Hello,
I have a question about the data observation window endpoint. For each patient, does the train/test dataset contain records from the beginning to the time of **PCR test** or the **time of hospitalization** (or 5 weeks if there is no hospitalization)?
I know that the uploaded dataset may contain everything. I'm just asking about the evaluation setting that will be used for evaluating our models by challenge organizers/evaluators.
Thank you in advance,
Hamed
Created by Hamed Fayyaz hamedcan Hi @oneilsh and @hamedcan,
Shawn, you're correct, the observation window should not _begin_ at the initial outpatient visit, it should _end_ at the observation window. Your models should be making predictions as though they were being used at the time of the outpatient visit. Apologies for the confusion on my part.
You are expected to trim the data yourselves and should be ignoring any data that occurs after the outpatient_visit_start_date. We chose not to do the trimming for everyone in order to give you the most flexibility in how you build the training and validation data for your internal model development. I'll be adding information to the challenge instructions to clarify this.
Hope this clears everything up!
@trberg Hi Tim,
I'm also curious about this question; browsing the data it looks like there are many events (e.g. lab measurements) that occurred long prior to the COVID index date (many months), as well as events that occurred long after hospitalization (by several months). Should we trimming the data ourselves? It sounds like from your answer for task 1 we should be ignoring any data (other than the outcome column for training purposes of course) that occurs after the outpatient_visit_start_date (assuming that during evaluation such data won't be available, as though the evaluation were performed at that point in time per patient). But then I wasn't sure what you meant by the observation window *beginning* with the initial outpatient visit...
Thanks for clarification!
~Shawn Hi @hamedcan,
I've updated the [Challenge Instructions](https://unite.nih.gov/workspace/report/ri.report.main.report.979b9415-adfe-48e8-bbf2-c6cf9b06e486) report in the project space. The observation window begins at the time of the initial outpatient visit. The Task 1 Gold Standard will most likely be imported tomorrow. In the meantime, I've added the Task 1 Gold Standard data generation script to the Resources folder, so you're welcome to look through that as well.
Thank you,
Tim