From my experience, ELISA can detect casein even already the denatured one at very small concentration, which I am not sure that the casein of such condition and amount can cause any allergy.

I think ELISA method may be applied to detect intact protein if proper antibody against the protein is available.

I am thinking that ELISA detecting cow's milk protein is suitable to detect contamination of dairy product in food processing but not to evaluate remained fraction of milk protein in hydrolysate formula as ELISA technically does not detect whole molecule of milk protein, just some specific part of the protein. Result of ELISA in hydrolysate cow's milk base protein formula will not reflect the true value of remained cow's milk protein.

When using one single polyclonal antibody raised against the whole purified protein, both intact protein and certain peptides can be detected.

I suppose that if you are interested in detecting whole proteins in hydrolisate formulas, you could use a sandwich ELISA with an antibody PAIR (one for coating and the other one for detection) targeting 2 specific epitopes on opposite sides of the protein sequence.

The choice of your target protein may be tricky, but this theoretically works. It would be useful if you could have the two antibodies from different species or otherwise have one of the pair (only the one used in detection) already coupled with peroxidase or such. 

However, in my experience, is can be difficult to obtain good antibodies against milk proteins such as caseins, especially when targeting a short peptide sequence. 

Lastly, products of partial hydrolysis can retain much of the allergenicity of the parent protein.