Another well done post, but a note of caution- I’ve seen this movie before. The study cited seems reasonable, but it is limited and a fairly small sample. Years ago, the Department of Defense did a study on privatizing one thing and from a single instance went ‘all in’ on the idea, applying it to anything that moved. Expansion to other applications did not fare as well, but they kept trying. I believe that there will be value gained by AI, but it should come the old fashioned way, by earning it, not on the basis of hype that dooms it to ‘success.’ We need more studies like the one you discuss to pave the way.
Thanks. I agree with you that the study in question is small and limited, but we have to start somewhere. Hopefully more research will arise which shows increased productivity in other areas.
I worry that AI will make healthcare less productive, in the sense that AI will increase the amount of revenue going into the system but not change health outcomes. Using ai to up code people in Medicare advantage is a good example.
Thanks for the comment. I'm not sure what you mean by "up code people in Medicare advantage". The research I discuss in this post concluded that opthamologists who use AI-enabled scanners for screening for diabetic retinopathy were able to screen more patients per hour than the doctors who didn't use these tools. That sounds like improved productivity to me.
Revenue on MAPD is based on how many risk codes someone has on their chart. So if diabetic retinopathy is a code then the insurance plans will want to do their best to make sure the doctor finds it and codes it. Often the doctors are in risk deals where finding codes increases their pay via revenue sharing. At scale some clinics become upcoding factories either up to or over the line of the law.
The cover story is that accurate coding is good for care coordination. The cynical story is that the same members on MAPD have more coding then equivalent traditional medicare people, which makes MAPD more expensive then trad medicare. The latest VC companies in the space try to use AI to find potential codes by looking at health data and prompt doctors to try and squeeze them into exams.
Short version, the more things they find wrong with you (whether or not they exist and whether or not they do anything effective to treat you) the more money goes into the medical system. Whenever the cost of a medical encounter is less then the codes it will generate the plans will try their very best to get you in front of a provider, even if they have to send someone to your house.
Another well done post, but a note of caution- I’ve seen this movie before. The study cited seems reasonable, but it is limited and a fairly small sample. Years ago, the Department of Defense did a study on privatizing one thing and from a single instance went ‘all in’ on the idea, applying it to anything that moved. Expansion to other applications did not fare as well, but they kept trying. I believe that there will be value gained by AI, but it should come the old fashioned way, by earning it, not on the basis of hype that dooms it to ‘success.’ We need more studies like the one you discuss to pave the way.
Thanks. I agree with you that the study in question is small and limited, but we have to start somewhere. Hopefully more research will arise which shows increased productivity in other areas.
I worry that AI will make healthcare less productive, in the sense that AI will increase the amount of revenue going into the system but not change health outcomes. Using ai to up code people in Medicare advantage is a good example.
Thanks for the comment. I'm not sure what you mean by "up code people in Medicare advantage". The research I discuss in this post concluded that opthamologists who use AI-enabled scanners for screening for diabetic retinopathy were able to screen more patients per hour than the doctors who didn't use these tools. That sounds like improved productivity to me.
Revenue on MAPD is based on how many risk codes someone has on their chart. So if diabetic retinopathy is a code then the insurance plans will want to do their best to make sure the doctor finds it and codes it. Often the doctors are in risk deals where finding codes increases their pay via revenue sharing. At scale some clinics become upcoding factories either up to or over the line of the law.
The cover story is that accurate coding is good for care coordination. The cynical story is that the same members on MAPD have more coding then equivalent traditional medicare people, which makes MAPD more expensive then trad medicare. The latest VC companies in the space try to use AI to find potential codes by looking at health data and prompt doctors to try and squeeze them into exams.
Short version, the more things they find wrong with you (whether or not they exist and whether or not they do anything effective to treat you) the more money goes into the medical system. Whenever the cost of a medical encounter is less then the codes it will generate the plans will try their very best to get you in front of a provider, even if they have to send someone to your house.