So in our data we store drug information as periods with a start and stop where a certain drug was given with some specified dosage and frequency.
Example: Patient X received Drug Y twice a week at a dosage of 20 mg from the first of March 20 to the 15th December 21.
We already had an issue due to not tracking the exact packaging, but we solve this by just mapping to the more generic drug concept instead. Information about the dosage taken by a patient (assuming perfect compliance) can be thus stored by putting the amount (using the matching unit from drug strength) taken during the entire period into ‘quantity’. So to get back to the original daily dosage you would divide that ‘quantity’ by the end date minus the start date.
Example: In the initial example we put 655 days x twice a week x 20 mg = 3743 mg into quantity, assuming that mg is the proper unit based on drug strength.
Now we also have periods that have not ended yet.
Example: Patient X receives Drug Y twice a week at a dosage of 20 mg since the first of March 20 with no plans of stopping.
With no end date we cannot calculate ‘quantity’. We could now impute the end date based on the last date when we confidently know that a patient actually took a medication, so to the end of the observation period. But adding an artificial end date is a pretty heavy change within the data. Medication retention studies are pretty common and there it is of vital importance to distinguish medications that were actually ended from medications that are still ongoing.
So, does anyone here have any guidance about how to include the dosage information for an ongoing treatment course like in our data?
But yes, people, when doing such analysis, should avoid using the records that are pushed all the way against the end of database. We need standardized methods for this.