How to use NarX for health care and patients data

To understand and make a simple scenario, I've an Excel file with data for some patients and wants to use it for prediction.

The details for 2 patients are as follows

Columns:

PatientID, Score1, Score2, Age, IsDisease, Date, Score1_F, Score2_F

Data:

1, 1, 2, 33, 0, 2/12/1983, 1.5, 2
1, 1.5, 2, 33, 0, 2/12/1983, 2, 2
1, 2, 2, 34, 0, 2/12/1983, 2.5, 2.5
1, 2.5, 2.5, 34, 0, 2/12/1983, ,
2, 3, 2, 25, 0, 3/7/1990, 3.5, 3
2, 3.5, 3, 25, 0, 3/7/1990, 4, 3.5
2, 4, 3.5, 26, 1, 3/7/1990, 4.5, 4
2, 4.5, 4, 25, 1, 3/7/1990, , 

I want to predict Score1, Score2 which I think makes them my y(t).

From the next record which can be after 6 months visit I copy Score1 and Score2 to Score1_F and Score2_F for y(t+1). The last values of Score1_F and Score2_F will be empty as they are the future values. So the age is also changing as the visit is after 6 months and also IsDisease which is 1/0 field changing for patient 2.

My x(t) will be age, IsDisease, and Date. In the real case there will be many other constraints but just to understand I want to run this first.

So now how to tackle this problem? How to train for each patient separately? What could be the delays? The nearest similar sample dataset I found is Pollution Mortality but I believe my case is different than that. I just ran NarX with NaN values in Score1_F and Score2_F in the last record for each patient and got some result but need to understand and analyse it. Also I've shortage of data. So what could be the good number. e.g if I can get 4 records for each patient (assuming ideal scenario) with 6 or 3 months interval, how many patients will be enough for this kind of prediction?