As far as I know you cannot to this if you don’t have all the information in the same row.
If your day information is always sequential, perhaps the node Lag Column can help you to retrieve the Val1 and Val2 from Day2 to Day1 and from there build your rule engine formula.
You can resolve this in multiple ways but the most adequate solution really depends on what your full data source is looking like and other scenario’s you might encounter. This is quite a narrow view.
For example; does each patient always only have two days throught the entire dataset? In that case the Missing Values nodes can be used. What should be done whenever a patient has multiple days (I suspect you are considering a time frame like a week/month) without a value, what should happen then?