GROUP BY affecting WHERE clause - sql

Question

I have a SQL query where I want to see claims where there is no service type associated with any of the lines. My query below is returning the one line where it meets the criteria. However, there is another line with a service type which means I don't want to see this. Because of the group by my query is seeing line and returning it and then looking at line two separately and removing it
select distinct a.claim_id,
count(a.receipt_id) as Count_Receipts,
sum(a.billed_amount) as Total_Billed_Amount,
(case
when a.service_type_id is null then
count(a.receipt_id)
else 0
end) test
from cd_roc_claim_item a,
cd_hospital b ,
cd_Claim_header c
where a.hospital_id = b.hospital_id
and a.claim_id=c.claim_id (+)
and b.hos_cat_id='PUBL'
and c.claim_status is null
and a.claim_id='123456'
group by a.claim_id,
a.service_type_id
having count(a.receipt_id) =1
and sum(a.billed_amount) in
('80','160','240','320','400','480','560','640','720','800')
;

Answer 1

If you just want to report the claims where there are no values for service_type in any claim item:
select ci.claim_id
, count(ci.receipt_id) as count_receipts
, sum(ci.billed_amount) as total_billed_amount
, count(*) as claim_items
from cd_roc_claim_item ci
left join cd_hospital ho
on ho.hospital_id = ci.hospital_id
left join cd_claim_header cl
on cl.claim_id = ci.claim_id
where cl.claim_status is null
group by ci.claim_id
-- , ci.service_type_id -- Edit (following comments): this was part of the problem
having count(ci.service_type_id) = 0
and sum(ci.billed_amount) in (80, 160, 240, 320, 400, 480, 560, 640, 720, 800);
CLAIM_ID COUNT_RECEIPTS TOTAL_BILLED_AMOUNT CLAIM_ITEMS
---------- -------------- ------------------- -----------
1 2 160 2
The problem is the hospital category, because every claim item could be for a different hospital with a different category. Do you want to see only claims where all items are for hospital category PUBL, or where there is at least one, or what?
Test setup:
create table cd_hospital
( hospital_id integer primary key
, hos_cat_id varchar2(10) );
create table cd_claim_header
( claim_id varchar2(10) primary key
, claim_status varchar2(10) );
create table cd_roc_claim_item
( claim_item_id integer generated always as identity primary key
, claim_id references cd_claim_header not null
, hospital_id references cd_hospital not null
, billed_amount number not null
, service_type_id integer
, receipt_id integer );
insert into cd_hospital (hospital_id, hos_cat_id) values (1, 'PUBL');
insert into cd_hospital (hospital_id, hos_cat_id) values (2, 'PRIV');
insert into cd_claim_header(claim_id, claim_status) values (1, null);
insert into cd_claim_header(claim_id, claim_status) values (2, 'CLOSED');
insert into cd_roc_claim_item (claim_id, hospital_id, billed_amount, service_type_id, receipt_id)
select ch.claim_id, ho.hospital_id, 80, null, 1
from cd_claim_header ch
cross join cd_hospital ho
where ch.claim_id = 1
union all
select ch.claim_id, ho.hospital_id, 640, 123, 1
from cd_claim_header ch
cross join cd_hospital ho
where ch.claim_id = 2;
Edit in response to comments:
To include a case where a claim has items with and without a service type, we'll update one row:
update cd_roc_claim_item set service_type_id = null
where claim_id = 2 and hospital_id = 1;
Now the test data looks like this:
select ci.claim_id, ch.claim_status, ci.hospital_id, ho.hos_cat_id, ci.service_type_id, ci.receipt_id
from cd_roc_claim_item ci
join cd_claim_header ch
on ch.claim_id = ci.claim_id
join cd_hospital ho
on ho.hospital_id = ci.hospital_id
order by ch.claim_id;
CLAIM_ID CLAIM_STATUS HOSPITAL_ID HOS_CAT_ID SERVICE_TYPE_ID RECEIPT_ID
---------- ------------ ------------ ---------- ---------------- -----------
1 2 PRIV 1
1 1 PUBL 1
2 CLOSED 2 PRIV 123 1
2 CLOSED 1 PUBL 1
As I understand it, you want to report claim 1 (aggregated to one row) because none of its items have a service type. Claim 2 should not be shown, because one of its items has a service type.