Hello everybody. Im new to STATA and this forum but I hope you can help me.
I am doing a research project where I intend to estimate the prevalence of venous thromboembolic events (vte) up to 30 days post surgery. I have merged two datasets. Dataset 1 contained information about patient characteristics and surgery date. Dataset 2 contained the date of an thromboembolic event ICD-10 code. After merging these to datasets I have created a new variable called 'interval' to estimate the duration between surgery and date of event to see if it occurred 30 days within surgery. Thereafter I created a variable 'vte', where events occurring 30 days, before surgery or non- events == 0.
The problem is that dataset 2 contains multiple observations per patient ID because the ICD-10 code was registered every time the patient had a consultation at the hospital. This means that the same patient occurs more that once in my merged dataset. For example one patient had an event 5 days after surgery but is also registered 87 days after surgery OR one patient had an event 7 days post surgery but also BEFORE surgery.
As my aim is to identify and compare non-events with events within 30 days I want to keep these patients and drop duplicates. The tricky thing is to remove the right duplicates without removing the first observation occurring 1
I hope you understand my question.
Thank you in advance.
Dear Hanneh
There is probably a solution which is much simpler, but here is my suggestion, given that I understood your question right.
For each individual you calculate the minimum number of days from surgery until vte, given that the event is not before the surgery. This will be the same number for all observations for each individual:
egen min = min( interval) if interval>=0, by(ID)
afterwards you make a new variable were you calculate the difference between the first diagnosed event (min) and all intervals. The rows for each individual with value 0, would be the first diagnose following surgery.
gen first = min-interval
then you deleted duplicates with "first" value
As I understand it, there is a 30-day time lag between non-event and event. You want to keep the data of the patients who have 30-day time lag between non-event and event. If I understand right, you can create a new variable and conditionally drop the others.
gen diff = abs( D_INDDTO-opr_dato)
drop if (diff !=30 & vte==0) or you can check if the variables belong to the same patient with ID using this:
drop if (diff !=30 & ID1=ID2)
I hope this helps, if you can describe the problem with single sheet (image), I can help more.
Best wishes,
Merve
Merve Kurt Thank you for answering :)
But I think you have misunderstood my question a bit. I want to identify patients who have had an event within 30 days post surgery (examples 2,17,28 days and so on). The problem is that some of these patients occur more than one time. As seen in the last screen shot: patient 1081 has 9 observations. I only want to keep the observation with the first registration post surgery, which in this case is interval = 9 and drop the duplicate observations of the patient.
In addition I would want to keep all the patients who have not had an event in order to compare the two groups with regard to patient characteristics later.
What do you mean with a single sheet?
Im sorry im so new in STATA and statistics.
Regards Hanneh.
Dear Hanneh
There is probably a solution which is much simpler, but here is my suggestion, given that I understood your question right.
For each individual you calculate the minimum number of days from surgery until vte, given that the event is not before the surgery. This will be the same number for all observations for each individual:
egen min = min( interval) if interval>=0, by(ID)
afterwards you make a new variable were you calculate the difference between the first diagnosed event (min) and all intervals. The rows for each individual with value 0, would be the first diagnose following surgery.
gen first = min-interval
then you deleted duplicates with "first" value
Dear @Jesper
Thank you so much! :) This worked almost perfectly! You understood my question completely.
However, I have one last question. The above worked great in creating unique observations for patients with events
- Badges
- Science topic