Partial or Total Overlap of two Date periods

Question

I'm trying to know which observations are active during specific episodes of time

My real objective is to know which Diagnosis (dx) are active during a Pregnancy period (9 months) by a patient. It has to be considered that a patient can have different number of pregnancies along her life and also a different number of diagnosis (the dx can or can't be active ).

I have tried foverlaps like here or here but weren't exactly what I was looking for. The real problem is this and it's well documented but not for R.

Here they make it work for SQL I think. So I hope it's solvable...

I also tried Non-Equi Joins like this but I'm not able to make it work the way I want...

Let's get into it:

I have one DB with patient (id) Diagnosis (dx), the time it was registered (InD_dx) and the time it ended (EndD_dx) like this:

id <- rep("a", 11)
InD_dx <- as.Date(c("2005-10-04","2005-10-06","2005-10-06", "2008-04-07", "2010-05-10", "2012-04-24", "2012-04-24", "2012-05-15", "2014-03-20", "2014-04-22", "2017-11-30"), format = "%Y-%m-%d")
EndD_dx <- as.Date(c("2020-12-31","2020-12-31","2020-12-31", "2008-11-05", "2011-01-17", "2015-07-20", "2013-01-01", "2015-07-20", "2015-12-04", "2020-12-31", "2020-12-31"), format = "%Y-%m-%d")
dx <- c("A", "B", "C", "P", "P", "D", "P", "E", "F", "G", "H")

DT = data.table(id,InD_dx,EndD_dx, dx)

    DT
    id   InD_dx    EndD_dx   dx
 1:  a 2005-10-04 2020-12-31  A
 2:  a 2005-10-06 2020-12-31  B
 3:  a 2005-10-06 2020-12-31  C
 4:  a 2008-04-07 2008-11-05  P
 5:  a 2010-05-10 2011-01-17  P
 6:  a 2012-04-24 2015-07-20  D
 7:  a 2012-04-24 2013-01-01  P
 8:  a 2012-05-15 2015-07-20  E
 9:  a 2014-03-20 2015-12-04  F
10:  a 2014-04-22 2020-12-31  G
11:  a 2017-11-30 2020-12-31  H

I selected the dx of pregnancy (P) and made a DB apart. And a DB with the dx different from P like this:

Pregnancies <- copy(DT[dx== "P"])
Pregnancies 

    id InD_dx     EndD_dx   dx
 1: a 2008-04-07 2008-11-05 P
 2: a 2010-05-10 2011-01-17 P
 3: a 2012-04-24 2013-01-01 P

Dx_Other_than_Pregnancies <- copy(DT[dx!= "P"])
Dx_Other_than_Pregnancies

   id InD_dx     EndD_dx    dx
 1: a 2005-10-04 2020-12-31 A
 2: a 2005-10-06 2020-12-31 B
 3: a 2005-10-06 2020-12-31 C
 4: a 2012-04-24 2015-07-20 D
 5: a 2012-05-15 2015-07-20 E
 6: a 2014-03-20 2015-12-04 F
 7: a 2014-04-22 2020-12-31 G
 8: a 2017-11-30 2020-12-31 H

I was expecting to merge them and group active diagnosis by pregnancy. Thats why I attemped with foverlap:

setkey(Dx_Other_than_Pregnancies, id, InD_dx, EndD_dx)

foverlaps(Pregnancies, Dx_Other_than_Pregnancies,
          by.x=c('id', 'InD_dx', 'EndD_dx'), type='within', mult= "all", nomatch = 0L)

and Non-Equi Joins:

   Dx_Other_than_Pregnancies[Pregnancies, on=.(id, dx_Ini<=dxIni , EndD_dx>=EndD_dx)]

and both end up like:

     id     InD_dx    EndD_dx dx   i.InD_dx  i.EndD_dx i.dx
 1:  a 2005-10-04 2020-12-31  A 2008-04-07 2008-11-05    P
 2:  a 2005-10-06 2020-12-31  B 2008-04-07 2008-11-05    P
 3:  a 2005-10-06 2020-12-31  C 2008-04-07 2008-11-05    P
 4:  a 2005-10-04 2020-12-31  A 2010-05-10 2011-01-17    P
 5:  a 2005-10-06 2020-12-31  B 2010-05-10 2011-01-17    P
 6:  a 2005-10-06 2020-12-31  C 2010-05-10 2011-01-17    P
 7:  a 2005-10-04 2020-12-31  A 2012-04-24 2013-01-01    P
 8:  a 2005-10-06 2020-12-31  B 2012-04-24 2013-01-01    P
 9:  a 2005-10-06 2020-12-31  C 2012-04-24 2013-01-01    P
10:  a 2012-04-24 2015-07-20  D 2012-04-24 2013-01-01    P

In the case with Non Equi Joins i.InD_dx and i.EndD_dx it's not outputed and EndD_dx becomes i.EndD_dx

Desired outcome

    id   InD_dx    EndD_dx   dx   i.InD_dx  i.EndD_dx i.dx
 1:  a 2005-10-04 2020-12-31  A 2008-04-07 2008-11-05    P
 2:  a 2005-10-06 2020-12-31  B 2008-04-07 2008-11-05    P
 3:  a 2005-10-06 2020-12-31  C 2008-04-07 2008-11-05    P
 4:  a 2005-10-04 2020-12-31  A 2010-05-10 2011-01-17    P
 5:  a 2005-10-06 2020-12-31  B 2010-05-10 2011-01-17    P
 6:  a 2005-10-06 2020-12-31  C 2010-05-10 2011-01-17    P
 7:  a 2005-10-04 2020-12-31  A 2012-04-24 2013-01-01    P
 8:  a 2005-10-06 2020-12-31  B 2012-04-24 2013-01-01    P
 9:  a 2005-10-06 2020-12-31  C 2012-04-24 2013-01-01    P
10:  a 2012-04-24 2015-07-20  D 2012-04-24 2013-01-01    P
11:  a 2012-05-15 2015-07-20  E 2012-04-24 2013-01-01    P

I dunno if I'm complicating things a little bit too much. Probably with a join 1:multiple and making differences between dates I would obtain what I want. This way it's efficient enough?

Any help would be appreciatated

Thanks in advance!

Waldi · Accepted Answer · 2021-03-31T15:51:11.043

type = 'within' is excluding some partial overlaps you're looking for.

Try:

DT = data.table(id,InD_dx,EndD_dx, dx)

setkey(DT,id,InD_dx,EndD_dx)

foverlaps(DT[dx=='P'],DT[dx!='P'],
          by.x = c("id", "InD_dx", "EndD_dx"),
          by.y = c("id", "InD_dx", "EndD_dx"))

   id     InD_dx    EndD_dx dx   i.InD_dx  i.EndD_dx i.dx
 1:  a 2005-10-04 2020-12-31  A 2008-04-07 2008-11-05    P
 2:  a 2005-10-06 2020-12-31  B 2008-04-07 2008-11-05    P
 3:  a 2005-10-06 2020-12-31  C 2008-04-07 2008-11-05    P
 4:  a 2005-10-04 2020-12-31  A 2010-05-10 2011-01-17    P
 5:  a 2005-10-06 2020-12-31  B 2010-05-10 2011-01-17    P
 6:  a 2005-10-06 2020-12-31  C 2010-05-10 2011-01-17    P
 7:  a 2005-10-04 2020-12-31  A 2012-04-24 2013-01-01    P
 8:  a 2005-10-06 2020-12-31  B 2012-04-24 2013-01-01    P
 9:  a 2005-10-06 2020-12-31  C 2012-04-24 2013-01-01    P
10:  a 2012-04-24 2015-07-20  D 2012-04-24 2013-01-01    P
11:  a 2012-05-15 2015-07-20  E 2012-04-24 2013-01-01    P

also note that you can always specify whatever non-equi join you want using x[y] notation. foverlaps is really just a convenience function for x[y] non-equijoins, implemented in the exact style of a specific Bioconductor function — Michael, Apr 02 '21 at 18:14
@Michael, overlaps can be found with x[y], but scenarios #1, #2 & #4 described in OP's question have to be tested, which makes `foverlaps` more practical. — Waldi, Apr 05 '21 at 09:02

Partial or Total Overlap of two Date periods

1 Answers1