Falg data when certain criteria met

[zimcom]

Greetings!

I have a Vital Sign dataset - systolic and diastolic blood pressure measured pre and post dose

Site	Subject	Visit	Date of Assessment	VS Test	Result
1	1001	Baseline	2019-12-06T20:01:00	Systolic_BP	111
1	1001	Baseline	2019-12-06T20:01:00	Diastolic_BP	75
1	1001	Baseline	2019-12-06T20:01:00	Pulse	89
1	1001	predose	2019-12-06T21:32:00	Systolic_BP	157
1	1001	predose	2019-12-06T21:32:00	Diastolic_BP	72
1	1001	postdose	2019-12-06T21:43:00	Systolic_BP	112
1	1001	postdose	2019-12-06T21:43:01	Diastolic_BP	64

I want to check the consecutive Systolic and Diastolic results be compared and if there is a drop of > 40 between predose and post dose, then flagged it as below

Site	Subject	Visit	Date of Assessment	VS Test	Result	Flag
1	1001	Baseline	2019-12-06T20:01:00	Systolic_BP	111
1	1001	Baseline	2019-12-06T20:01:00	Diastolic_BP	75
1	1001	Baseline	2019-12-06T20:01:00	Pulse	89
1	1001	predose	2019-12-06T21:32:00	Systolic_BP	157
1	1001	predose	2019-12-06T21:32:00	Diastolic_BP	72
1	1001	postdose	2019-12-06T21:43:00	Systolic_BP	112	Drop of > 40mmHg
1	1001	postdose	2019-12-06T21:43:01	Diastolic_BP	64

Thank you!

[ballardw · Accepted solution]

To be clear you want to compare Predose Systolic with Postdose Systolic and Predose Diastolic with Postdose Diastolic for the same subject?

Do you ever have more than one predose measure other than perhaps that baseline?

I would suggest having 2 different flag variables so you don't get confused, one each for systolic and diastolic.

My start would look something like the following untested code.

data want;
   set have;
   by site subject;
   retain presys predia;
   if first.subject then call missing(presys, predia);
   If visit='predose' and vs_test = "Systolic_BP" then presys=result;
   If visit='postdose' and vs_test = "Systolic_BP" then FlagSys = ((presys - result)>40 );
   If visit='predose' and vs_test = "Diastolic_BP" then presys=result;
   If visit='postdose' and vs_test = "Diastolic_BP" then FlagDia = ((predia - result)>40 );
   label
      FlagSys = "Systolic Flag drop > 40"
      FlagDia = "Systolic Flag drop > 40"
   ;
   drop presys predia;
run;

The flags are numeric 1 when true and 0 for false. There are some good reasons to include both, and are set only on the post dose visit for each measure. You could make custom format if you really need to see text like "drop > 40".

Reasons to have two variables: If you use proc report or tabulate to create summaries of the data you can use the N statistic with each to get how many comparisons were performed (I would expect this to be subjects but may be wrong), the Sum statistic would tell you how many had the drop and Mean statistic would give you the percentage of drops in a somewhat cleaner manner than would be needed with character values for two different results in the same variable.

I am assuming your data is in an appropriate order, sorted by site, subject and that "date of assessment". If that is not the case then a sort should be done prior.

[zimcom]

Thank you !!!