Question

I need a difference between two values based on their row number. The CTE returns data like

What I want to do is have row2 read 60 in the rehabWait column (92-32) and row three to be 60 as well (152-92`) etc. until there is a change in patientid. So for row 11 I’d like for rehabwait to be 110 (114-4). The query I have will run but I it returns all NULLS

with x as 
    (
     SELECT row_number() over (order by patientid, admissiondate, claimsfromdate, datediff(dd,admissiondate, claimsfromdate))as rn,
     patientid, admissiondate, claimsfromdate,
            DATEDIFF(dd, admissiondate, claimsfromdate) as rehabWait, hcpcs
     FROM    Claims
     WHERE   hcpcs in ('g0151', '97001', '97002', '9339') and
             claimsfromdate > admissiondate 
      group by patientid, admissiondate, claimsfromdate, hcpcs
      --adding this group by clause will keep rehabWait from showing up
      --however many times they patient has the HCPCS code for rehab
    )


select  x.patientid
        ,x.admissiondate
        ,x.claimsfromdate

        ,(select x2.rehabWait-x.rehabwait from x where x.patientid=x2.patientid 
        and x.rn > x2.rn and x.admissiondate=x2.admissiondate and x.claimsfromdate=x2.claimsfromdate
        )
        from x inner join
        x as x2 on x.patientid=x2.patientid and x.admissiondate=x2.admissiondate and x.claimsfromdate = x2.claimsfromdate

Answer 1

with x as 
(
 SELECT row_number() over (PARTITION BY patientid order by patientid, admissiondate, claimsfromdate, datediff(dd,admissiondate, claimsfromdate))as rn,
 patientid, admissiondate, claimsfromdate,
        DATEDIFF(dd, admissiondate, claimsfromdate) as rehabWait, hcpcs
 FROM    #Claims
 WHERE   hcpcs in ('g0151', '97001', '97002', '9339') and
         claimsfromdate > admissiondate 
  group by patientid, admissiondate, claimsfromdate, hcpcs
  --adding this group by clause will keep rehabWait from showing up
  --however many times they patient has the HCPCS code for rehab
)
select  x.patientid
    ,x.admissiondate
    ,x.claimsfromdate
    , CASE WHEN x2.rn = 1 
           THEN x.rehabWait 
           ELSE  x2.rehabWait - x.rehabWait END AS rehabWait
    from x INNER join
    x as x2 on x.patientid=x2.patientid AND CASE WHEN x2.rn = 1 AND x.rn = 1 THEN x2.rn - 1 ELSE x.rn END = x2.rn - 1 

The two CASE statements in the select are making sure that you get the first rows. The PARTITION BY is making sure that every patient Id starts at rn = 1 and ascends.

EDIT: The first query I gave you was going to lose rows 1 and 10 in your example.