I am using multiple imputation to handle missingness in my dataset of 83 variables across 250 participants. I intend to use 12 of these variables, which comprise a self-compassion questionnaire, in a series of CFA models.

Multiple imputation was conducted with mice and I am using lavaan and lavaan.mi for the factor analysis.

imp2 <- mice(merged_data, m = 5, maxit = 10, 
             predictorMatrix = predM, 
             method = meth, seed = 12345)

# lavaan model
cfa1 <- ' scs  =~ NA*SCS_1 + SCS_2 + SCS_3 + SCS_4 + SCS_5 + 
SCS_6 + SCS_7 + SCS_8 + SCS_9 + SCS_10 + SCS_11 + SCS_12
scs ~~ 1*scs'

library(lavaan.mi)
cfa1out <- cfa.mi(cfa1, data=imp2, ordered=c("SCS_1", "SCS_2", "SCS_3", "SCS_4", "SCS_5", "SCS_6",
                                            "SCS_7", "SCS_8", "SCS_9", "SCS_10", "SCS_11", "SCS_12"),
                 estimator = "WLSMV")

summary(cfa1out)
semTools::fitMeasures(cfa1out, fit.measures = c("all"),
                      test = "D2", pool.robust = TRUE, asymptotic = TRUE)

All of this seems to run fine. The problem is that half of the indicator variables I am entering into the lavaan CFA model need to be reverse scored. How can I reverse score a select group of variables that were treated with multiple imputation across 5 datasets? Additionally, how can I sum these and other variables to derive total or subscale scores, since I imputed at the item level?

Thanks in advance for your assistance.