Get total number of records for groups of data in a table

Question

I'm using a SQL Server 2008R2 database. I have a query that I use to get some lab-result data from our database for patients who have had more than 2 visits to our office. Here is the query:

;WITH PatientData (patient_id,last_name, first_name, ethnic_group,
race, icdcode, encounter_date, test_performed, result_value,
result_units )
AS
(
SELECT pat.patient_id, pat.last_name, Pat.first_name, pat.ethnic_group,
pat.race, pl.icdcode, enc.encounter_date, lab.test_performed, lab.result_value,
lab.result_units
FROM dbo.dem_patient pat
RIGHT JOIN dbo.med_problemlist as pl
ON pat.patient_id = pl.patient_id
JOIN dbo.enc_encounter as enc
ON pat.patient_id = enc.patient_id
JOIN dbo.med_labresult as lab
ON pat.patient_id = lab.patient_id
WHERE pl.icdcode like '250%' AND
enc.encounter_date >= '01/01/2012' AND enc.encounter_date <= '12/31/2012' AND
pat.last_name != 'test' AND (lab.test_performed = 'HgbA1c' OR 
lab.test_performed = 'Hemoglobin A1c')
GROUP BY pat.patient_id, pat.last_name, Pat.first_name, pat.ethnic_group,
pat.race, pl.icdcode, enc.encounter_date, lab.test_performed, lab.result_value,
lab.result_units
)
SELECT patient_id,last_name, first_name, ethnic_group,
race, icdcode, encounter_date, test_performed, result_value,
result_units
FROM PatientData pd1 
WHERE EXISTS (SELECT patient_id 
                FROM PatientData pd2 
                WHERE pd2.patient_id  = pd1.patient_id  
                GROUP BY patient_id 
                HAVING COUNT(*)>1 )
ORDER BY pd1.race, pd1.ethnic_group

This works fine for getting a list of patients and their lab result values. It returns something like this:

patient_id    encounter_date    test_performed    result_value    result_units
00001         01/15/2012        HgbA1c            5.6             %
00001         05/03/2012        HgbA1c            8.6             %
00025         02/02/2012        HgbA1c            9.1             %
00064         07/01/2012        HgbA1c            7.6             %

However, I'd now like to group this data, and get a total number of results between different values of results. For example, I'd like to have results that look something like this:

LessThan7%    7%To8%     8%To9%    GreaterThan9%
775           289        365       154

I've been able to do something similar in the past utilizing the CROSS JOIN function. However, never with a query this complicated. I could really use some help.

Thank you!

UPDATE:

On the advice below, I've updated the query to the following:

;WITH PatientData (patient_id,last_name, first_name, ethnic_group,
race, icdcode, encounter_date, test_performed, result_value,
result_units )
AS
(
SELECT pat.patient_id, pat.last_name, Pat.first_name, pat.ethnic_group,
pat.race, pl.icdcode, enc.encounter_date, lab.test_performed, lab.result_value,
lab.result_units
FROM dbo.dem_patient pat
RIGHT JOIN dbo.med_problemlist as pl
ON pat.patient_id = pl.patient_id
JOIN dbo.enc_encounter as enc
ON pat.patient_id = enc.patient_id
JOIN dbo.med_labresult as lab
ON pat.patient_id = lab.patient_id
WHERE pl.icdcode like '250%' AND
enc.encounter_date >= '01/01/2012' AND enc.encounter_date <= '12/31/2012' AND
pat.last_name != 'test' AND (lab.test_performed = 'HgbA1c' OR 
lab.test_performed = 'Hemoglobin A1c')
GROUP BY pat.patient_id, pat.last_name, Pat.first_name, pat.ethnic_group,
pat.race, pl.icdcode, enc.encounter_date, lab.test_performed, lab.result_value,
lab.result_units
)
select sum(case when isnumeric(result_value) = 1
       then (case when result_value < 7.0 then 1 else 0 end) 
       end) as [LessThan7%],
       sum(case when isnumeric(result_value) = 1
       then (case when result_value >= 7.0 and result_value < 8.0 then 1 else 0 end) 
       end) as [7%to8%],
       sum(case when isnumeric(result_value) = 1
       then (case when result_value >= 8.0 and result_value < 9.0 then 1 else 0 end) 
       end) as [8%to9%],
       sum(case when isnumeric(result_value) = 1
       then (case when result_value >= 9.0 then 1 else 0 end)
       end) as [GreaterThan9%]
FROM PatientData pd1 
WHERE EXISTS (SELECT patient_id 
                FROM PatientData pd2 
                WHERE pd2.patient_id  = pd1.patient_id  
                GROUP BY patient_id 
                HAVING COUNT(*)>1 )

However, I'm still getting the "Arithmic overflow error converting varchar to data type numeric"

Any additional help is greatly appreciated.

Answer 1

This is a conditional aggregation:

with query as (your query here)
select sum(case when result_value < 0.07 then 1 else 0 end) as [LessThan7%],
       sum(case when result_value >= 0.07 and result_value < 0.08 then 1 else 0 end) as [7%to8%],
       sum(case when result_value >= 0.08 and result_value < 0.09 then 1 else 0 end) as [8%to9%],
       sum(case when result_value >= 0.09 then 1 else 0 end) as [GreaterThan9%]
from query

Arrange your query to go in the with clause.

If result_value is stored as a character, then try this:

select sum(case when isnumeric(result_value) = 1
                then (case when result_value < 0.07 then 1 else 0 end)
            end) as [LessThan7%],
       sum(case when isnumeric(result_value) = 1
                then result_value >= 0.07 and result_value < 0.08 then 1 else 0 end)
           end) as [7%to8%],
       sum(case when isnumeric(result_value) = 1
                then (case when result_value >= 0.08 and result_value < 0.09 then 1 else 0 end)
            end) as [8%to9%],
       sum(case when isnumeric(result_value) = 1
                then (case when result_value >= 0.09 then 1 else 0 end)
           end) as [GreaterThan9%]

You need the nested case statements to ensure that isnumeric() runs before the attempted conversion.