如何检查是否有日期重叠？

Question

Trying to figure out if there is a 60 days (two 30 days supply) overlap for two or more drugs. Following table has a person name, day of fill/re-fill, drugid and days of supply.

Red high-lighted dates are overlapped for 60 days supply for these drugids 25605 and 25700.

How do I construct a SQL to figure out if there is at least a 60 days overlap? Any suggestion would be appreciated. Thank you

Answer 1

This type of calculation is cumbersome in SQL Server because the date functions are clunky and there is no least() or greatest() function.

But, you can do it. It is essentially a self join with an aggregation. The key idea is determining the days of overlap. The beginning and ending of the overlap period is either the dos or dos + days_supply . This is just a complex case statement:

select t.member, t.drug_id, t2.drug_id,
       sum(datediff(day, (case when t.dos <= t2.dos then t2.dos else t.dos end),
                    (case when dateadd(day, t.days_supply, t.dos) <= dateadd(day, t2.days_supply, t2.dos)
                          then dateadd(day, t.days_supply, t.dos)
                          else dateadd(day, t2.days_supply, t2.dos)
                     end)
           ) as days_overlap
from table t join
     table t2
     on t.member = t2.member and t.drug_id < t2.drug_id and
        t.dos between dateadd(day, -t.days_supply, t2.dos) and
                      dateadd(day, t2.days_supply, t2.dos)
group by t.member, t.drug_id, t2.drug_id;

The code might have an off-by-one error -- overlaps are always a bit confusing, but some real data usually clears things up quickly.

Note: This assumes that the drugs do not overlap with themselves. This is an important assumption; otherwise the days of overlaps will be overcounted.

Answer 2

An easy way (and you can build this up by parts), is this:

select  s1.Member, s2.DOS, s1.Drug_ID [First Drug], s2.Drug_ID [Second Drug]
from    Stream s1
join    Stream s2
    on  s2.Member   = s1.Member
    and s2.Drug_ID  > s1.Drug_ID
    and DateDiff( Day, s1.DOS, s2.DOS ) between 0 and 60
group by s1.Member, s2.DOS, s1.Drug_ID, s2.Drug_ID
having Sum( s1.Days_Supply ) >= 60;

The comparison s2.Drug_ID > s1.Drug_ID eliminates the double hit caused by comparing drug A with drug B and again drug B with drug A. The DateDiff limits the comparison to fills in the last 60 days relative to the fill being examined.

This produces a single row from the data stream you provided:

Member DOS        First Drug  Second Drug
------ ---------- ----------- -----------
Joe    06/25/2015 25605       25700

Which means that the drug fill of drug 25700 on 06/25/2015 caused a 60 day overlap with drug 25605. Of course, since both fills occurred on the same day, you could just as well say that it was the fill of drug 25605 which caused the overlap, but as that is a distinction without a real difference, it doesn't matter. Check it against real data to make sure it produces relevant results.

One important assumption made is that there will be no doses left over of a drug when that same drug is filled again (hoarding). Of course, you probably have no way of determining that.