HMHospitales and Hannover Risk Prognosis

[jansoe]

Predict ICU and death risk for

• HMHospitales dataset
• Hannover dataset

[jansoe]

HMHospitales dataset seems not suited as only 6 persons go to ICU and 16 death. Maybe use as external validation.

[jansoe]

For both datasets quantitative inspection of Brixia-Score seems not to be correlated with any risk.
Questions: Is it a problem of domain transfer or the suitability of Brixia for risk prediction ...?

[jansoe]

Next step: Investigate direct training on risk prediction for Hannover