Add treatment episode start and end dates as arguments to CMA_per_episode

[Masswear]

As an alternative to calculating CMA_per_episode based on gap lenght, give option to provide episode start and end dates as arguments, e.g.:

CMA_per_episode(CMA="CMA1",
                data=med.events,
                maximum.permissible.gap = NULL,
                episode.start.date = "start_episode",  #Name of variable containing episode start date as string or date object
                episode.end.date = "end_episode"; #Name of variable containing episode end date as string or date object
                date.format="%m/%d/%Y")

[ddediu]

I guess that means that I have to call your code? If so, could you pls give me skeleton example that I can expand on?

[Masswear]

No, it means that the user should be able to override the internal calculation of episodes with compute_internal_gaps by providing episode start- and end dates per patient per medication class. Either as separate variable in the dataset or as separate dataset.

[ddediu]

I've been looking at this: currently, the episodes do not explicitly care about medication class (i.e., only a change in medication class might trigger a new episode), so what I propose is:

In CMA_per_episode() if you pass me an extra argument treat.episodes of type data.frame and structure:

• patid the patient ID
• episode.ID the episode unique ID (increasing sequentially)
• episode.start the episode start date
• end.episode.gap.days the corresponding gap days of the last event in this episode
• episode.duration the episode duration in days
• episode.end the episode end date

(i.e., in the same format as compute.treatment.episodes() would have returned), I can use it internally instead of calling compute.treatment.episodes().

Would this work for you?

[Masswear]

In principle yes, but the data.frame structure would ideally look like this:

• patid the patient ID
• episode.ID the episode unique ID (increasing sequentially)
• medication.class the medication class for which this episode is valid
• episode.start the episode start date
• episode.end the episode end date

This would reflect the situation in clinical practice, where medications are prescribed for specific durations, but durations don't need to be the same for each medication.

end.episode.gap.days is not known before CMA calculation and episode.duration is redundant when episode.start and episode.end are given.

[Masswear]

Users can now specify the parameter treat.epi for CMA_per_episode, in the form of a data.frame containing the treatment episodes. It must contain the patient ID (as given in ID.colname), the episode unique ID (increasing sequentially, episode.ID), the episode start date (episode.start), the episode duration in days (episode.duration), and the episode end date (episode.end).

If treat.epi is specified, it overrides the default computation of treatment episodes by gap days. The medication.class column is not required because CMA_per_episode should only be used for single treatments. This option would make sense for a polypharmacy version of the function.