Raw LLM Responses

I don't work for Cigna, but another insurer and what they are doing is definitely not the standard. What is supposed to happen: an algorithm looks at a prior authorization, determines if the policy for approval is being met (i.e. what tests have been done, blood pressure, etc) and recommends approval or further review. A nurse is supposed to check the results and in the case of denial it gets sent to a medical director to verify the denial if it is final and not due to insufficient information (there can be back and forth if they need more info). Algorithms in this case can actually help ease the burden by pulling out information from long medical records which can be dozens of pages long. So a good version of this would be an AI that looks for criteria which matches the approval policy and surfaces it for nurse / doctor review. Even going as far as auto-approval. This is good for the patient and doctor because it saves time. Denials should always be manual review. In the world of AI, there should always be a manual review of a denial in these cases. An algorithm can't be held accountable, but a person can. A doctor should always have the final say. What makes this worse is not just the fact that the AI is denying approval, but the doctors who are rubber stamping it are actively being discouraged from disagreeing with the algorithm denials. This is the exact opposite of how algorithms should be used responsibly. You can disagree with the system we have - the US healthcare system truly sucks - but this goes way beyond the healthcare system and into a completely irresponsible and repugnant way to use algorithms.