I just started using Stata. I would like to create a panel data for each patient and count number of different active medications in a given month. In my dataset, there are five columns:
patient_id, first_detected i.e. the first time (month-year) given patient started to take the given medication, last_detected i.e. the last time given patient started to take the given medication, category i.e. the category of the medication.
Code:
* Example generated by -dataex-. To install: ssc install dataex clear input float(patient_id first_detected last_detected category) 1 648 701 665 1 658 661 721 1 658 662 850 1 658 666 1033 1 658 666 1761 1 658 666 2140 1 658 695 239 1 658 695 843 1 658 695 873 1 658 695 1037 1 658 695 1039 1 658 695 1507 1 658 695 2022 1 658 695 2828 1 658 695 2859 1 658 695 2892 1 658 701 533 1 659 701 2499 1 662 666 1105 1 666 695 1245 1 695 695 1040 1 695 695 1109 1 695 695 2478 1 695 695 2888 1 695 701 1862 2 622 637 1293 2 622 637 2892 2 622 643 171 2 622 662 850 2 622 662 1118 2 622 662 1396 2 622 662 2408 2 622 691 1331 2 622 695 2022 2 622 701 1716 2 622 701 2137 2 622 701 2139 2 622 701 2139 2 622 701 264 2 622 701 1308 2 622 701 2911 2 622 701 . 2 627 630 2067 2 627 637 2310 2 628 701 . 2 631 665 1827 2 632 637 70 2 634 701 570 2 636 639 851 2 636 642 2772 2 636 700 849 2 636 700 974 2 636 701 . 2 636 701 . 2 636 701 . 2 637 701 . 2 637 701 285 2 638 642 109 2 638 643 288 2 639 639 2 2 639 639 1918 2 640 701 . 2 641 687 1394 2 642 642 1526 2 642 642 2381 2 642 673 2300 2 694 694 568 2 694 694 1070 2 694 699 2047 2 694 701 2761 2 694 701 2761 2 694 701 2761 2 694 701 2761 2 694 701 2761 2 694 701 2761 2 694 701 2761 2 694 701 2761 2 694 701 2761 2 694 701 2761 2 694 701 2761 2 694 701 2761 2 694 701 2761 2 694 701 2761 2 694 701 2761 2 694 701 2761 2 694 701 2761 2 694 701 2761 2 694 701 2761 2 695 701 2761 2 695 701 2761 2 697 701 2761 3 648 701 665 3 695 695 239 3 695 695 841 3 695 695 843 3 695 695 858 3 695 695 863 3 695 695 869 3 695 695 873 3 695 695 1037 end format %tm first_detected format %tm last_detected
My second and third questions are on modelling advice.
Do I need to use all the data available for each patient at the monthly level or I can simply use a single observation for each event? In the former case, there would be lot of zeros whereas in the latter case, DV would always be 1 and I will use the information available from the month before the adoption of new medication. Do I need to specify the patient_id when I stset the data? I would like to stratify by patient_id/
I also would like to model abonnement of drugs, do I need a separate regression for that or I can actually model both adoption and abonnement in a single regression, like using a competing event specification?
I purchased the Event History Analysis book by Paul Allison, although I think understood the concepts, I still would love to have some advice on the modelling decision.
Thanking you in advance.
0 Response to Counting the number of total active and inactive medication by patient and general advice on Cox regression
Post a Comment