How do you correct dates in a data.frame based on another column by groups

I have a data.frame containing drug prescriptions (of the same drug) including the date of the prescription and the amount of daily doses. There is one row per drug prescription. Each patient in the data.frame has a unique idnr and each unique idnr can have several fillings of prescriptions (ie several rows). The data is arranged by first idnr then filldate.

The purpose of the desired output is to create a variable indicating discontinuation of treatment which will be defined as <(-30) doses left at time of subsequent refill. The date of discontinued treatment will then be filldate + daily_doses + lag(doses_left_at_time_of_subsequent_refill) calculated from the row which had <(-30) doses left at time of the subsequent refill. A patient can have multiple dates of discontinuation.

Example data:

idnr     daily_doses     filldate
1.       10            2000-01-01
1.       10            2000-01-12
1.       10            2000-01-15
1.       10            2000-01-20
1.       10            2002-03-21
1.       10            2002-03-30
1.       10            2002-04-20
2.       10            2004-05-01
2.       20            2004-05-11
2.       10            2004-05-24
2.       10            2004-06-01
3.       20            2000-01-01
3.       10            2010-03-04
3.       10            2010-03-04
3.       10            2010-03-04
3.       10            2010-03-08
3.       10            2012-07-08
3.       10            2012-07-18
3.       10            2012-07-30
3.       10            2012-08-15
3.       10            2012-09-25

As data.frame:

data.frame(
         idnr = c(1L,1L,1L,1L,1L,1L,1L,2L,2L,2L,
                  2L,3L,3L,3L,3L,3L,3L,3L,3L,3L,3L),
  daily_doses = c(10L,10L,10L,10L,10L,10L,10L,10L,
                  20L,10L,10L,20L,10L,10L,10L,10L,10L,10L,10L,10L,10L),
     filldate = as.Date(c("2000-01-01","2000-01-12","2000-01-15",
                  "2000-01-20","2002-03-21","2002-03-30","2002-04-20",
                  "2004-05-01","2004-05-11","2004-05-24","2004-06-01",
                  "2000-01-01","2010-03-04","2010-03-04","2010-03-04","2010-03-08",
                  "2012-07-08","2012-07-18","2012-07-30","2012-08-15",
                  "2012-09-25"), format = "%Y-%m-%d")
)

I want to identify those who have not refilled their prescription within 30 days with the following conditions:

If a patient has filled prescriptions earlier than needed, illustrated by amount of days covered by the daily doses, the doses should be assumed to be used up before starting the next fill. In other words, there should be a variable indicating how many doses are left at the next fill date so that this information is not lost when defining date of discontinuation. If there is no subsequent fill date the variable should indicate there is -Inf doses left

If a patient however has filled prescriptions later than supposed to, according to the amount of days covered by the daily dose, the variable indicating how many doses there are left on the next prescription should not be adjusted.

Desired output:

idnr     daily_doses     filldate    doses_left_at_time_of_subsequent_refill
1.       10            2000-01-01       -1
1.       10            2000-01-12       7
1.       10            2000-01-15       12
1.       10            2000-01-20       -769
1.       10            2002-03-21       1
1.       10            2002-03-30       10
1.       10            2002-04-20       -Inf
2.       10            2004-05-01       0
2.       20            2004-05-11       7
2.       10            2004-05-24       11
2.       10            2004-06-01       -Inf
3.       20            2000-01-01       -3715
3.       10            2010-03-04       10
3.       10            2010-03-04       20
3.       10            2010-03-04       26
3.       10            2010-03-08       -817
3.       10            2012-07-08       0
3.       10            2012-07-18       -2
3.       10            2012-07-30       -6
3.       10            2012-08-15       -31
3.       10            2012-09-25       -Inf

I have tried by shifting the fill date forward but i can only account for just 1 previous row in the groups of idnr. This was done using the lag() function but since patients has different number of prescriptions and since some patients have hundred of prescriptions it is not efficient using lag(). I also tried the cumsum() function but then I have issues with doses hoarding up from previous treatment episodes (which should already be discontinued by the -30 doses left at time of subsequent refill definition and thus not accounted for in newer treatment episodes).