Present: Aslak, Hans, Øystein, Rune, Arezoo
Status
Rune:
We still don't have enough access to the data.
...
We need informed consent from patients to work on free text.
Arezoo:
ICD-10 codes from Australia ...
...
Summary: First look at one trajectories, then start splitting them into meaningful separate trajectories.
Øystein
talks to the people who translated ICD-10 into Norwegian.
...
Summary: We need to classify trajectories (but first we do just ONE big multi-year trajectory for each patient)
Aslak
How is co-morbidity related to trajectories?
...
Summary: We need (ASAP) a graph showing That Patient, visiting That Department. Next Step: Show this in PE. (Rune, just do it ;-)
Hans
What about mapping between text/ontology? Ø: We don't have any text... Just classifying chronic or not, etc.
I'm trying to summarize the text in the health records.
A: How many patients do you need? Vector-space-model need a lot of training data.
Planning to extract sentences. A: The statements are important. Should be built into a system for real-time note-writing to store the (suggested) codes directly, instead of storing free text.
We want to use the results to rank recommendations. When a doctor is looking at a patients trajectory, the recommendations that best fit this patient show up first.