Dear Stata society,

I am currently analysing the impact of Covid19 on health professionals work, through survey responses (n>2500) where I want to investigate the relationship between respondents: 1) feeling of safety while working; 2) Sense of competence in providing care; 3) impact on general health; 4) impact on work morale during Covid19.
These dependent variables are given on a 4-point Likert scale, where (1 indicating no or low impact/low sense of safety/competence and 4 = High impact/ High sense of safety/competence).

Independent variables are: Country's income class, clinical vs. non-clinical role and levels of restrictions during the pandemic (also an ordered variable, ranging from 1 = No restrictions to 4 = full lock-down).

Initially, I wanted to utilize the ologit model, but when performing the Brant test, I see that the assumption of proportional odds are violated (overall chi2<0.01).

This might come across as a very generic question, but does anyone have any advice in how to proceed? Would the multinomial model be a more appropriate way to go?

Thank you for all your help!