Can this query be solved using something besides 2 CTEs? - tsql

Question

I’m writing a query against a table of fictional insurance clams called CLAIMS, using RANDOMLY GENERATED FICTIONAL NAMES AND DATA.
There are 5 distinct categories in the column called PRIMARY_DX:
Alcoholism, Anxiety Disorder, Depression, Psychosis, Substance Use Disorder
The other main columns are PATIENT_ID and CLAIM_PAID_AMT
I want to sum up the CLAIM_PAID_AMT per PATIENT per PRIMARY_DX and list only the top 5 patients who have the highest sum per PRIMARY_DX
The only way I could think to do this was with two Common Table Expressions, where in CTE1 I partition by PRIMARY_DX and PATIENT_ID and SUM the CLAIM_PAID_AMT for each PATIENT.
Then in CTE2 I use a ROW_NUMBER function on CTE1, to partition by PRIMARY_DX and sort by the TotalClaims DESC and select the top 5 from each PRIMARY_DX.
I’ve been writing SQL for less than 2 years and was wondering if this could be accomplished in one CTE or perhaps with some form of Cross Apply?
I’m including my code and the output below.
;WITH CTE1 AS
(
select PRIMARY_DX, PATIENT_ID, TotalClaims = SUM(CLAIM_PAID_AMT)
OVER (PARTITION BY PRIMARY_DX, PATIENT_ID ORDER BY PATIENT_ID, CLAIM_PAID_AMT DESC)
from claims
)
,
CTE2 AS
(SELECT *, RowCounter = ROW_NUMBER() OVER (PARTITION BY PRIMARY_DX ORDER BY TotalClaims DESC) FROM CTE1)
select CTE2.PRIMARY_DX, CTE2.TotalClaims from CTE2
where RowCounter <= 5
order by CTE2.PRIMARY_DX, CTE2.TotalClaims DESC
Alcoholism 3737.51 Joe Smith
Alcoholism 3282.07 Suzie Homemaker
Alcoholism 3207.72 Joey Strummer
Alcoholism 3040.52 Rusty Nailfile
Alcoholism 2997.02 Big Ben
Anxiety Disorder 3291.14 Norman Pigsty
Anxiety Disorder 3113.05 Billy Bob
Anxiety Disorder 3101.13 Rachel Antarctica
Anxiety Disorder 3058.52 John John
Anxiety Disorder 3021.98 Kathy Europa
Depression 3466.14 Freda Beagallly
Depression 3279.25 Ron Jeremize
Depression 3140.43 Sharon Sharonaz
Depression 3119.26 Allie Kat
Depression 3118.54 Biff Biffstoferson
Psychosis 3098.13 James Monopoly
Psychosis 2991.23 Leon Erroneously
Psychosis 2857.69 Lucie Ratched-McMurphy
Psychosis 2678.88 Billy Bibbitz
Psychosis 2602.24 Sam Zypperzsky
Substance Use Disorder 3435.27 Donald Duckaronawitz
Substance Use Disorder 3300.33 Mickey Mousetrap
Substance Use Disorder 3285.41 Hector Heathercoatz
Substance Use Disorder 3179 Erin GoBragh
Substance Use Disorder 3147.09 Bono Edgerstein

Answer 1

You should only need one sub-query or CTE since you can use the aggregate within the ROW_NUMBER().
Here is an approach using the sub-query:
SELECT *
FROM (
SELECT PRIMARY_DX, PATIENT, SUM(CLAIM_PAID_AMT) AS CLAIM_PAID_AMT,
ROW_NUMBER() OVER (PARTITION BY PRIMARY_DX ORDER BY SUM(CLAIM_PAID_AMT) DESC) AS RowId
FROM Claims GROUP BY PRIMARY_DX, PATIENT
) T
WHERE RowId <= 5
And if you prefer CTE:
;WITH CTE AS (
SELECT PRIMARY_DX, PATIENT, SUM(CLAIM_PAID_AMT) AS CLAIM_PAID_AMT,
ROW_NUMBER() OVER (PARTITION BY PRIMARY_DX ORDER BY SUM(CLAIM_PAID_AMT) DESC) AS RowId
FROM Claims GROUP BY PRIMARY_DX, PATIENT
) SELECT * FROM CTE WHERE RowId <= 5