SQL: Putting an individuals distinct diagnosis into one horizontal row - sql

Question

I'm using Microsoft SQL Server 2008 for a mental health organization.
I have a table that lists all of out clients and their diagnoses, but each diagnoses that a client has is in a new row. I want them all to be in a single row listed out horizontally with the date for each diagnosis. Some people have just one diagnosis, some have 20, some have none.
Here's an example of how my data sort of looks now (only with a lot few clients, we have thousands):
And Here's the format I'd like it to end up:
Any solutions you could offer or hints in the right direction would be great, thanks!

Answer 1

In order to get the result, I would first unpivot and then pivot your data. The unpivot will take your date and diagnosis columns and convert them into rows. Once the data is in rows, then you can apply the pivot.
If you have a known number of values, then you can hard-code your query similar to this:
select *
from
(
select person, [case#], age,
col+'_'+cast(rn as varchar(10)) col,
value
from
(
select person,
[case#],
age,
diagnosis,
convert(varchar(10), diagnosisdate, 101) diagnosisDate,
row_number() over(partition by person, [case#]
order by DiagnosisDate) rn
from yourtable
) d
cross apply
(
values ('diagnosis', diagnosis), ('diagnosisDate', diagnosisDate)
) c (col, value)
) t
pivot
(
max(value)
for col in (diagnosis_1, diagnosisDate_1,
diagnosis_2, diagnosisDate_2,
diagnosis_3, diagnosisDate_3,
diagnosis_4, diagnosisDate_4)
) piv;
See SQL Fiddle with Demo.
I am going to assume that you will have an unknown number of diagnosis values for each case. If that is the case, then you will need to use dynamic sql to generate the result:
DECLARE #cols AS NVARCHAR(MAX),
#query AS NVARCHAR(MAX)
select #cols = STUFF((SELECT ',' + QUOTENAME(col+'_'+cast(rn as varchar(10)))
from
(
select row_number() over(partition by person, [case#]
order by DiagnosisDate) rn
from yourtable
) t
cross join
(
select 'Diagnosis' col union all
select 'DiagnosisDate'
) c
group by col, rn
order by rn, col
FOR XML PATH(''), TYPE
).value('.', 'NVARCHAR(MAX)')
,1,1,'')
set #query = 'SELECT person,
[case#],
age,' + #cols + '
from
(
select person, [case#], age,
col+''_''+cast(rn as varchar(10)) col,
value
from
(
select person,
[case#],
age,
diagnosis,
convert(varchar(10), diagnosisdate, 101) diagnosisDate,
row_number() over(partition by person, [case#]
order by DiagnosisDate) rn
from yourtable
) d
cross apply
(
values (''diagnosis'', diagnosis), (''diagnosisDate'', diagnosisDate)
) c (col, value)
) t
pivot
(
max(value)
for col in (' + #cols + ')
) p '
execute(#query);
See SQL Fiddle with Demo. Both queries give the result:
| PERSON | CASE# | AGE | DIAGNOSIS_1 | DIAGNOSISDATE_1 | DIAGNOSIS_2 | DIAGNOSISDATE_2 | DIAGNOSIS_3 | DIAGNOSISDATE_3 | DIAGNOSIS_4 | DIAGNOSISDATE_4 |
------------------------------------------------------------------------------------------------------------------------------------------------------------------------
| John | 13784 | 56 | Depression | 03/13/2012 | Brain Injury | 03/14/2012 | Spinal Cord Injury | 03/15/2012 | Hypertension | 03/16/2012 |
| Kate | 2643 | 37 | Bipolar | 03/11/2012 | Hypertension | 03/12/2012 | (null) | (null) | (null) | (null) |
| Kevin | 500934 | 25 | Down Syndrome | 03/18/2012 | Clinical Obesity | 03/19/2012 | (null) | (null) | (null) | (null) |
| Pete | 803342 | 34 | Schizophenia | 03/17/2012 | (null) | (null) | (null) | (null) | (null) | (null) |

Answer 2

For this type of pivoting, I think the aggregate/group method is feasible:
select d.case, d.person,
max(case when seqnum = 1 then diagnosis end) as d1,
max(case when seqnum = 1 then diagnosisdate end) as d1date,
max(case when seqnum = 2 then diagnosis end) as d2,
max(case when seqnum = 2 then diagnosisdate end) as d2date,
. . . -- and so on, for as many groups that you want
from (select d.*, row_number() over (partition by case order by diagnosisdate) as seqnum
from diagnoses d
) d
group by d.case, d.person

Answer 3

Since you are dealing with sensitive medical information, identifyiable information (name age etc) shouldn't be stored in the same table as the medical information. Also, if you extract out the person info into its own table and a Diagnosis table that has the personID foreign key you can establish the 1 to many relationship you want.

Answer 4

Unless you use Dynamic SQL, the PIVOT operator will not work here. I assume that patients can come in on any date. The PIVOT operator works with a finite and predefined number of columns. Your options are to use Dynamic SQL to create the PIVOT table, or to use Excel or a reporting tool like SSRS to do a Pivot report.
I think the Dynamic SQL option would not be practical here, since, you could end up having hundreds of columns for each of the patient visit dates.
If you want to explore the Dynamic SQL option anyway, have a look here:
https://www.simple-talk.com/blogs/2007/09/14/pivots-with-dynamic-columns-in-sql-server-2005/