Extract numeric entries from column B that match unique (first occurrence of) alphanumeric items in column A

[StructuredWater]

Dear Reader,

[Col A] [Col -] [Col -] [Col B]
URN DIEDLATER? Surg? LOS
ABC40643 FALSE TRUE 176
ABC40643 FALSE TRUE 176
ABC50503 FALSE TRUE 123
ABC64663 FALSE TRUE 62
ABC64663 FALSE TRUE 62
ABC52133 FALSE TRUE 124
ABC52133 FALSE TRUE 124
ABC55433 FALSE TRUE 132
ABC55433 FALSE TRUE 132
ABC34303 FALSE TRUE 78
ABC83573 FALSE TRUE 85
ABC83573 FALSE TRUE 85

The data shows length of stay (LOS) in surgical survivors (Surg? = TRUE, DIEDLATER? = FALSE) who had selected diagnoses. I want to generate quartiles from the LOS column for unique instances of the URN. Thus, for the data above I want to generate quartiles (1, 2 and 3) for the unique items 176, 123, 62, 124, 132, 78, 85. If two different URNs happen to have the same LOS, then the LOS is counted twice - it is the uniqueness of the URNs and not of the LOS's that determines if an LOS ends up (possibly more than once) in the final list to be passed to the QUARTILE function.
At the moment I am somewhat despondent. I can't figure out how to pull out the LOS's to pass to the QUARTILE function, and could do with some help.
Kind regards,
JohnM

 

[StructuredWater]

Further information: Excel 2003. The data shown above shows the filtered rows from my list meeting my criteria for patients who survived and had surgery. I was wanting to summarise the data just for those rows shown.

My list sheet has named ranges for the URN column (URNRange), the DIEDLATER? column (DeathsRange), the surgery column (SurgRange) and the length of stay column (LOSRange).

Thanks to Aladin Akyurek for some useful posts which got me started. My solution was:
=QUARTILE(IF((FREQUENCY(MATCH(INDIRECT(URNRange),INDIRECT(URNRange),0),MATCH(INDIRECT(URNRange),INDIRECT(URNRange),0))>0)*(OFFSET(INDIRECT(DeathsRange),0,0,ROWS(INDIRECT(DeathsRange))+1,1)=FALSE)*(OFFSET(INDIRECT(SurgRange),0,0,ROWS(INDIRECT(SurgRange))+1,1)=TRUE),OFFSET(INDIRECT(LOSRange),0,0,ROWS(INDIRECT(LOSRange))+1,1)),1)

This is probably a kludge but I had to work around the extra array element returned by FREQUENCY. I did this by increasing the size of the named ranges by one row, knowing that the extra 'empty' row of data would not return an LOS as it did not meet the selection criteria (eg being empty, it had no Surgery item so the surgery condition could not be true).

Is there an alternative? Could I have somehow resized the array returned by FREQUENCY to drop the unwanted final element so I would not need to resize the ranges?