SQL query to only pull most recent value

<knee-jerk reaction>

Which column defines 'most recent'?  If you mean the most recent o2.completedate, then you'll need a MAX.

SELECT obsvalue, p.person_id, p.last_name, p.first_name, p.person_nbr, 
   MAX(o2.completeddate) as most_recent_completeddate
FROM order_ o2
   INNER JOIN ncsuds_lab_test_desc d ON RTRIM(o2.actText) = RTRIM(d.test_desc) and d.order_ind = 'Y' AND d.selected_diag_id = 42
   INNER JOIN person p on o2.person_id = p.person_id
WHERE o2.obsvalue IS NOT NULL
   AND ISNUMERIC(o2.obsvalue) = 1
   AND o2.completeddate > '20150101'
GROUP BY obsvalue, p.person_id, p.last_name, p.first_name, p.person_nbr
ORDER BY completeddate desc

Select all Open in new window

btw note the use of indentation and capitalizing all SELECT .. FROM ... keywords.  Makes it easier to read.   Also I removed the DISTINCT as the GROUP BY handles that.

Check out SQL Server GROUP BY Solutions for some extra info on grouping.

I would try ROW_NUMBER() here:

SELECT * FROM (
  SELECT ROW_NUMBER() OVER (p.person_id, p.person_nbr ORDER BY completeddate desc) rowno, 
         obsvalue, p.person_id, p.last_name, p.first_name, p.person_nbr, o2.completeddate
    FROM order_ o2
   INNER JOIN ncsuds_lab_test_desc d ON RTRIM(o2.actText) = RTRIM(d.test_desc) and d.order_ind = 'Y' AND d.selected_diag_id = 42
   inner join person p on o2.person_id = p.person_id
   WHERE o2.obsvalue IS NOT NULL
     AND ISNUMERIC(o2.obsvalue) = 1
     and o2.completeddate > '20150101') insl
  WHERE rowno = 1 

Select all Open in new window

This is getting closer but still returns duplicates for many of the persons we're returning the values for. In other words the first two rows in the returned data set are the same person with a value for Sep and another for Nov. It is returning both and I'd only want it to return the entry for Nov.

There is a syntax error near the first Order By. Not sure that can be wrapped inside parens like that.

The ROW_NUMBER() requires SQL Server 2005 or newer. The OVER (PARTITION BY ... ORDER BY ...) clause should be OK because it belongs to the ROW_NUMBER().
More info: https://msdn.microsoft.com/en-us/library/ms186734.aspx

If you are using different SQL engine then tell the version please.

BTW, multiplicities on output can also be caused by the INNER JOIN ncsuds_lab_test_desc  definition. To join two tables via text description column is not a standard. What is the reason to join ncsuds_lab_test_desc table?

I don't quite understand what the Row_Number Over Partition does but it seemed to work. Thanks

>This is getting closer but still returns duplicates for many of the persons we're returning the values for.
Please show us some sample data, both your data source and what the above queries are returning.

ROW_NUMBER() generates line numbers for your data and the OVER() defines rules for such numbering, namely grouping (partitioning) and the order of the data.

To get the latest result for each person we used p.person_id for grouping (p.person_nbr is not necessary probably) and ordered the data by  completeddate DESC.

Thanks to the descending order the latest result has ROW_NUMBER() = 1 for each person.

I agree the ROW_NUMBER() is a little bit unclear at the first view.

BTW, this part of your query is still unclear to me:
   INNER JOIN ncsuds_lab_test_desc d ON RTRIM(o2.actText) = RTRIM(d.test_desc) and d.order_ind = 'Y' AND d.selected_diag_id = 42

but that's not important maybe.

Thank you this is helpful. I'm still unclear on the inner join too, but it doesn't seem to filter anything out that I need so I have to trust the vendor on this one.