旋转聚合：得到错误的结果

Question

Table 1:

CREATE TABLE Main (patient VARCHAR(5), surgery_date date);
INSERT INTO Main VALUES
   ('P01','2018-01-01'),
   ('P02','2019-01-02'),
   ('P03','2019-01-03'),
   ('P04','2020-07-07');

Table 2:

CREATE TABLE Additional (patient VARCHAR(5), record_date date, bpm integer);
INSERT INTO Additional VALUES
   ('P01','2018-01-01',60),
   ('P01','2018-02-01',49),
   ('P01','2018-02-10',52),
   ('P01','2018-03-10',57),
   ('P01','2018-04-10',57),
   ('P02','2019-01-02',46),
   ('P02','2019-02-02',55),
   ('P02','2019-03-02',55),
   ('P03','2019-01-03',60),
   ('P03','2019-02-03',49),
   ('P03','2019-03-03',49),
   ('P04','2021-07-07',71),
   ('P04','2021-08-07',49),
   ('P04','2021-09-07',49);

第一个 table 有手术日期，第二个 table 有 bpm 读数 post 手术......我想让 % 的患者有平均 bpm（日历上的平均值）个月）> 50 在第一个和第二个日历月 post 手术。（需要忽略手术后第二个日历月的所有数据）

这是我的尝试：

SELECT
   FORMAT(surgery_date,'yyyy-MM') as [Month],
   count(*) as [New_Surgery],
   (100*SUM(COALESCE(CASE WHEN b1.bpm>50 THEN 1 ELSE 0 END,0))) as [Month1_percentage],
   (100*SUM(COALESCE(CASE WHEN b2.bpm>50 THEN 1 ELSE 0 END,0))) as [Month2_percentage]
FROM Main a
LEFT JOIN Additional b1 On b1.patient=a.patient and b1.record_date=a.surgery_date
LEFT JOIN Additional b2 On b2.patient=a.patient and b2.record_date=DATEADD(M,1,a.surgery_date)
GROUP BY FORMAT(surgery_date,'yyyy-MM')

但这并没有给我应有的效果。

我的期望是：

Month   New_Surgery Month1_percentage   Month2_percentage
2018-01 1   100 100
2019-01 2   50  50
2020-07 1   100 0

Answer 1

您的 JOIN 条件仅在恰好一个月或恰好 2 个月返回记录 months.I 已将其更改为包括第一个月的所有测试和第二个月的所有测试。此外，您忘记了将案例总和除以案例数。我添加了 /COUNT(*)。小数据集意味着结果非常simple:0, 50 or 100%.

SELECT
   FORMAT(surgery_date,'yyyy-MM') as [Month],
   count(*) as [New_Surgery],
   (100*SUM(COALESCE(CASE WHEN b1.bpm>50 THEN 1 ELSE 0 END,0)))/COUNT(*) as [Month1_percentage],
   (100*SUM(COALESCE(CASE WHEN b2.bpm>50 THEN 1 ELSE 0 END,0)))/COUNT(*) as [Month2_percentage]
FROM Main a
LEFT JOIN Additional b1 On b1.patient=a.patient 
   and b1.record_date <= DATEADD(month, 1 , surgery_date)
LEFT JOIN Additional b2 On b2.patient=a.patient 
   and b2.record_date<=DATEADD(M,2,a.surgery_date)
      and b2.record_date>DATEADD(month, 1 , surgery_date)
GROUP BY FORMAT(surgery_date,'yyyy-MM')

这给出：

Month   | New_Surgery | Month1_percentage | Month2_percentage
:------ | ----------: | ----------------: | ----------------:
2018-01 |           2 |                50 |               100
2019-01 |           4 |                50 |                50
2020-07 |           1 |                 0 |                 0

Answer 2

我觉得我 over-complicated 通过使用几个 CTE 分成几个阶段，首先将行限制为仅需要的行，然后在计算之前旋转 bpm 列百分比，但它确实 return 您想要的结果。

您的 main table 对于这些结果来说是多余的，因为所有数据都存在于 Additional table 中；如果这里省略了要检索的其他列，您当然可以加入 main table。

with d as (
  select *,
    First_Value(record_date) over(partition by patient order by record_date) Sdate,
    Avg(bpm) over(partition by patient, Year(record_date), Month(record_date)) av
    from Additional
), m as (
    select patient, record_date, Format(sdate,'yyyy-MM') [Month], 
      case when Month(record_date)=Month(sdate) and av >= 50 then av end m1,
      case when Month(record_date)>Month(sdate) and av >= 50 then av end m2
    from d
    where DateDiff(month, sdate, record_date) < 2
), p as (
    select Min([Month])[Month], 
      Count(m1) * 1.0 / Count(*) * 100 M1Percent, 
      Count(m2) * 1.0 / Count(*) * 100 M2Percent, 
      Count(distinct patient) [New_Surgery]
    from m
    group by Year(record_date), Month(record_date)
)
select [month], Max([New_Surgery]) [New_Surgery],
    Max(M1Percent) Month1_Percentage, 
    Max(M2Percent) Month2_Percentage 
from p
group by [Month]
order by [Month]

Working DB<>Fiddle

旋转聚合：得到错误的结果

Pivot the aggregate: getting wrong result

sql

sql-server

pivot