Update process_repeat_events.R

[A-Whitmarsh]

Update code to include episodes that started before index_date. If patient is in episode at index_date, update index_date to be end of first episode and then remove first episode.

[elsie-h]

I've made a few changes:

• Remove patients for whom the updated index_date > end_date_outcome (link to code)
• Create a dataset to keep track of how many patients are removed for each value of population and outcome_name (link to code)
• Replace inner_join with left_join to make sure we keep patients with no events in the dataset (e.g. here)
• Update the code to derive data_repeat_events_episodes_long to reduce the number of joins and group_by() commands (link to code)

@A-Whitmarsh please check your happy with these changes.

Additionally, please confirm whether patients can start an episode on index_date or end_date_outcome (as implemented in these lines)

[elsie-h]

@venexia - before merging this branch, we should discuss:

• if index_date falls during an episode, should the new index_date be episode_end or episode_end+1
• can an episode start on index_date or end_date_outcome (and can stset in Stata handle this? If not, maybe we need to -/+1 to index_date/end_date_outcome?)

[elsie-h]

@venexia I've transformed the repeat events episodes data to wide before saving, as discussed on the call.

I do wonder whether it would be better to use survival::tmerge() to format the data with start/stop times at this stage, or whether that would cause problems when analysing it in Stata?