Program design

[Chung-Li]

Hi all,

I have difficulty designing my program.

In my data set, I have hundreds of patients.

Each patient is recorded with his/her blood pressure, and these records are ordered by hour.

What I am trying to do is that if the the difference between starting time and current time is greater than 24 hour, then record should switch to another day.

It's a bit too abstract, so allow me to present an example:

OBS    ID    SBP    Hour

 1          1      115       15

 2          1      113       17

 3          1      121       23

 4          1      120        0

 5          1      112        5

 6          1      108       13

 7          1      110       15

 8          1      113       16

 9          1      115       17

10         1      113       18

11         1      121       23

12        1      120        0

13        1      112        5

14        1      108       13

15        1      110       15

16        1      113       16

I want to add a variable "Day", and the value is

1, for OBS=1 to 6

2, for OBS=7 to 14 (Since time at 7th-OBS equals time at 1st-OBS, it means this monitoring device has started over 24 hour, so the variable "Day" should added one) 

3, for OBS=15 16 (Since it's the second time that the time at nth-OBS equals the time at 1st-OBS, it means this monitoring device has started over 24*2 hour, so the variable "Day" should added two)

Regarding this question, on one hand, I post it here hope someone can save me from the mud.

On the other hand, I'm searching information about "do-loop in SAS".

If anyone has any idea, I'll appreciate it.

Thanks in advance! 

Best,

Bill

[Kurt_Bremser]

Step 1: convert your time values to real SAS datetime values, then you can use the power of the date/datetime interval functions intnx() and intck().

Well-designed programs are the result of well-designed data.

So go back to the source and retrieve the actual timestamps of the measurements.

[Chung-Li]

Kurt Bremser:

Thank you for your advice, I'll check it!

[RW9]

I would agree with @Kurt_Bremser there on datetimes etc.

I would also suggest that you should look at SDTM models provided by CDISC.  These are pretty standard across the pharma industry, and if you look at the Phuse code repository you might find some code ready to work with them.  The data you present here is not sufficient for task in hand, what does your import agreement state for structure?

[Chung-Li]

RW9:

Thank you for your advice, I'll try this later!

[Astounding]

Questions:

Why does observation 9 not indicate a reset and a new day?

What if observation 7 was just not part of the data?  The hour would skip from 13 to 16, going past the original starting point of 15.  Wouldn't that also indicate a new day?

[Chung-Li]

Astounding:

Sorry for misleading.
That's typo, and I've changed it.

[ChrisNZ · Accepted solution]

Sometimes, especially when data comes from a medical device like here, there is no choice on what you get.

Like this?

data WANT;
  retain START_HOUR MIDNIGHT_PASSED;
  drop   START_HOUR MIDNIGHT_PASSED;
  set HAVE;
  by ID;
  if first.ID then do;
    START_HOUR=HOUR;
    DAY=1;        
    MIDNIGHT_PASSED=0;
  end;
  if HOUR<lag(HOUR) then MIDNIGHT_PASSED=1;
  if MIDNIGHT_PASSED & HOUR>=START_HOUR then do;
    DAY+1;
    MIDNIGHT_PASSED=0;
  end;
run;

SAS Output

OBS	ID	SBP	Hour	DAY
1	1	115	15	1
2	1	113	17	1
3	1	121	23	1
4	1	120	0	1
5	1	112	5	1
6	1	108	13	1
7	1	110	15	2
8	1	113	16	2
9	1	115	17	2
10	1	113	18	2
11	1	121	23	2
12	1	120	0	2
13	1	112	5	2
14	1	108	13	2
15	1	110	15	3
16	1	113	16	3

[Chung-Li]

ChrisNZ:

It's glad to have someone who feels the same way about the data structure (in the medical field)!

Sometimes you just have to adapt to the data, instead of asking data to be the way you want.

Thank you for your help, really!