Aggregation over multiple rows SQL Server 2008

Question

This is an extension of this query here. I have a result set that currently is correct and looks like this:

MD Name | # PTS | Med Staff | AVG LOS | AVG Bench LOS | AVG DIFF
MCP     |    4  | Ortho SX  | 8.000000| 2.650000      | 5.350000

I want to add onto this result, the total patients seen for the department and the same averages but for the departments as a whole. I was able to get the total patients per department to sum up correctly by doing the following as another column added onto the original query above.

(SELECT
    COUNT(DISTINCT V.PT_ID) 
    FROM smsdss.pract_dim_v p
    JOIN smsmir.vst_rpt v
    on v.adm_pract_no = p.src_pract_no
    WHERE pv.med_staff_dept = p.med_staff_dept
    AND v.adm_dtime BETWEEN @STARTDATE AND @ENDATE
    AND v.vst_type_cd = 'I'
    AND p.spclty_desc != 'NO DESCRIPTION'
    --AND pv.spclty_desc NOT LIKE 'HOSPITALIST%'
    AND v.drg_std_days_stay IS NOT NULL
    AND p.pract_rpt_name != '?'
    AND p.orgz_cd = 's0x0'
    AND p.med_staff_dept IN (
    'INTERNAL MEDICINE',
    'FAMILY PRACTICE',
    'SURGERY'
    )
) AS '# PTS For Dept'

I manually checked the sums, they are correct. I know want to add AVG LOS, AVG BENCH LOS and AVG DIFF over p.med_staff_dept. This is where I start getting wack-a-doo (technical term) results.

Here is the part where I try to get the AVG LOS per department:

(SELECT
    AVG(v.len_of_stay)
    FROM smsdss.pract_dim_v p
    JOIN smsmir.vst_rpt v
    ON v.adm_pract_no = p.src_pract_no
    WHERE pv.med_staff_dept = p.med_staff_dept
    AND v.adm_dtime BETWEEN @STARTDATE AND @ENDATE
    AND v.vst_type_cd = 'I'
    AND p.spclty_desc != 'NO DESCRIPTION'
    --AND pv.spclty_desc NOT LIKE 'HOSPITALIST%'
    AND v.drg_std_days_stay IS NOT NULL
    AND p.pract_rpt_name != '?'
    AND p.orgz_cd = 's0x0'
    AND p.med_staff_dept IN (
    'INTERNAL MEDICINE',
    'FAMILY PRACTICE',
    'SURGERY'
    )
) AS 'DEPT ALOS'

Two of my groups get the right numbers when checked against excel, but the middle group of rows is vastly incorrect, like so:

# PTS FOR DEPT | DEPT ALOS
330            | 5.284848 <-- SAME IN EXCEL
737            | 4.952510 <-- EXCEL SHOWS 196 AND 6.214285
111            | 3.072072 <-- SAME IN EXCEL

This is my FROM Clause for the Query in total:

FROM smsmir.vst_rpt vr
JOIN smsdss.pract_dim_v pv
ON vr.adm_pract_no = pv.src_pract_no

WHERE vr.adm_dtime BETWEEN @STARTDATE AND @ENDATE
AND vr.vst_type_cd = 'I'
AND pv.spclty_desc != 'NO DESCRIPTION'
--AND pv.spclty_desc NOT LIKE 'HOSPITALIST%'
AND vr.drg_std_days_stay IS NOT NULL
AND pv.pract_rpt_name != '?'
AND pv.orgz_cd = 's0x0'
AND pv.med_staff_dept IN (
'INTERNAL MEDICINE',
'FAMILY PRACTICE',
'SURGERY'
)
GROUP BY pv.pract_rpt_name, pv.med_staff_dept, pv.spclty_desc
ORDER BY pv.med_staff_dept, AVG(vr.len_of_stay - vr.drg_std_days_stay)DESC

Desired final output:

MD NAME | # PTS | MED STAFF | AVG LOS | AVG BENCH | AVG DIFF | DEPT TOT | AVG LOS | AVG BNCH | AVG DIFF
MCP     |   4   | Ortho SX  | 8.000000| 2.650000  | 5.350000 | 300      | 4.990000| 4.500000 | 0.490000

Full code:

DECLARE @STARTDATE DATETIME
DECLARE @ENDATE DATETIME

SET @STARTDATE = '2013-05-01'
SET @ENDATE = '2013-05-31'

SELECT DISTINCT pv.pract_rpt_name AS 'PHYSICIAN'
, COUNT(DISTINCT vr.pt_id) AS '# PTS'
--, pv.spclty_desc AS 'SPECIALTY'
, pv.med_staff_dept AS 'MED STAFF'
, AVG(vr.len_of_stay) AS 'AVG LOS'
, AVG(vr.drg_std_days_stay) AS 'AVG DRG LOS BENCH'
, AVG(vr.len_of_stay - vr.drg_std_days_stay) AS 'OPPORTUNITY'
, (SELECT
    COUNT(DISTINCT V.PT_ID) 
    FROM smsdss.pract_dim_v p
    JOIN smsmir.vst_rpt v
    on v.adm_pract_no = p.src_pract_no
    WHERE pv.med_staff_dept = p.med_staff_dept
    AND v.adm_dtime BETWEEN @STARTDATE AND @ENDATE
    AND v.vst_type_cd = 'I'
    AND p.spclty_desc != 'NO DESCRIPTION'
    AND pv.spclty_desc NOT LIKE 'HOSPITALIST%'
    AND v.drg_std_days_stay IS NOT NULL
    AND p.pract_rpt_name != '?'
    AND p.orgz_cd = 's0x0'
    AND p.med_staff_dept IN (
    'INTERNAL MEDICINE',
    'FAMILY PRACTICE',
    'SURGERY'
    )
) AS '# PTS For Dept'
-- currently not working properly
, (SELECT
    AVG(V.len_of_stay)
    FROM smsmir.vst_rpt v
    JOIN smsdss.pract_dim_v p
    ON v.adm_pract_no = p.src_pract_no
    WHERE pv.med_staff_dept = p.med_staff_dept
    AND v.adm_dtime BETWEEN @STARTDATE AND @ENDATE
    AND v.vst_type_cd = 'I'
    AND p.spclty_desc != 'NO DESCRIPTION'
    AND pv.spclty_desc NOT LIKE 'HOSPITALIST%'
    AND v.drg_std_days_stay IS NOT NULL
    AND p.pract_rpt_name != '?'
    AND p.orgz_cd = 's0x0'
    AND p.med_staff_dept IN (
    'INTERNAL MEDICINE',
    'FAMILY PRACTICE',
    'SURGERY'
    )
) AS 'DEPT ALOS'

FROM smsmir.vst_rpt vr
JOIN smsdss.pract_dim_v pv
ON vr.adm_pract_no = pv.src_pract_no

WHERE vr.adm_dtime BETWEEN @STARTDATE AND @ENDATE
AND vr.vst_type_cd = 'I'
AND pv.spclty_desc != 'NO DESCRIPTION'
AND pv.spclty_desc NOT LIKE 'HOSPITALIST%'
AND vr.drg_std_days_stay IS NOT NULL
AND pv.pract_rpt_name != '?'
AND pv.orgz_cd = 's0x0'
AND pv.med_staff_dept IN (
'INTERNAL MEDICINE',
'FAMILY PRACTICE',
'SURGERY'
)
GROUP BY pv.pract_rpt_name, pv.med_staff_dept, pv.spclty_desc
ORDER BY pv.med_staff_dept, AVG(vr.len_of_stay - vr.drg_std_days_stay)DESC

None of the Internal Medicine numbers are coming out correctly.

Answer 1

The problem I found is right here:

, (SELECT
    COUNT(DISTINCT V.PT_ID) 
    FROM smsdss.pract_dim_v p
    JOIN smsmir.vst_rpt v
    on v.adm_pract_no = p.src_pract_no
    WHERE pv.med_staff_dept = p.med_staff_dept
    AND v.adm_dtime BETWEEN @STARTDATE AND @ENDATE
    AND v.vst_type_cd = 'I'
    AND p.spclty_desc != 'NO DESCRIPTION'
    AND pv.spclty_desc NOT LIKE 'HOSPITALIST%' <-- should be p.spclty not pv.
    AND v.drg_std_days_stay IS NOT NULL
    AND p.pract_rpt_name != '?'
    AND p.orgz_cd = 's0x0'
    AND p.med_staff_dept IN (
    'INTERNAL MEDICINE',
    'FAMILY PRACTICE',
    'SURGERY'
    )
) AS '# PTS For Dept'

By changing the pv.spclty_desc to p.spclty_desc in both column select statements the problem was fixed. However, this does not tell me why in the statement only Internal Medicinie failed and not the other two.