Dieser Beitrag ist auch verfügbar auf:
English 
Français (French) 
Deutsch (German)
Entering ICD-10 numbers is a lonely and laborious sport that takes time away from our patients every day. How about instead we formulate a meaningful epicrisis and the computer picks the numbers from it?
The efforts to do so already exist. Initial work dates back to 2015, currently achieving an F-value(balancing between precision and sensitivity) of 95% (Fallisard et al. 2022). A crucial bottleneck for a classification model is the availability of sufficiently annotated texts. For this purpose, the working groups were able to make use of the death certificates. However, to further develop this model on real epicrises, a collective effort is needed.
If you like this idea, please answer the question above (is anonymous) and share this post on your social networks!
