If the patient gets fired and loses coverage, the assertion won't matter.  Same deal if a lab result previously relied on gets corrected and the change in information changes the result.  Same issue if an employer changes what they pay for.  None of these are any different in the CRD world than in the old world.  If you pick up the phone and use a portal and get back a response, you can 'usually' rely on that response.  However, there are edge cases where things might change and an assertion of 'covered' or 'no prior auth needed' might turn out to be false when all the cards are down.  In some cases, the fact that you'd previously received a contrary notification might  result in the initial assertion being respected.  I.e. If you were told that no prior authorization was needed, you did the service, and then are told prior authorization was needed after all, the payer might waive the requirement.  But it's not guaranteed. The whole system currently relies on receiving responses that are "almost always accurate" and managing the "almost" bit through exception processes when necessary.  CRD can't and won't change that.  It just makes getting the "almost always accurate" responses much faster and easier.  (And ideally, nudges the 'almost' to be even more accurate.)   We will change "However, this doesn't mean that circumstances can't change in a way that invalidates an answer provided" to "However, in rare situations, circumstances may change in a way that invalidates an answer provided.  E.g. coverage is terminated or changed by the employer, data in the record is subsequently found to be erroneous, etc." Will evaluate whether this language should be moved into a 'best practices/usage' section rather than the technical specification.