Question

我正在处理代表人群索赔数据的数据提取。最终目标是能够将特定的主张与独特的遭遇相关联。我们定义遭遇的方式是一连串连续的或在相同范围内的声明。意思是说，迄今为止的索赔要求后面紧跟着一个索赔要求的起始日期，或者与另一项索赔要求的＆起始日期相同。但是，由于没有键可以真正表明一组声明代表特定的遭遇，因此我们必须实现为我们执行此操作的逻辑。

在此示例中，我们已经按“患者与设施”对数据进行了分组。

CREATE TABLE #BillingData (
ClaimID VARCHAR(MAX)
,Patient VARCHAR(MAX)
,Facility VARCHAR(MAX)
,ClaimFromDate DATE
,ClaimToDate DATE
,GroupID VARCHAR(MAX)
)

INSERT INTO #BillingData
VALUES
('Claim1','JOHN DOE','NURSING HOME','2000-01-01','2000-01-01','1')
,('Claim2','JOHN DOE','NURSING HOME','2000-01-01','2000-01-31','1')
,('Claim3','JOHN DOE','NURSING HOME','2000-01-01','2000-01-31','1')
,('Claim4','JOHN DOE','NURSING HOME','2000-01-01','2000-01-31','1')
,('Claim5','JOHN DOE','NURSING HOME','2000-02-01','2000-02-29','1')
,('Claim6','JOHN DOE','NURSING HOME','2000-03-01','2000-03-31','1')
,('Claim7','JOHN DOE','NURSING HOME','2000-04-01','2000-04-30','1')
,('Claim8','JOHN DOE','NURSING HOME','2000-08-01','2000-08-31','1')
,('Claim9','JOHN DOE','NURSING HOME','2000-09-01','2000-09-30','1')
,('Claim10','JANE DOE','HOME HEALTH','2000-02-01','2000-02-29','2')
,('Claim11','JANE DOE','HOME HEALTH','2000-03-01','2000-03-31','2')
,('Claim11','JANE DOE','HOME HEALTH','2000-03-01','2000-03-31','2')

我想最终看到的是一个遭遇ID。这是将各个声明放入遇到的ID中的方法：

第1,2,3,4,5,6,7项针对Encounter 1.1；索赔8,9在遭遇1.2中；索赔10,11在遭遇2.1中

任何指导将不胜感激！

Answer 1

您需要找到差距，即当前的开始日期大于任何以前的结束日期（加上一天）

with gaps as
 (
   select *,
      case
        when max(ClaimToDate) -- max previous end date
             over (partition by Patient, Facility
                   order by ClaimFromDate, ClaimID
                   rows between unbounded preceding and 1 preceding)
            < dateadd(day,-1,ClaimFromDate) -- compare to current start date 
        then 1 -- gap
        else 0 -- overlapping ranges
      end as flag
   from #BillingData
 )
select *,
   sum(flag) -- cumulative sum over 0/1 to create encounter number 
   over (partition by Patient, Facility
         order by ClaimFromDate, ClaimID
         rows unbounded preceding) +1 as encounter
from gaps
order by GroupID, encounter, ClaimFromDate

请参见fiddle

SQL连续日期范围以前分组的数据

1 个答案: