在第一个最高分之后找到最低分

Question

我正在寻找一个 SQL 查询。

表 1：源表

表 2：结果表

附上病人桌。在这里，我想找到每个患者的最高分、最高分日期、最低分和最低分日期。棘手的部分是，如果患者在两个不同的日期（10/5/2018 和 8/4/2020）具有相同的最高分（这里是 9），我们需要取最早的高分（10/5/2018） .同样，如果患者在两个不同日期（2019 年 3 月 1 日和 2020 年 4 月 2 日）的最低分相同（此处为 6），我们应该取最新的低分（2020 年 4 月 2 日）

表 1：源表包含单个患者的所有分数。患者 ID 是该表的主键。我想要一个看起来像表 2 的结果表。

我试过了

SELECT distinct
    pat.PAT_NAME 'Patient Name'
    , pat.PAT_ID
    , CAST(pat.CONTACT_DATE AS DATE) 'Service Date'
    , pat.MEAS_VALUE 'Score'
    , [Row Number] = rank() OVER (PARTITION BY pat.PAT_ID ORDER BY CAST(pat.MEAS_VALUE AS int) DESC, CONTACT_DATE asc)
FROM Patient pat            
WHERE pat.PAT_ID = 'A112233'

但是这段代码可以显示我最高或最低的分数。但这并不能满足我的所有要求。

Answer 1

如果表 t 包含以下列：PatientName、PatientID、ServiceDate、Score。像这样的：

;with high_low_cte(PatientID, high_rn, low_rn) as(
    select
      PatientID,
      row_number() over (partition by PatientID order by Score, ServiceDate asc),
      row_number() over (partition by PatientID order by Score, ServiceDate desc)
    from
      t)
select * from high_low_cte where high_rn=1 and low_rn=1;

更新问题后：

;with high_low_cte([Patient Name], PAT_ID, [Service Date], Score, high_rn, low_rn) as (
    SELECT distinct
      pat.PAT_NAME 'Patient Name'
      ,pat.PAT_ID
      ,CAST(pat.CONTACT_DATE AS DATE) 'Service Date'
      ,pat.MEAS_VALUE 'Score'
      ,high_rn=row_number() OVER (PARTITION BY pat.PAT_ID ORDER BY CAST(pat.MEAS_VALUE AS int) DESC, CONTACT_DATE asc) 
      ,low_rn=row_number() OVER (PARTITION BY pat.PAT_ID ORDER BY CAST(pat.MEAS_VALUE AS int) asc, CONTACT_DATE asc) 
    FROM  
      Patient pat                                 
    WHERE                  
    pat.PAT_ID='A112233')
select hld1.*, hld2.Score [Low_Score], hld2.[Service Date] [Low Service Date]
from high_low_cte hld1 join high_low_cte hld2 on hld1.PAT_ID=hld2.PAT_ID
where
  hld1.high_rn=1
  and hld2.low_rn=1;

Answer 2

您可以通过row_number() 使用条件聚合：

select PatientName, PatientID, 
       min(case when ;
_hi = 1 then ServiceDate end) as high_date,
       max(score) as high_score,
       min(case when seqnum_lo = 1 then ServiceDate end) as low_date,
       min(score) as low_score
from (select t.*
             row_number() over (partition by PatientID order by Score desc, ServiceDate asc) as seqnum_hi,
             row_number() over (partition by PatientID order by Score, ServiceDate desc) as seqnum_lo
      from t
     ) t
group by PatientName, PatientID;

Answer 3

首先，得到最高分和最低分，然后找出它发生在哪一天

;with scores as (  -- get highest/lowest score per patientId
   select patientID, max(score) hScore, min(score) lScore 
   from table
   group by patientID
), dates as (
   -- get first date a patient had the highest score
   select patientID, min(ServiceDate) dateHScore, score
   from table t
   inner join scores s on s.patientid = t.patientid and s.hScore= t.score
   group by patientID, score
   union all
   -- get the last date a patient had the lowest score
   select patientID, max(ServiceDate) dateLScore, score
   from table t
   inner join scores s on s.patientid = t.patientid and s.lScore = t.score
   group by patientID, score
)
select t.patientName, scores.*
from table t
inner join scores s on s.patientid= t.patientid

如果您想在同一行的不同列上获得最高/最低分，只需拆分第二个 cte 并在选择中适当地加入它们

Answer 4

我还没有测试过，但也许像这样的东西可以完成这项工作：

select x.PatientName, x.PatientId,
  case when x.HSD=1 then 'Highest' else 'Lowest' as ScoreType,
  Score, ScoreDate
from (
  select
    p.PAT_NAME as PatientName,
    p.PAT_ID as PatientId,
    p.MEAS_VALUE as Score,
    p.CONTACT_DATE as ScoreDate,
    row_number() over (partition by p.PatientId order by p.MEAS_VALUE desc, p.CONTACT_DATE desc) as HSD,
    row_number() over (partition by p.PatientId order by p.MEAS_VALUE asc, p.CONTACT_DATE asc) as LSD
  from Patient p
  ) x
where 
  x.HighestScore=1 or
  x.LowestScoreDate=1

您应该收到两行。您需要对数据进行透视才能获得预期的输出。