Query to fill in missing dates in a series of rows

Question

Query to fill in missing dates in a series of rows

bnordberg

SSCertifiable

Points: 5480
More actions
December 16, 2008 at 8:18 am

#204047

I have a table with a series of events for a patient. Say they were admitted on day 1, transferred on day 3, then discharged on day 6. I would only have 3 rows, I need to create a resultset with 6 rows - filling in the blanks with a new day, some dummy event ("Stay" or something) and the last location they were in that day.
So my table would look like:
person_id event_type event_date location_id
1Admit2008-12-05 12:00:00.0001
1Transfer2008-12-05 00:00:00.0002
1Discharge2008-12-07 00:00:00.0002
and I need
1Admit2008-12-05 12:00:00.0001
1Transfer2008-12-05 00:00:00.0002
1 Stay 2008-12-06 00:00:00 2
1Discharge2008-12-07 00:00:00.0002
My table would look something like
CREATE TABLE [dbo].[adt](
[person_id] [int] NULL,
[event_type] [varchar](10) NULL,
[event_date] [datetime] NULL,
[location_id] [int] NULL
)
Here are some test data
INSERT INTO [adt] ([person_id],[event_type],[event_date],[location_id])VALUES(1,'Admit','Dec 5 2008 12:00PM',1)
INSERT INTO [adt] ([person_id],[event_type],[event_date],[location_id])VALUES(1,'Transfer','Dec 5 2008 12:00AM',2)
INSERT INTO [adt] ([person_id],[event_type],[event_date],[location_id])VALUES(1,'Discharge','Dec 7 2008 12:00AM',2)
INSERT INTO [adt] ([person_id],[event_type],[event_date],[location_id])VALUES(2,'Admit','Dec 7 2008 12:00AM',2)
INSERT INTO [adt] ([person_id],[event_type],[event_date],[location_id])VALUES(2,'Discharge','Dec 10 2008 12:00AM',2)
INSERT INTO [adt] ([person_id],[event_type],[event_date],[location_id])VALUES(3,'Admit','Dec 8 2008 12:00AM',1)
INSERT INTO [adt] ([person_id],[event_type],[event_date],[location_id])VALUES(3,'Transfer','Dec 8 2008 12:00AM',2)
INSERT INTO [adt] ([person_id],[event_type],[event_date],[location_id])VALUES(3,'Transfer','Dec 10 2008 12:00AM',3)

Viewing 15 posts - 1 through 15 (of 20 total)

You must be logged in to reply to this topic. Login to reply

Lynn Pettis SSC Guru Points: 442467 More actions · Answer 1

Don't have time at the moment to work on this, have a few meetings to attend this morning, but I'd like you to do a search on SSC for the Tally Table. This is a good place to start on your particular requirement.

ChrisM@Work SSC Guru Points: 186127 More actions · Answer 2

A similar question is answered here.

^{“Write the query the simplest way. If through testing it becomes clear that the performance is inadequate, consider alternative query forms.” - Gail Shaw}

For fast, accurate and documented assistance in answering your questions, please read this article.
Understanding and using APPLY, (I) and (II) Paul White
Hidden RBAR: Triangular Joins / The "Numbers" or "Tally" Table: What it is and how it replaces a loop Jeff Moden

bnordberg SSCertifiable Points: 5480 More actions · Answer 3

Yes, but those are all simple examples, throw in a couple extra fields to group your data upon - like my example.

Lynn Pettis SSC Guru Points: 442467 More actions · Answer 4

Okay, so based on those simple examples, what have you come up with and what problems have you encountered?

bnordberg SSCertifiable Points: 5480 More actions · Answer 5

the location_id needs to maintain the previous value - not the min/max or a cartesian product. And I need each day for each person between their admit and discharge.

running something like

SELECT n.[Date] AS Date, b.person_id,Coalesce(b.event_type, 'Stay') AS Event, b.location_id, b.adt_id

FROM dbo.MyCalendarTally('12/1/2008','12/30/2008') n

LEFT JOIN adt b ON b.event_Date = n.[Date]

LEFT JOIN adt x ON x.event_Date = (SELECT MAX(event_Date) AS eventDate FROM adt WHERE event_Date <= n.[Date])

order by n.date, b.person_id

maintains neither. I'm wondering if I should go down the CTE with a row number and then hook to a tally table. But I'm stumped for the moment - hence my posting here!

ChrisM@Work SSC Guru Points: 186127 More actions · Answer 6

bnordberg (12/16/2008)
the location_id needs to maintain the previous value - not the min/max or a cartesian product. And I need each day for each person between their admit and discharge.
running something like
SELECT n.[Date] AS Date, b.person_id,Coalesce(b.event_type, 'Stay') AS Event, b.location_id, b.adt_id
FROM dbo.MyCalendarTally('12/1/2008','12/30/2008') n
LEFT JOIN adt b ON b.event_Date = n.[Date]
LEFT JOIN adt x ON x.event_Date = (SELECT MAX(event_Date) AS eventDate FROM adt WHERE event_Date <= n.[Date])
order by n.date, b.person_id
maintains neither. I'm wondering if I should go down the CTE with a row number and then hook to a tally table. But I'm stumped for the moment - hence my posting here!

This is ugly as a warthog but seems to work...

[font="Courier New"]DROP TABLE #adt

CREATE TABLE #adt (

[person_id] [int] NULL,

[event_type] [varchar](10) NULL,

[event_date] [datetime] NULL,

[location_id] [int] NULL

)

INSERT INTO #adt ([person_id],[event_type],[event_date],[location_id])

SELECT 1,'Admit', CONVERT(DATETIME,'Dec 5 2008 12:00PM', 100), 1 UNION ALL

SELECT 1,'Transfer', CONVERT(DATETIME,'Dec 5 2008 12:00AM', 100), 2 UNION ALL

SELECT 1,'Discharge', CONVERT(DATETIME,'Dec 7 2008 12:00AM', 100), 2 UNION ALL

SELECT 2,'Admit', CONVERT(DATETIME,'Dec 7 2008 12:00AM', 100), 2 UNION ALL

SELECT 2,'Discharge', CONVERT(DATETIME,'Dec 10 2008 12:00AM', 100), 2 UNION ALL

SELECT 3,'Admit', CONVERT(DATETIME,'Dec 8 2008 12:00AM', 100), 1 UNION ALL

SELECT 3,'Transfer', CONVERT(DATETIME,'Dec 8 2008 12:00AM', 100), 2 UNION ALL

SELECT 3,'Transfer', CONVERT(DATETIME,'Dec 10 2008 12:00AM', 100), 3 UNION ALL

SELECT 3,'Transfer', CONVERT(DATETIME,'Dec 14 2008 12:00AM', 100), 4

SELECT person_id, event_type, event_date,

[location_id] = ISNULL(location_id, (SELECT TOP 1 location_id

FROM #adt

WHERE person_id = d.person_id AND event_date < d.event_date

ORDER BY event_date DESC))

FROM (SELECT * FROM #adt

UNION ALL

SELECT DISTINCT a.person_id,

CAST('wait' AS VARCHAR(10)) AS event_type,

DATEADD(DD, 0, n.number) AS event_date,

CAST(NULL AS INT) AS location_id

FROM Numbers n, #adt a, (SELECT person_id, MIN(event_date) AS MINevent_date, MAX(event_date) AS MAXevent_date

FROM #adt

GROUP BY person_id) b

WHERE NOT EXISTS (SELECT 1 FROM #adt WHERE person_id = a.person_id AND event_date = DATEADD(DD, 0, n.number))

AND b.person_id = a.person_id

AND DATEADD(DD, 0, n.number) > b.MINevent_date

AND DATEADD(DD, 0, n.number) < b.MAXevent_date

) d

ORDER BY person_id, event_date, location_id

[/font]

and if it works, then it's time to make it pretty and fast.

^{“Write the query the simplest way. If through testing it becomes clear that the performance is inadequate, consider alternative query forms.” - Gail Shaw}

For fast, accurate and documented assistance in answering your questions, please read this article.
Understanding and using APPLY, (I) and (II) Paul White
Hidden RBAR: Triangular Joins / The "Numbers" or "Tally" Table: What it is and how it replaces a loop Jeff Moden

bnordberg SSCertifiable Points: 5480 More actions · Answer 7

Thanks Chris!

that appears functional. Now to tune and see how it works with a few hundred million rows!

Jeff Moden SSC Guru Points: 1004403 More actions · Answer 8

bnordberg (12/17/2008)
Thanks Chris!
that appears functional. Now to tune and see how it works with a few hundred million rows!

Why on Earth do you need to create what is essentially a potwad of duplicated data? Not being a smart guy here... it could make a difference...

--Jeff Moden

RBAR is pronounced "ree-bar" and is a "Modenism" for Row-By-Agonizing-Row.
First step towards the paradigm shift of writing Set Based code:
________Stop thinking about what you want to do to a ROW... think, instead, of what you want to do to a COLUMN.

Change is inevitable... Change for the better is not.

Helpful Links:
How to post code problems
How to Post Performance Problems
Create a Tally Function (fnTally)

ChrisM@Work SSC Guru Points: 186127 More actions · Answer 9

Jeff Moden (12/17/2008)
bnordberg (12/17/2008)
Thanks Chris!
that appears functional. Now to tune and see how it works with a few hundred million rows!
Why on Earth do you need to create what is essentially a potwad of duplicated data? Not being a smart guy here... it could make a difference...

Totally agree Jeff, this is a reporting requirement.

^{“Write the query the simplest way. If through testing it becomes clear that the performance is inadequate, consider alternative query forms.” - Gail Shaw}

For fast, accurate and documented assistance in answering your questions, please read this article.
Understanding and using APPLY, (I) and (II) Paul White
Hidden RBAR: Triangular Joins / The "Numbers" or "Tally" Table: What it is and how it replaces a loop Jeff Moden

bnordberg SSCertifiable Points: 5480 More actions · Answer 10

The data in this format will allow for contract tracing of MRSA infections. Our statisticians can pipe this through some algorithims in SAS to see if we can't find patterns of infection/spread.

Jeff Moden SSC Guru Points: 1004403 More actions · Answer 11

bnordberg (12/17/2008)
Thanks Chris!
that appears functional. Now to tune and see how it works with a few hundred million rows!

Are you sure it works? When I run it, I only get the original rows...

person_id event_type event_date location_id

----------- ---------- ------------------------------------------------------ -----------

1 Transfer 2008-12-05 00:00:00.000 2

1 Admit 2008-12-05 12:00:00.000 1

1 Discharge 2008-12-07 00:00:00.000 2

2 Admit 2008-12-07 00:00:00.000 2

2 Discharge 2008-12-10 00:00:00.000 2

3 Admit 2008-12-08 00:00:00.000 1

3 Transfer 2008-12-08 00:00:00.000 2

3 Transfer 2008-12-10 00:00:00.000 3

3 Transfer 2008-12-14 00:00:00.000 4

--Jeff Moden

RBAR is pronounced "ree-bar" and is a "Modenism" for Row-By-Agonizing-Row.
First step towards the paradigm shift of writing Set Based code:
________Stop thinking about what you want to do to a ROW... think, instead, of what you want to do to a COLUMN.

Change is inevitable... Change for the better is not.

Helpful Links:
How to post code problems
How to Post Performance Problems
Create a Tally Function (fnTally)

Jeff Moden SSC Guru Points: 1004403 More actions · Answer 12

bnordberg (12/18/2008)
The data in this format will allow for contract tracing of MRSA infections. Our statisticians can pipe this through some algorithims in SAS to see if we can't find patterns of infection/spread.

Got it... lemme see what I can do... yep, I know... Chris posted code for this already...

--Jeff Moden

RBAR is pronounced "ree-bar" and is a "Modenism" for Row-By-Agonizing-Row.
First step towards the paradigm shift of writing Set Based code:
________Stop thinking about what you want to do to a ROW... think, instead, of what you want to do to a COLUMN.

Change is inevitable... Change for the better is not.

Helpful Links:
How to post code problems
How to Post Performance Problems
Create a Tally Function (fnTally)

ChrisM@Work SSC Guru Points: 186127 More actions · Answer 13

Jeff Moden (12/18/2008)
bnordberg (12/18/2008)
The data in this format will allow for contract tracing of MRSA infections. Our statisticians can pipe this through some algorithims in SAS to see if we can't find patterns of infection/spread.
Got it... lemme see what I can do... yep, I know... Chris posted code for this already...

Jeff, I was hoping you'd take the bait...proof of concept is one thing, turning prototype code into 'nasty fast' production code is entirely another. My query will not scale well. Yours will, and I look forward to learning from it.

^{“Write the query the simplest way. If through testing it becomes clear that the performance is inadequate, consider alternative query forms.” - Gail Shaw}

For fast, accurate and documented assistance in answering your questions, please read this article.
Understanding and using APPLY, (I) and (II) Paul White
Hidden RBAR: Triangular Joins / The "Numbers" or "Tally" Table: What it is and how it replaces a loop Jeff Moden

Jeff Moden SSC Guru Points: 1004403 More actions · Answer 14

Chris Morris (12/18/2008)
Jeff, I was hoping you'd take the bait...proof of concept is one thing, turning prototype code into 'nasty fast' production code is entirely another. My query will not scale well. Yours will, and I look forward to learning from it.

Heh... this one isn't quite as easy as folks would think and still maintain any performance in the face of scalability at all... hope I can live up to your good expectations and mine. 😛

--Jeff Moden

RBAR is pronounced "ree-bar" and is a "Modenism" for Row-By-Agonizing-Row.
First step towards the paradigm shift of writing Set Based code:
________Stop thinking about what you want to do to a ROW... think, instead, of what you want to do to a COLUMN.

Change is inevitable... Change for the better is not.

Helpful Links:
How to post code problems
How to Post Performance Problems
Create a Tally Function (fnTally)