6 ms·

You're not getting cheaper healthcare by using more machine learning, it's a hard problem for a reason.

thr0wawayf00 4y ago
It's actually not. The CBO already built these models last year and reported that they reduced costs in nearly all cases, across the board[0]. It's only a hard problem because some billionaires want us to believe that it's a hard problem. From the summary: > The report makes many sound assumptions but also some questionable ones that are overly pessimistic. Yet, overall, its bottom-line estimates should reassure those concerned about the economic feasibility of single payer: The CBO projects that such reform would achieve universal coverage, bolster provider revenues for clinical services, and eliminate almost all copayments and deductibles—even as overall health care spending fell. 0: https://www.healthaffairs.org/do/10.1377/forefront.20210210.190243/full/ https://www.healthaffairs.org/do/10.1377/forefront.20210210....
- sumy23 4y ago
  There are political concerns as well. As the article mentions, healthcare spending is 17% of U.S. GDP (which is an absolutely insane figure). Healthcare jobs account for 14% of the workforce according to Census Bureau's 2019 American Community Survey. You can't get the first number down (17% of GDP) without impacting the second number (14% of jobs). I agree that single-payer would be a net benefit, but the U.S. hasn't shown much appetite for such tough political decisions.
  - nopenopenopeno 4y ago
    The same logic can be applied to literally anything in a capitalist society. If the current way of doing things employs people, we can’t possibly do it any other way.
    - sumy23 4y ago
      Yes, but 14% of jobs is a huge number. Healthcare is one of the biggest, if not the biggest, employment sectors.
      - nopenopenopeno 4y ago
        Then replace 1% of it at a time. This is extremely basic stuff if you have any genuine interest in actually doing it.
      - bombcar 4y ago
        If you really want to solve it, and have the political willpower, you just do it over time. Nationalize all hospitals and insurance companies, and continue to run them as they currently are, and slowly streamline systems over decades, and don't hire replacements for retirees.
  - robocat 4y ago
    To agree: most wealthy countries spend somewhere between 10% and 12%[1] of GDP on healthcare. The USA is an outlier in %GDP, and total spending per capita[2], for lower life expectancy - check this graph: https://en.wikipedia.org/wiki/File:Life_expectancy_vs_healthcare_spending.jpg https://en.wikipedia.org/wiki/File:Life_expectancy_vs_health... [1] https://data.worldbank.org/indicator/SH.XPD.CHEX.GD.ZS?most_recent_value_desc=true https://data.worldbank.org/indicator/SH.XPD.CHEX.GD.ZS?most_... [2] https://en.wikipedia.org/wiki/List_of_countries_by_total_health_expenditure_per_capita https://en.wikipedia.org/wiki/List_of_countries_by_total_hea...
- thatfrenchguy 4y ago
  Once you have universal healthcare, it’s still going to be hard. You still have a lot of clever providers trying to extract as much money as possible from the system and an insurer that needs to find out what the right price and cost/benefit analysis is. It’s super super hard, and the US is not used to do that.
0xy 4y ago
It's a hard problem because the real problem is convincing a brain surgeon making $450,000 a year that he should be partially or wholly replaced with AI models. It's the same reason costs to build subway tunnels has dramatically increased in NYC, labor unions have the process by the throat and will accept not even a cent in reduced labor costs, even when improvements would improve safety or efficiency.
- the_d3f4ult 4y ago
  I am not aware of any AI model or technology that functions as anything beyond a toy in the healthcare space. There's very little utility to labelling things in healthcare - even in fields like ophthalmology, dermatology or radiology where companies have made some advances. Humans already do it very well and it's about 1% of the actual battle. Show me an AI that that can make a diagnosis, counsel a patient's family emergently, take the patient to surgery and do a craniotomy then nurse the patient for weeks in the neuro-ICU and I'll be super impressed. Moreover, it's the patient, not the neurosurgeon that has to make the decision to trust the AI.
  - 0xy 4y ago
    Humans are terrible at giving healthcare. There are 8x more deaths in the US due to medical malpractice than from firearms. Clearly humans aren't very good at labeling if they get it wrong a huge portion of the time.

It's actually not. The CBO already built these models last year and reported that they reduced costs in nearly all cases, across the board[0]. It's only a hard problem because some billionaires want us to believe that it's a hard problem. From the summary: > The report makes many sound assumptions but also some questionable ones that are overly pessimistic. Yet, overall, its bottom-line estimates should reassure those concerned about the economic feasibility of single payer: The CBO projects that such reform would achieve universal coverage, bolster provider revenues for clinical services, and eliminate almost all copayments and deductibles—even as overall health care spending fell. 0: https://www.healthaffairs.org/do/10.1377/forefront.20210210.190243/full/ https://www.healthaffairs.org/do/10.1377/forefront.20210210....

sumy23 4y ago
There are political concerns as well. As the article mentions, healthcare spending is 17% of U.S. GDP (which is an absolutely insane figure). Healthcare jobs account for 14% of the workforce according to Census Bureau's 2019 American Community Survey. You can't get the first number down (17% of GDP) without impacting the second number (14% of jobs). I agree that single-payer would be a net benefit, but the U.S. hasn't shown much appetite for such tough political decisions.
- nopenopenopeno 4y ago
  The same logic can be applied to literally anything in a capitalist society. If the current way of doing things employs people, we can’t possibly do it any other way.
  - sumy23 4y ago
    Yes, but 14% of jobs is a huge number. Healthcare is one of the biggest, if not the biggest, employment sectors.
    - nopenopenopeno 4y ago
      Then replace 1% of it at a time. This is extremely basic stuff if you have any genuine interest in actually doing it.
    - bombcar 4y ago
      If you really want to solve it, and have the political willpower, you just do it over time. Nationalize all hospitals and insurance companies, and continue to run them as they currently are, and slowly streamline systems over decades, and don't hire replacements for retirees.
- robocat 4y ago
  To agree: most wealthy countries spend somewhere between 10% and 12%[1] of GDP on healthcare. The USA is an outlier in %GDP, and total spending per capita[2], for lower life expectancy - check this graph: https://en.wikipedia.org/wiki/File:Life_expectancy_vs_healthcare_spending.jpg https://en.wikipedia.org/wiki/File:Life_expectancy_vs_health... [1] https://data.worldbank.org/indicator/SH.XPD.CHEX.GD.ZS?most_recent_value_desc=true https://data.worldbank.org/indicator/SH.XPD.CHEX.GD.ZS?most_... [2] https://en.wikipedia.org/wiki/List_of_countries_by_total_health_expenditure_per_capita https://en.wikipedia.org/wiki/List_of_countries_by_total_hea...
thatfrenchguy 4y ago
Once you have universal healthcare, it’s still going to be hard. You still have a lot of clever providers trying to extract as much money as possible from the system and an insurer that needs to find out what the right price and cost/benefit analysis is. It’s super super hard, and the US is not used to do that.

0xy 4y ago

It's a hard problem because the real problem is convincing a brain surgeon making $450,000 a year that he should be partially or wholly replaced with AI models. It's the same reason costs to build subway tunnels has dramatically increased in NYC, labor unions have the process by the throat and will accept not even a cent in reduced labor costs, even when improvements would improve safety or efficiency.

the_d3f4ult 4y ago
I am not aware of any AI model or technology that functions as anything beyond a toy in the healthcare space. There's very little utility to labelling things in healthcare - even in fields like ophthalmology, dermatology or radiology where companies have made some advances. Humans already do it very well and it's about 1% of the actual battle. Show me an AI that that can make a diagnosis, counsel a patient's family emergently, take the patient to surgery and do a craniotomy then nurse the patient for weeks in the neuro-ICU and I'll be super impressed. Moreover, it's the patient, not the neurosurgeon that has to make the decision to trust the AI.
- 0xy 4y ago
  Humans are terrible at giving healthcare. There are 8x more deaths in the US due to medical malpractice than from firearms. Clearly humans aren't very good at labeling if they get it wrong a huge portion of the time.