You're looking for explanations in the format P(x1,x2,x3...)->R where P is a "function" (the modeling of your theory), x1, x2 etc are measurable factors and R is a result
Now, you can't measure all the factors, but you can have a good set of them that explains a result closely.
Hence the question is, are there phenomena for which you can measure all the x's that give a good prediction for your result? (or even having good proxies for the x's) And that's not even going into the problem of establishing the theory in the first place
In physics you can repeat experiments multiple times, few experiments are unethical and you can have very exact measurements.
Now compare this with medicine. Also compare the difference between general predictions for a whole population to doing precise predictions for one specific individual. How many x's do you thing would affect risk of cancer/risk of cardiac disease or just the risk of a weird mutation that changes risks slightly (and that medicine hasn't even heard about it)
There's your limit