory factor analysis part), which in turn can be included for your analysis. Yes, you can take step-wise regression approach for SEM by simply adding/removing variables. I do not know what statistical software you use, but SEM can be applied in SPSS (AMOS), SAS (tcalis), R (lavaan or openmx packages), or STATA.