They probably can’t, assuming that they haven’t adopted a more complex (and, consequently, probably less reliable at what it purports to represent) diagnostic categorization than is generally used for other purposes in healthcare; the usual systems can distinguish admitting and primary diagnoses and diagnoses that are neither admitting nor primary (though I don’t know that the distinction is in what is shared for public health reasons), but it can’t distinguish between “diagnosis that are part of the reason that they were admitted to or not yet released from the hospital” and “diagnoses that aren’t part of that despite being present”.

At least, that’s my understanding based on a couple decades of contact with health IT systems and the associated data models.