Question

I’m looking at ways of creating contingency tables in SQL Server 2008. They don’t necessarily have to appear in the 2×2 typical matrix. I’d just like to know if anyone out there can see a better solution than mine.

Please refer to the picture for clarity. The red letters are names of the squares for simplicity’s sake. The label X+ means X is present in that cell and the opposite is true as well.

I will label my queries with the letter of box in the table that they represent

A

       select count(*) from 
    (

    select distinct p.patientid
        from Patient as p
        inner join icdpatient as picd on picd.patientid = p.patientid
        and picd.admissiondate = p.admissiondate
        and picd.dischargedate = p.dischargedate
        inner join tblicd as t on t.icd_id = picd.icd_id
        where t.icdText like '%x%' 
    ) as t
    inner join 
    (
    select distinct p.patientid
        from Patient as p
        inner join icdpatient as picd on picd.patientid = p.patientid
        and picd.admissiondate = p.admissiondate
        and picd.dischargedate = p.dischargedate
        inner join tblicd as t on t.icd_id = picd.icd_id
        where t.icdText like '%y%'
    ) as s on s.patientid=t.patientid

---

B
select count(*) from 
(

select distinct p.patientid
    from Patient as p
    inner join icdpatient as picd on picd.patientid = p.patientid
    and picd.admissiondate = p.admissiondate
    and picd.dischargedate = p.dischargedate
    inner join tblicd as t on t.icd_id = picd.icd_id
    where t.icdText like '%x%' 
) as t
left join 
(
select distinct p.patientid
    from Patient as p
    inner join icdpatient as picd on picd.patientid = p.patientid
    and picd.admissiondate = p.admissiondate
    and picd.dischargedate = p.dischargedate
    inner join tblicd as t on t.icd_id = picd.icd_id
    where t.icdText like '%y%'
) as s on s.patientid=t.patientid
where s.patientid is null

---

C
select * from 
(

select distinct p.patientid
    from Patient as p
    inner join icdpatient as picd on picd.patientid = p.patientid
    and picd.admissiondate = p.admissiondate
    and picd.dischargedate = p.dischargedate
    inner join tblicd as t on t.icd_id = picd.icd_id
    where t.icdText like '%x%' 
) as t
right join 
(
select distinct p.patientid
    from Patient as p
    inner join icdpatient as picd on picd.patientid = p.patientid
    and picd.admissiondate = p.admissiondate
    and picd.dischargedate = p.dischargedate
    inner join tblicd as t on t.icd_id = picd.icd_id
    where t.icdText like '%y%'
) as s on s.patientid=t.patientid
where t.patientid is null

D
This one I’m a little iffy about but I think I’m going to do something like

declare @d int
set @d = (select count(distinct p.patientid) from Patient as p) - b -c

This aims to find the entire population and subtracting those ONLY with X and ONLY with y

Answer 1

Yes! There are easier ways. Assuming your join produces no duplicate patients:

select (case when t.icdText like '%x%' then 'X' else 'NO-X' end) as X,
       (case when t.icdText like '%y%' then 'Y' else 'NO-Y' end) as Y,
       count(*) as cnt
from Patient p inner join
     icdpatient picd
     on picd.patientid = p.patientid and
        picd.admissiondate = p.admissiondate and
        picd.dischargedate = p.dischargedate inner join
     tblicd t
     on t.icd_id = picd.icd_id
group by  (case when t.icdText like '%x%' then 'X' else 'NO-X' end),
           (case when t.icdText like '%y%' then 'Y' else 'NO-Y' end)

Otherwise replace the count(*) with:

count(distinct patientid)

This should give you the information you need for the contingency table.