Closed delwen closed 2 years ago
Prepping the IV exclusion variables now, so we can choose to use as needed. Looking at our filters at https://github.com/quest-bih/clinical-dashboard/blob/main/prep/prep-intovalue.R#L9
I think we only need:
Right?
Disregard:
While it's no problem to add the boolean flags, I'm not sure we should filter for iv_status
since that excludes 545 trials that would still be eligible to be inspected for prospective reg...
iv_status = if_else(
recruitment_status %in% c("Completed" , "Terminated" , "Suspended", "Unknown status"), TRUE, FALSE
)
While I'm here, I can also switch start date filter (by 2018) to another boolean in case we change our mind to include trials started after 2018 (22 trials) [or less logically, with na start date (1 trial)]
@bgcarlisle : may mean some minor tweaks on your end too
Ok, I made the updates to prospective reg dataset and pushed to the branch: https://github.com/maia-sh/intovalue-data/commit/88c1103e98dd5ae54b2a6fad61d804e825f79fcd
@bgcarlisle : for the dashboard you could now update to exclude based on start_by_2018
and iv_interventional
. As said above, i would not exclude based on iv_status
but open for counterarguments.
@delwen for the text, i reran the query to ctgov so use today's date for latest query. also updated in query.log
iv_completion
→ agree to disregardiv_interventional
→ agree to keepiv_status
→ given that DRKS and CT.gov prospective reg data is displayed in the same plot (with toggle), it could be confusing if we apply different iv_status
inclusion/exclusion criteria to these datasets. But I see the argument to extend beyond completed trials. Would suggest we either re-apply our iv_status
filter to the updated CT.gov dataset (and lose trials) or disregard the iv_status
filter in both datasets. Note that we probably couldn't drop this filter completely; we'd probably still need to filter out trials that are 'Not yet recruiting' and 'Withdrawn' (we'd need to check this again and check equivalence in DRKS).has_german_umc_lead
→ let's wait for update from Nicoiv_status
→ ah I forgot about that! indeed could be confusing...
no duplicated in this dataset (I check that assertion)
@maia-sh: check for start 2006; filter start_by_2018
and iv_interventional
and iv_status
and let @bgcarlisle know its ready.
Changes pushed. I checked for 2006 min (all good). @bgcarlisle here is the dataset with exclusions filtered so ready to use.
@delwen Here's some draft text just for content This dataset includes trials led by a German UMC (based on manual evaluation [checking with Nico]) and registered in ClinicalTrials.gov. Trials were started between 2006 and 2018 inclusive per registry data available on 2021-10-06. In line with the primary IntoValue dataset, all are interventional trials with the eligible statuses ("Completed" , "Terminated" , "Suspended", "Unknown status").
Just adding that we decided to keep the iv_status
exclusions for both the CT.gov and DRKS datasets even though we don't strictly need to only focus on completed trials for prospective registration. This means we lose ~500 CT.gov trials, but we gain parity between the plots and avoid having too many separate data sources with different criteria.
We can revisit this later if requested.
Nico suggested we limit the range of the DRKS prospective registration plot to start from 2006 (in line with the CT.gov plot). We'd lose 10 trials. I tend to agree. Again, parity but also because the earlier dates might cause confusion. Thoughts?
@bgcarlisle @maia-sh agree?
Yeah agreed! This should do it! https://github.com/quest-bih/clinical-dashboard/commit/9cff831baf944b1786aabf814d6e77ad6a35cde7
TODO: detail the inclusion/exclusion criteria for the different prospective registration datasets in more detail in the dashboard methods/tooltips.
Done.
For the prospective registration metric, we will include more recent trials using a dataset that Nico previously extracted from the AACT (15.03.2019). It includes trials with a start date between 2006 – 2018 (2018 still included completely but not 2019).
TODO: update the Methods text accordingly.