For fun you could also keep track of other related data. How far fetched the prediction is. How confident the voice is. The time frame of the prediction. Just an idea...
The voice always speaks in the same tone.
Predictions are usually something I want to do or am planning to do. Plans it usually tells me as soon as they are conceived. Predictions come often when I am making plans, it tells me whether they will work or not.
HOWEVER, to answer your question, such an exercise could be totally futile as in actuality it changes its mind a lot. Sometimes it tells me something I intend to do won't work, without divulging why. I try it anyway and it kinda works, and I figure out why the point of failure. I fix it, and now the voice tells me it will work now, making the initial prediction kind of moot.
Sometimes something happens and something the voice initially thought was a great idea it now says will be an absolute failure. By that time I am often committed so I tell it there's no choice now, we can only keep going, and it kinda agrees and we do the thing anyway. Sometimes the voice is just plain wrong about things.
The predictions aren't so much predictions as they are opinions. Also, studying it with too hard a lens seems to bring out the worse in the condition, so I avoid it.
It would, however, require a lot of commitment from the patient and willingness to face the condition and even make the symptoms worse. And, considering how crude our tools are at this stage, wouldn’t likely yield much useful information.