Can regression to the mean be corrected by linear mixed effects?

I am wondering if regression to the mean can be corrected by using linear mixed effects models in the following case.

model1 = lme(fixed = fat ~ visit  + diabetes  + Sex  + age , random = ~ visit| PatientID, data = df,control = lmeControl(opt = "optim"))

As visit is nested in patient ID. The model knows that visit1 from patient 1 is the baseline of visit2 from patient1.

alternatively, I am wondering if correcting the variance of visit would help?

mod2.lme <- update(model1, weights = varIdent(~ 1|visit))

Am I right?

Thanks!