Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

A petabyte of health insurance prices per month

blog.turquoise.health

Post reply on HN
Print view View original

191–200 of 275 posts

Re: A petabyte of health insurance prices per month

#191

Post by yetanotherloss »

redandblack wrote: post #82
FireBeyond wrote: post #60

I used to work for a company that writes claims benefit management software. This: > Notice how every item has a price that requires external information to understand: > Per diem rates are paid for each day a patient is in the hospital. We need to know how long the patient will be in the hospital to know the total amount. > Rates for Cardiac Studies require knowing the price the hospital will bill in the future. The…

Will LLMs eventually help with this - will be great if it can create structured data out of this to learn/classify

Why build LLMs when the problem does not need to exist in the first place. The rest of the civilized world does not need LLMs to solve this because they didn't build this dumpster fire in the first place.

Re: A petabyte of health insurance prices per month

#192

Post by AnthonyMouse »

lotsofpulp wrote: post #173
AnthonyMouse wrote: post #150

Earlier quoted context omitted.

Health insurance company executives are though, aren't they? Their compensation is the "administrative overhead." And a lot of the administrative overhead is proportional to expenses. Commissions are commonly a fixed percentage. How much it's worthwhile to spend on fraud prevention is in proportion to the size and amount of claims. So when premiums and claims costs go up, actual administrative costs go up, but shareh…

>Health insurance company executives are though, aren't they? Their compensation is the "administrative overhead." Yes, and considering they health insurance company executives are not all the richest people in the US, there must exist some pressure to contain their compensation. >once you take lowering claims costs off the table as a way to make more money. This is a pretty big assumption. Surely, UHC/Elevance/CVS/C…

> Yes, and considering they health insurance company executives are not all the richest people in the US, there must exist some pressure to contain their compensation.

Sure there is. But they're also not poor, so there must be some pressure to keep shareholders from paying them minimum wage.

And if they make the shareholders more money, the shareholders will be willing to pay them more. If the way they do that is by making the same margin on a higher cost base, that also allows the shareholders to pay them more. So the incentives all line up to have higher medical costs.

> Surely, UHC/Elevance/CVS/Cigna/Humana risk losing clients if they let their costs go up, and hence their premiums go up, and then a competitor offers their customers lower premiums.

But their competitors have the same incentives.

Suppose you could lower your costs. One thing you could do is keep charging the same amount of money and just make more money, but now that's prohibited. So already we have a disincentive to lower costs right there. Maybe we don't care about this one, but this one is often combined with the second one, and the incentive to achieve the cost reduction is what enables them to do the one we actually like.

Which is to lower prices to try to get more customers. So let's say they lower their premiums by 20% and that gets them 10% more customers. If their absolute profit per customer stays the same, now they're making 10% more money -- great. But now their absolute profit per customer isn't allowed to stay the same. It has to go down by the 20% their costs went down. Meanwhile the lower premiums only got them 10% more customers, so they're losing money on net. Why would they do that?

Re: A petabyte of health insurance prices per month

#193

Post by swalsh »

DoreenMichele wrote: post #182
thomascgalvin wrote: post #15

There are separate charge codes for everything , and all of those codes need to be reflected in the pricing data. There's a price to slap on a bandaid. There's a price to give someone a Tylenol. Hell, W61.61XA is the medical code for "Bitten by duck, initial encounter." Presumably, this means there's also a code for "Bitten by duck, again." Medical billing is broken , and it's no surprise that the amount of data is o…

I worked in health insurance. The code is supposed to make it easy to pay. It doesn't. We could do away with the codes. In addition to having to look up the codes, I read records all day, called and wrote providers for additional info, consulted my technical lead regularly to see if it was covered and once had to print the entire file and get it reviewed by the retired surgeon who came in once a week to read surgical…

I'm in healthcare analytics, the codes are for me.

Re: A petabyte of health insurance prices per month

#194

Post by yardie »

paulddraper wrote: post #189
lotsofpulp wrote: post #106

Earlier quoted context omitted.

Per federal law (ACA), health insurers in the US are required to spend 80% to 85% of revenues on healthcare expenses. Their profit margins are 2% to 6%. Therefore, the higher the revenue, the higher the profit. Where is the incentive to deny legitimate claims?

> Where is the incentive to deny legitimate claims? Because they want to be competitive and gain customers. Which would you sign up for? Insurance that costs $800/month or $1200/month?

Competitive? In every job I've had it was basically the same 3-4 HMOs: Aetna, Bluecross, Cigna, and United Healthcare. If you are getting employer provided healthplans, as most Americans are, those are your options, 4. With so few options you aren't choosing the most competitive, you're choosing the lease worst.

Re: A petabyte of health insurance prices per month

#195

Post by owenmarshall »

boredpeter wrote: post #180
Alupis wrote: post #126

Earlier quoted context omitted.

It's not when you want statistics about all injury types, treatments, etc.

You don’t need arbitrary codes to get statistics for something that can be typed in a description field.

Speaking as someone who has seen the inside of an EHR system, their data models, and what data actually ends up stored there:

lmao

Re: A petabyte of health insurance prices per month

#196

Post by secabeen »

elcritch wrote: post #128
Buttons840 wrote: post #102

Companies are going to grow around this data. An entire industry might grow around this data, and it will be an industry that I support every time I pay an insanely high insurance premium or medical bill. 15% of our workers are in the medical system [0], and this doesn't even include the surrounding insurance industries or pharma industries (I think). Someone has to pay all these people, and that someone is me when I…

There's a similar issue in academia. Tuition costs have risen far faster than inflaction but the number of professors per student is roughly similar. Where does all the extra go? Administrators and bureaucracy!

There's two things here. First, payment for the cost of education has two major parts: tuition, and state allocations to higher ed. When tuition goes up, it can mean that education costs have risen, that state allocations have gone down, or both. Although there have been some understandable increases in the cost of education, the main driver in the increase in the nameplate tuition is the decreases in state allocations. https://fivethirtyeight.com/features/fancy-dorms-arent-the-m...

Second, the nameplate tuition number that looks to have risen is not even the amount that most students pay, as most students get financial aid. The correct number to judge whether tuition has gone up is the Net tuition number. Most reports of the net tuition number show that it has gone up much much less than the nameplate number, but it's also not public in the same way. If we had an anonymized national database of net tuition paid by all students receiving any amount of a federal financial aid, it would go a long way towards improving the quality of the discussion around higher education costs.

Re: A petabyte of health insurance prices per month

#197

Post by yardie »

jimnotgym wrote: post #153

These kind of tales make me very happy to live in the UK with the NHS. Not perfect...but surely better than this?

I've been under the care of both. As an American I can say our system isn't that bad. You see the trick is to always be healthy, don't ever get sick.

/s

Re: A petabyte of health insurance prices per month

#198

Post by oldgradstudent »

mkmk wrote: post #185
oldgradstudent wrote: post #91

For reference, here is the Israeli Ministry of Health price list as of July 1st 2023: https://www.gov.il/BlobFolder/dynamiccollectorresultitem/moh... Edit: Direct link to an Excel spreadsheet.

FYI The above is direct link to .xlsx file

Thanks, noted.

Re: A petabyte of health insurance prices per month

#199

Post by paulddraper »

yardie wrote: post #194
paulddraper wrote: post #189

Earlier quoted context omitted.

> Where is the incentive to deny legitimate claims? Because they want to be competitive and gain customers. Which would you sign up for? Insurance that costs $800/month or $1200/month?

Competitive? In every job I've had it was basically the same 3-4 HMOs: Aetna, Bluecross, Cigna, and United Healthcare. If you are getting employer provided healthplans, as most Americans are, those are your options, 4. With so few options you aren't choosing the most competitive, you're choosing the lease worst.

Humana, Kaiser, Oscar.

But yes, there is an intense economy-of-scale that produces relatively few insurance companies.

(Maybe eventually only one.)

Re: A petabyte of health insurance prices per month

#200

Post by DoreenMichele »

swalsh wrote: post #193
DoreenMichele wrote: post #182

Earlier quoted context omitted.

I worked in health insurance. The code is supposed to make it easy to pay. It doesn't. We could do away with the codes. In addition to having to look up the codes, I read records all day, called and wrote providers for additional info, consulted my technical lead regularly to see if it was covered and once had to print the entire file and get it reviewed by the retired surgeon who came in once a week to read surgical…

I'm in healthcare analytics, the codes are for me.

We were drowning in data. It was a Fortune 200 company and every few years they completely overhauled their system for looking up state exceptions etc, which meant longstanding employees could no longer use it efficiently and new people still didn't magically get it instantaneously and inevitably some important thing could no longer be accessed at all.

It was all in-house, homegrown software, I tried more than once to suggest moving to GIS -- a map-based system -- for some things and was ignored.

I'm curious how that works for you because I just have trouble imagining that actually works. Like what value does that provide that you cannot do some other way and which justifies the systemic cost burden?

Please edify me a smidgen.

Post reply on HN