I would be interested to know to what extent the prediction of cardiovascular risk goes beyond a linear combination of the other factors such as hypertension, d

I would be interested to know to what extent the prediction of cardiovascular risk goes beyond a linear combination of the other factors such as hypertension, diabetes, age. I.e. if we just prediction these variables with the network then run a linear regression is it less accurate? This would give some indication as to whether there is additional independent information that is useful in the retinal imaging.