Friends: Please keep bringing these up. We will open a forum of use cases soon. Please be ready to pitch in, this is a community effort. In particular, be ready to build ATLAS or R packages. Or find friends who can do that for you.
@Carlos_Areia, the prediction group is working on who will be readmitted, so it seems like that would be the most natural fit for your question. I don’t recall offhand who’s leading that subgroup, though.
Are there open source or published collections of the de-identified/limited use data that would support these types of analyses? I would be happy to contribute AI/ML techniques applied to other sciences, but do not have access to supporting data.
MEPS or other national level survey data may be an option, but not sure if they are workable for your thought. I am in the traditional data analysis side, such as insurance claim data or EHR…
I second this desire to characterize readmission rates and post-discharge complications of COVID-19 patients. As hospitals come out of the surge, and need to start reporting all-cause 30 day readmission rates to the federal CMS, such information will help educate ourselves and CMS on whether post-discharge COVID patients are similar enough to non-COVID that they should be pooled into the existing quality metrics vs. developing separate metrics.