Thank you for this package! I have been exploring how we could use it at Karolinska, Sweden and found some features that would make it more useful for us.
Is Your Feature Request Related to a Problem? Please Describe
I have been exploring a different approach to the specialty predictions. I did not have access to consultation sequences in my data but a rich set of events in the ED. So, I build a model with a transformer architecture that uses the events and meta data like visit reason, age, sex etc to predict the ward where the patient will end up. I used torch for this and managed to generate probabilities per ward that I could use as weights for the distributions of admissions for patients currently in the ED. Thereafter I struggled to incorporate my model into the aggregation pipeline due to the objects looking quite different from the approach you suggest.
Describe the Solution You'd Like
I think some modifications to support a wider range of specialty models when creating the full pipeline would be required and then maybe a standardised/documented interface to implement when using another model architecture for specialty.
Describe Alternatives You've Considered
No response
Additional Context
No response
Thank you for this package! I have been exploring how we could use it at Karolinska, Sweden and found some features that would make it more useful for us.
Is Your Feature Request Related to a Problem? Please Describe
I have been exploring a different approach to the specialty predictions. I did not have access to consultation sequences in my data but a rich set of events in the ED. So, I build a model with a transformer architecture that uses the events and meta data like visit reason, age, sex etc to predict the ward where the patient will end up. I used torch for this and managed to generate probabilities per ward that I could use as weights for the distributions of admissions for patients currently in the ED. Thereafter I struggled to incorporate my model into the aggregation pipeline due to the objects looking quite different from the approach you suggest.
Describe the Solution You'd Like
I think some modifications to support a wider range of specialty models when creating the full pipeline would be required and then maybe a standardised/documented interface to implement when using another model architecture for specialty.
Describe Alternatives You've Considered
No response
Additional Context
No response