Splitting Data from a Column into two Columns - sql-server

Question

I have data in a column called "Medication_Description". First I need to
then find out who that received Tylenol also received a second drug (Ibuprofen here).
I also only want the top two results for each MRN (i.e. patient). I only want data for the past year.
Later I will be plugging this into an SSRS report where I will determine the percentage of patients that received both drugs.
I've played with a couple different ways of getting this to work but can't get it working quite right.
The data in this table looks like this:
As for desired results, I'd like to have something like this:
Blockquote
For MRN 654321, no Ibuprofen was administered so it returns NULL (it could also return another drug name - doesn't matter too much. I just need to be able to count the results later to determine a percentage).
For MRN 246824, only one dose of Ibuprofen was administered so the second line is NULL.
Below is my latest attempt but (as you can see) Med1 and Med2 will always reflect the same exact data - how can I make Med1 reflect one medication and Med2 reflect a second?
SELECT [MRN], [Med1], [Med2], [Row_Num], [Department_Name], [Date]
FROM
( SELECT [MRN], [Medication_Description] AS [Med1], [Medication_Description] AS [Med2],
ROW_NUMBER() OVER (PARTITION BY [MRN]
ORDER BY [Medication_Description] DESC
)
AS [Row_Num],
[Date],
[Department_Name]
FROM T_Med_Orders
WHERE [DATE] BETWEEN dateadd(year,-1,getdate()) AND getdate()
AND [Department_Name] LIKE 'ICU'
--First Med Must Match "Tylenol" but 2nd should match any result...???
AND Medication_Description LIKE '%Tylenol%'
)
tmp
WHERE
[Row_Num] <= 2
AND Med1 LIKE '%Tylenol%'
--AND Med2 LIKE '%Ibuprofen%'
AND [DATE] BETWEEN dateadd(year,-1,getdate()) AND getdate()
ORDER BY [MRN]

Answer 1

You have a solid start, but I think you're being too ambitious with your query. While there are ways to optimize the query, using Partitions may be overkill for the current requirements. What I ended up doing was to make two CTEs where each is filtered to the individual medicine being identified. I can LEFT JOIN those to the source table and filter it to only show results where the CTE values are NOT NULL. I can also apply the other DeptName and Date clauses, although I have not done so in my code snippet.
The main drawback is that this format would require more and more CTEs if you wanted to expand to include other medicines, swiftly reducing optimization further. But without knowing how MedicationDescription is formatted (or if it even has a standard format) I can't write that for you.
WITH Tylenol_CTE AS
(SELECT *, 'Tylenol' AS [FilteredMedicine]
FROM #Temp
WHERE Medication_Description LIKE '%Tylenol%')
,Ibuprofen_CTE AS
(SELECT *, 'Ibuprofen' AS [FilteredMedicine]
FROM #Temp
WHERE Medication_Description LIKE '%Ibuprofen%')
SELECT t.*
, Tylenol_CTE.[FilteredMedicine] AS Med1
, Ibuprofen_CTE.[FilteredMedicine] AS Med2
FROM #Temp t
LEFT JOIN Tylenol_CTE
ON t.MRN = Tylenol_CTE.MRN
AND t.Date = Tylenol_CTE.Date
LEFT JOIN Ibuprofen_CTE
ON t.MRN = Ibuprofen_CTE.MRN
AND t.Date = Ibuprofen_CTE.Date
WHERE Ibuprofen_CTE.Medication_Description IS NOT NULL
AND Tylenol_CTE.Medication_Description IS NOT NULL