Multiple imputation analyses

Dear all,
I have a methodological question about multiple imputation analyses.
I have 5 variables Q1-Q5 on ordinal level and i plan to crate a one variable on oridnal level (modified Rankin Scale, the method is aproved in some studies).
My question is: should i perform Multiple imputation analyses on each variable Q1 - Q5 or it is better to calculate mRS and perform multiple imputation analyses on mRS?

(When i performed MI analyses on Q1-Q5 i have got lovely data, but i have not got pooled data so i could calculate mRS. I have done missing value analysis showing that the missing values were missing at random).

Thank you very much for your answer.

Best regards!