Some highlights - Between 2019 and 2022, UnitedHealthcare, Humana, and CVS each denied prior authorization requests for post-acute care at far higher rates than they did for other types of care,...
Some highlights -
Between 2019 and 2022, UnitedHealthcare, Humana, and CVS each denied prior authorization
requests for post-acute care at far higher rates than they did for other types of care, resulting in
diminished access to post-acute care for Medicare Advantage beneficiaries.
One program that CVS developed suggested the company should focus on cases it
assigned “a significant probability to be denied
In a May 2019 presentation, CVS determined that it had saved more than $660
million the previous year by denying prior authorizations requests its Medicare
Advantage beneficiaries submitted for inpatient facilities. A majority of these savings
came from “denied admissions.”
My personal favorite, which tells me everything we need to know -
CVS’s testing of a predictive model for inpatient admissions for the company’s
Medicare Advantage beneficiaries showed that a model built to “Maximize Approvals”
jeopardized profits by approving too many cases the company felt should be denied.
Now I understand the scope of this investigation is limited to Medicare and Medicaid Advantage insurers but if this is what is being done on the "public" (I know its not really public) side of insurance, I cannot even begin to imagine what might be going on in the private side.
Have you been denied an insurance claim and want to appeal? Fight fire with fire (and "AI") Fight Health Insurance and improve your chances!
I wondered when we would start hearing about this. Back in March, someone I know from my data science master's program said he had started a new job building "an Al-driven solution that will use...
I wondered when we would start hearing about this. Back in March, someone I know from my data science master's program said he had started a new job building "an Al-driven solution that will use full-text health diagnosis documents to create risk predictions that are then used in pricing life and health re-insurance." Any bets on how long AI has been in use in the insurance world?
Algorithms and actuary tables predate computers, so as long as insurance has existed. Pricing for risk is how health insurance used to work in the US. We still have discounts for nonsmokers which...
Algorithms and actuary tables predate computers, so as long as insurance has existed. Pricing for risk is how health insurance used to work in the US. We still have discounts for nonsmokers which are a thinly-veiled tax on smokers.
indeed, but traditionally the predictions were interpretable (it's pretty easy to see why smoking might raise your health insurance premiums). not so for an opaque matrix of weights
indeed, but traditionally the predictions were interpretable (it's pretty easy to see why smoking might raise your health insurance premiums). not so for an opaque matrix of weights
...a matrix of weights is just a complicated algorithm though. It's still easy to understand what could raise rates. Insurers are well aware of what raises and lowers rates because that's the...
...a matrix of weights is just a complicated algorithm though. It's still easy to understand what could raise rates. Insurers are well aware of what raises and lowers rates because that's the difference between making a profit and going bankrupt. It may be more difficult for a layperson to understand, but an insurer can still tell you what raises and lowers rates. Part of the reason insurers don't talk about every factor that affects rates is because it incentivizes people lying to try lowering their rates.
"Do you have a trampoline in your backyard?"
"Will that raise my rates?"
"Yes, it's a significant liability risk."
"Then no, I don't. I've never owned a trampoline in my life."
It's easy to see some things that raise rates, but there's absolutely no guarantee that the ML model isn't using some other perverse pattern as well. You can surmise some things from the black box...
It's still easy to understand what could raise rates.
It's easy to see some things that raise rates, but there's absolutely no guarantee that the ML model isn't using some other perverse pattern as well. You can surmise some things from the black box but AI explainability is nowhere close to good enough for you to identify everything that led to a given rate increase in practice. This is very different from keeping these factors hidden from insured folks -- even the people who design and train the AI models can't identify everything the model's using to make a given decision (and if they say they can, they're liars).
But we need to think about corporate profits! Kidding. Could not agree more. As someone who has worked in a number of non-clinical healthcare and health IT roles. It's INSANE the amount of...
But we need to think about corporate profits!
Kidding. Could not agree more. As someone who has worked in a number of non-clinical healthcare and health IT roles. It's INSANE the amount of "bureaucrats" and "admins" that exist for essentially no reason but to support the broken healthcare system.
For example - I worked for an HIE (Health Information Exchange) these types of orgs were created as a result of the US HITECH act. Essentially it forced electronic health records (something pretty new at the time) to be more portable and interoperable with other healthcare orgs and software. We had an initiative to collect payment information from all of the insurers on the different procedures and then provide that information to hospitals to allow for more effective pricing. Well it ran for a few years, we correlated all of this data and when we were going to publish it, all of the insurance providers did not renew their contracts with us. Huge portion of the business gone - why? Insurance companies got wise to what was going to happen so they pulled the plug. Since we lost the contract half of the company got laid off (myself included) and the business ended up getting consolidated into another.
The US healthcare system is a broken nightmare of red tape and deception.
Some highlights -
requests for post-acute care at far higher rates than they did for other types of care, resulting in
diminished access to post-acute care for Medicare Advantage beneficiaries.
assigned “a significant probability to be denied
million the previous year by denying prior authorizations requests its Medicare
Advantage beneficiaries submitted for inpatient facilities. A majority of these savings
came from “denied admissions.”
My personal favorite, which tells me everything we need to know -
Now I understand the scope of this investigation is limited to Medicare and Medicaid Advantage insurers but if this is what is being done on the "public" (I know its not really public) side of insurance, I cannot even begin to imagine what might be going on in the private side.
Have you been denied an insurance claim and want to appeal? Fight fire with fire (and "AI") Fight Health Insurance and improve your chances!
What a coincidence! My Starfinder group had a company by the same name that it used for smuggling runs.
I wondered when we would start hearing about this. Back in March, someone I know from my data science master's program said he had started a new job building "an Al-driven solution that will use full-text health diagnosis documents to create risk predictions that are then used in pricing life and health re-insurance." Any bets on how long AI has been in use in the insurance world?
Algorithms and actuary tables predate computers, so as long as insurance has existed. Pricing for risk is how health insurance used to work in the US. We still have discounts for nonsmokers which are a thinly-veiled tax on smokers.
indeed, but traditionally the predictions were interpretable (it's pretty easy to see why smoking might raise your health insurance premiums). not so for an opaque matrix of weights
...a matrix of weights is just a complicated algorithm though. It's still easy to understand what could raise rates. Insurers are well aware of what raises and lowers rates because that's the difference between making a profit and going bankrupt. It may be more difficult for a layperson to understand, but an insurer can still tell you what raises and lowers rates. Part of the reason insurers don't talk about every factor that affects rates is because it incentivizes people lying to try lowering their rates.
It's easy to see some things that raise rates, but there's absolutely no guarantee that the ML model isn't using some other perverse pattern as well. You can surmise some things from the black box but AI explainability is nowhere close to good enough for you to identify everything that led to a given rate increase in practice. This is very different from keeping these factors hidden from insured folks -- even the people who design and train the AI models can't identify everything the model's using to make a given decision (and if they say they can, they're liars).
This, and many other reasons, are why we should have universal healthcare.
But we need to think about corporate profits!
Kidding. Could not agree more. As someone who has worked in a number of non-clinical healthcare and health IT roles. It's INSANE the amount of "bureaucrats" and "admins" that exist for essentially no reason but to support the broken healthcare system.
For example - I worked for an HIE (Health Information Exchange) these types of orgs were created as a result of the US HITECH act. Essentially it forced electronic health records (something pretty new at the time) to be more portable and interoperable with other healthcare orgs and software. We had an initiative to collect payment information from all of the insurers on the different procedures and then provide that information to hospitals to allow for more effective pricing. Well it ran for a few years, we correlated all of this data and when we were going to publish it, all of the insurance providers did not renew their contracts with us. Huge portion of the business gone - why? Insurance companies got wise to what was going to happen so they pulled the plug. Since we lost the contract half of the company got laid off (myself included) and the business ended up getting consolidated into another.
The US healthcare system is a broken nightmare of red tape and deception.
We're having issues with Humana right now. It's exhausting. There's always something.