MAX or MIN on SQL table value with condition - sql

Question

Hi I was hoping someone could tell me a better way to do this, what I have works, but it is slow and is causing my query to run over 4 seconds. The tables all have indexes on and I can't see anything in the execution plans in particular.
I want to get the min assessmentId where initial = 1, and also the max assessmentId altogether from my table.
Is there a better way that joining the table twice?
select
r.PatientId,
MAX(r.ReferralId),
MIN(a.AssessmentOneId),
MAX(a1.AssessmentOneId)
from dbo.Referral r
inner join dbo.Patient p on p.PatientId = r.PatientId
left join dbo.AssessmentOne a on a.ReferralId = r.ReferralId and a.Initial = 1
left join dbo.AssessmentOne a1 on a1.ReferralId = r.ReferralId
where
p.AccountId = #pAccountId
group by r.PatientId
I've also tried the following using sub queries, but I am still getting bad performance.
select
r.PatientId,
MAX(r.ReferralId),
(Select MIN(a.AssessmentOneId) from dbo.AssessmentOne a where a.ReferralId = MAX(r.ReferralId) and a.Initial = 1),
(Select MAX(a.AssessmentOneId) from dbo.AssessmentOne a where a.ReferralId = MAX(r.ReferralId))
from dbo.Referral r
inner join dbo.Patient p on p.PatientId = r.PatientId
where
p.AccountId = #pAccountId
group by r.PatientId
Any help you can give would be appreciated

Answer 1

For this query:
select r.PatientId, MAX(r.ReferralId), MIN(a.AssessmentOneId),
MAX(a1.AssessmentOneId)
from dbo.Referral r inner join
dbo.Patient p
on p.PatientId = r.PatientId left join
dbo.AssessmentOne a
on a.ReferralId = r.ReferralId left join
dbo.AssessmentOne a1
on a1.ReferralId = r.ReferralId
where p.AccountId = #pAccountId
group by r.PatientId;
I would recommend indexes on Patient(AccountId, PatientId), Referral(PatientId, ReferralId), and AssessmentOne(ReferalId, Initial).
The double join is a bit strange. So, I would write this as:
select p.PatientId, max(r.ReferralId), max(a.AssessmentOneId),
max(case when a.Initial = 1 then a.AssessmentOneId end)
from dbo.Referral r inner join
dbo.Patient p
on p.PatientId = r.PatientId left join
dbo.AssessmentOne a
on a.ReferralId = r.ReferralId
where p.AccountId = #pAccountId
group by p.PatientId;
I doubt this really affects performance, but it seems simpler to me.

Answer 2

The referral table is a bridge table for the m:n relation of patient and assessment. As you want certain assessment IDs per patient, it's necessary to join the tables, what you are already doing. However, as we must look up all assessments per patient to get the maximum number, we can get the minimum initial assessment ID on-the-fly:
select
p.patientid,
max(r.referralid),
min(case when initial = 1 then assessmentoneid end),
max(assessmentoneid)
from dbo.patient p
join dbo.referral r on on r.patientid = p.patientid
left join dbo.assessmentone a on a.referralid = r.referralid
where p.accountid = #paccountid
group by p.patientid;
You'll want these indexes:
patient (accountid, patientid)
referral (patientid, referralid)
assessmentone (referralid, assessmentoneid, initial)