[HN Gopher] Judge rules $400M algorithmic system illegally denie...
___________________________________________________________________
Judge rules $400M algorithmic system illegally denied Medicaid
benefits
Author : rntn
Score : 260 points
Date : 2024-08-29 17:25 UTC (5 hours ago)
(HTM) web link (gizmodo.com)
(TXT) w3m dump (gizmodo.com)
| blackeyeblitzar wrote:
| > The TennCare Connect system--built by Deloitte and other
| contractors for more than $400 million--is supposed to analyze
| income and health information to automatically determine
| eligibility for benefits program applicants. But in practice, the
| system often doesn't load the appropriate data, assigns
| beneficiaries to the wrong households, and makes incorrect
| eligibility determinations, according to the decision from Middle
| District of Tennessee Judge Waverly Crenshaw Jr.
|
| Not at all surprised to see the name of a big consulting firm
| like Deloitte in something like this. How much money and
| productivity is lost to these leeches across our entire economy?
|
| Leaving aside this particular case: how can potential recipients
| of benefits even know how or why they were denied to bring such
| lawsuits in the first place? Especially if they are forced into
| arbitration? For example I feel like private health insurance
| companies, particularly Aetna, deny many claims falsely as a
| typical approach to avoid having to pay out as much. And patients
| are then subjected to a long drawn out process with hours of wait
| times, hours of calls, and constant vigilance. This method of
| avoiding payouts by creating expenses for patients should be
| illegal. But how can anyone see what's happening and be in a
| position to challenge it without even a basic level of
| transparency?
| OutOfHere wrote:
| IMHO, the state should be the one to approve/deny claims
| (except if they're duplicate claims within a time-period
| specific to the code). Leaving it to the insurance company to
| process the claim is akin to letting robbers play cops.
| throwaway902984 wrote:
| > _how can potential recipients of benefits even know how or
| why they were denied to bring such lawsuits in the first place?
| Especially if they are forced into arbitration?_
|
| You get a letter in the mail, with a review of the decision
| they made. There is a section where they have to explain why
| they denied you. It is a direct, "We don't think you meet this
| criteria" statement.
|
| Awful letters, imo.
| miohtama wrote:
| > The TennCare Connect system--built by Deloitte and other
| contractors for more than $400 million
|
| There you have the reason. The government software should be open
| source because
|
| - People need to be able to make public officers accountable
| (Article XV - The society has the right of requesting an account
| from any public agent of its administration.
| https://en.m.wikipedia.org/wiki/Declaration_of_the_Rights_of...)
|
| - You should be able to fire incompetent contractors, and replace
| them with someone else, for which open source code base makes it
| much more practical and the contractor cannot use any trade
| secrets clauses to make this impossible
|
| > Deloitte was a major beneficiary of the nationwide
| modernization effort, winning contracts to build automated
| eligibility systems in more than 20 states, including Tennessee
| and Texas.
|
| So much tax payers money wasted.
| OutOfHere wrote:
| I would go further and say that not only the software but also
| the execution of the software should also be open, akin on a
| non-financial blockchain database, with the rules, execution,
| and decision recorded permanently (under an anonymous ID, of
| course). Anyone should be able to easily replay the execution.
| This would allow for citizen monitoring of government actions.
| If not for this, at least a weekly export of all anonymized
| processed claims, plus the corresponding logic rules, should be
| made available on the web, similarly facilitating a local
| replay.
|
| The problem with a scheduled export is that a corrupt state
| government will gladly change your records for you (in their
| favor) when you go to court. This is more difficult to do with
| a blockchain.
| striking wrote:
| I would really prefer my health information not be exposed in
| any way, "anonymized" or otherwise, thanks.
| OutOfHere wrote:
| I don't think you realize this already happens. It is
| already available to hundreds of organizations, both
| domestic and foreign. Any organization can license it
| really at a price. What I said would only democratize it,
| also mandating an appropriate level of anonymization which
| is currently inconsistent or missing. Also, think of the
| bigger picture, of the clues such data will yield for
| advances in healthcare.
| KittenInABox wrote:
| Can you link where I can license this information?
| OutOfHere wrote:
| If you offer to give $10+ million to any health insurance
| company or hospital IT company, they will gladly give it
| to you, plus more fees for updated batches. In practice
| it will cost much less if you negotiate well. There also
| are commercial data aggregators that specialize in it,
| that GPT can probably name for you. (I don't want to name
| names here.) At multiple employers, I have worked with
| this data from various vendors.
| Larrikin wrote:
| Which specific vendors?
| striking wrote:
| I think this is plainly untrue in any of the situations
| relevant to what we're discussing.
|
| Some "de-identified" health information can be used or
| disclosed without restriction under HIPAA, but it needs
| to genuinely not be possible to correlate back to an
| individual, which I don't think is something that's
| possible with what you're describing (at least if prior
| medical history and decisions factor into the execution
| of the program). Some health information can be released
| in the public interest (like for research) or for
| bettering healthcare operations, but that's not this
| either.
|
| It might be worth taking a look at
| https://www.hhs.gov/hipaa/for-professionals/privacy/laws-
| reg....
|
| This is excluding stuff like 23AndMe genome sharing /
| Ancestry.com law enforcement collaboration, which might
| be what you're alluding to, but I believe it to be
| irrelevant in this case.
| OutOfHere wrote:
| > it needs to genuinely not be possible to correlate back
| to an individual
|
| What you say is a theoretical goal. In practice, despite
| good efforts, there is no hard rule imposed by reality
| that 100% of records will totally remain anonymized.
| striking wrote:
| Maybe it's unclear why I've brought this up. I understand
| it's not possible to keep everything locked down in a way
| that prevents my information from leaking.
|
| And as unfortunate as that is, that's not my concern with
| your original comment.
|
| I do not want that problem made worse. I don't care if
| it's already kind of bad. I have no idea why the problem
| already being bad would justify making it worse.
| OutOfHere wrote:
| I guess you forgot the part where you're a mortal human,
| and where biomedical data helps discover patterns to
| solve problems. Just you wait until you get old. Also
| forgot the part about exposing statewide corruption wrt
| claims.
| striking wrote:
| I'm not interested in your snark. I'm aware of my
| mortality and of statewide corruption. Those are not good
| reasons to publish my health data publicly even in an
| anonymized form. The people who might be able to fix my
| mortality are already allowed to have my data (whether
| I'm happy or not about that is moot as you've made clear
| yourself). That does not change my belief that the scheme
| you propose for "exposing statewide corruption" is a bad
| idea.
| cdumler wrote:
| I think way he meant a hashed chain: one where the history
| is tamper proof. The important part is that the data and
| logic used to process claim is digitally documented and
| sealed by hash. You get a copy of the hash in your claims.
| The hash chain is public or has to be given to an
| independent third-party. If you get into a kicking contest
| in court, the company they have to reproduce documentation
| with the hash. The hashing prevents forging the original
| documents in the process.
| KittenInABox wrote:
| Nnnnnnah, I don't want claims to be public. There's no real
| way to anonymize claims. e.g. someone getting specialized
| immune therapy in west virginia is pretty damn identifiable.
| Or someone being paid out in a claim of 300k+ for some highly
| specific antiviral.
| OutOfHere wrote:
| > I don't want claims to be public.
|
| I don't think you realize this already happens. It is
| already available to hundreds of organizations, both
| domestic and foreign. Any organization can license it
| really at a price. What I said would only democratize it,
| also mandating an appropriate level of anonymization which
| is currently inconsistent or missing. Also, think of the
| bigger picture, of the clues such data will yield for
| advances in healthcare.
|
| > someone getting specialized immune therapy in west
| virginia is pretty damn identifiable.
|
| Yes, despite the best anonymization, some can still be
| identified, but it will take significant effort, and be
| limited to a very small number of people, like the one in
| WV. Overall though, the ratio of benefits to harms from
| transparency would exceed 999:1.
| Terr_ wrote:
| > blockchain [...] replay
|
| That's burning down your house to kill a spider. Even in a
| closed system infinite replay-ability is a big headache, and
| adding "blockchain" adds a hundred new problems. There's
| already legal system which is a direct dependency, and it is
| more efficient and manageable. (Yes, that's right, the US
| legal system actually looks sane by comparison.)
|
| Instead, make a legal requirement that patients (and
| sometimes other stakeholders) can demand the data/logic
| behind the company decisions.
|
| If you still want very tamper-proof write-only records, that
| can be done _way_ more easily /quickly/cheaply with an old-
| school distributed database with fixed trusted nodes, run by
| multiple institutions that are not likely to conspire
| together.
| OutOfHere wrote:
| > make a legal requirement that patients (and sometimes
| other stakeholders) can demand the data/logic behind the
| company decisions.
|
| Huh. This is already available, and it never works in the
| interest of the patient. Unless one has expensive
| attorneys, no individual has the time and patience to jump
| through all the hoops needed to get it to work. The point
| is to eliminate friction completely.
|
| I do reiterate that scheduled data+logic exports from the
| government would alternatively also help, allowing the
| execution to similarly be replayed. A problem with a
| scheduled export, however, is that a corrupt state
| government will gladly change your records for you (in
| their favor) when you go to court. This is more difficult
| to do with a blockchain.
| troupo wrote:
| Ah yes, because blockchain is the 100% true source of
| ultimate truth.
|
| Despite the little overlooked fact that data in it will
| be input by external systems and people.
| OutOfHere wrote:
| If someone receives a letter in the mail with a claim
| rejection that contradicts what the blockchain says, that
| then is a very easy win in court.
| colejohnson66 wrote:
| The problem is not people's denials being lies, but that
| they _actually were denied_ illegally.
| OutOfHere wrote:
| We're now going in circles. This has already been
| covered. Jump to parent comment #2 in the hierarchy.
| giantg2 wrote:
| It shouldn't be auditable by anyone. There should be an audit
| trail though, and it should be available to the subject of
| that process on request.
| OutOfHere wrote:
| That would be something, but the issue is that most people
| don't have the time or capacity to audit their own records,
| or to hire an attorney to escalate it. Journalists on the
| other hand specialize in this; they can audit everyone's
| records and expose wrongdoing at a state level.
| giantg2 wrote:
| A simple request for the audit record is absolutely
| something that consumers should be able to do. There's no
| guarantee that journalist is going to look at your record
| to get you help. It won't help in edge cases and those
| people will still be screwed.
| OutOfHere wrote:
| > There's no guarantee that journalist is going to look
| at your record
|
| The journalists look at aggregates first, and then
| individuals help tell the story. The point is that both
| individual-audit-records and full data+logic exports
| serve their unique purpose. Anyone should be able to use
| the data+logic to replay the execution, whether for
| oneself or for the entire community.
| giantg2 wrote:
| "Anyone should be able to use the data+logic to replay
| the execution, whether for oneself or for the entire
| community."
|
| No. You can't guarantee anonymity. If you want to have
| someone audit them, use the GAO or similar.
| abduhl wrote:
| >> People need to be able to make public officers accountable
| (Article XV - The society has the right of requesting an
| account from any public agent of its administration. https://en
| .m.wikipedia.org/wiki/Declaration_of_the_Rights_of...)
|
| I agree with this sentiment, but what is the relationship
| between the linked French Declaration and American law? Why not
| try to build the argument based on FOIA:
| https://en.wikipedia.org/wiki/Freedom_of_Information_Act_(Un...
| forrestthewoods wrote:
| Four hundred million dollars is unfathomable.
|
| Has Deloitte ever built any software that is good? I certainly
| have a super strongly negative bias against their work. But hey
| maybe someone here has worked with them and was impressed...
| akira2501 wrote:
| The contract is likely structured into "Scope of Work" and
| "Ongoing Maintenance and Support" to help justify this cost.
| tomrod wrote:
| > Has Deloitte ever built any software that is good?
|
| Not in my experience.
| amonon wrote:
| The United Stated Digital Service is pushing for more
| government software to be open source[1]. I have no idea how
| effective it is, but it made me happy to see
|
| [1] https://playbook.usds.gov/ ctrl+f "Default to open"
| jacoblambda wrote:
| Not just the USDS (as the USDS is a small part of a larger
| effort), the GSA as a whole is pushing for this. They
| maintain a lot of open source software for the federal and
| state governments and they try to do open by default and FOSS
| COTS where they can.
|
| Just the main GSA github org has over 1000 repos in it.
|
| https://github.com/GSA/
|
| https://open.gsa.gov/
| tomrod wrote:
| It's extremely effective. Not just USDS either, but also 18F
| (now GSA TTS). Also, much of the UK government.
|
| The people win when the sun can shine.
| geraldwhen wrote:
| There is no competent government contractor. The only companies
| which survive RFP operate in a way to maximize cost and
| minimize value.
| tomrod wrote:
| There is a subset pushing the change that. Typically people
| connected in some way to USDS or GSA TTS who then go into
| private business. It shows up in surprising places.
| Flozzin wrote:
| Maybe it should be open source. It is using our tax dollars
| after all. But I think this failure isn't a failure of the
| software. But a failure of using software. And of have complex
| laws/criteria of who can be helped. We should craft laws and
| programs that are uncomplex. 400 million dollars on admin
| software to ultimately deny people care they want. That just
| shouldn't be a thing. We should have spent that money on
| helping people, and using it to 'eat' the cost of accepting too
| many people.
|
| Our justice system acknowledges that it's flawed and it's
| flawed with the idea of letting guilty people go, in order to
| make sure we aren't charging innocent people.(granted we are
| failing at that). But we should be crafting care with the
| intention of accidentally helping people who may have no
| qualified so that all qualified people would get care.
| lenerdenator wrote:
| I'll do you one better:
|
| We should just have a publicly-funded health insurance plan
| that applies to every American at a basic level. Then you
| don't have to have a massive bureaucracy to figure out who's
| eligible, who's not, how long they are eligible, etc.
|
| We've successfully spent more money trying to deny people
| public benefits than we probably would have just providing a
| basic level of public benefits to everyone.
| ToucanLoucan wrote:
| I cannot fathom that it's nothing more than a strange
| coincidence that we have spent more money making sure as
| few people access any benefits as possible than it would
| cost to simply provide things, and that those circumstances
| directly benefit some of the largest corporate donors in
| the USA. Like, every business that has employees on welfare
| is receiving a subsidy. The fact that Walmart and other
| large employers (I think maybe Amazon got caught out doing
| this?) pass out literature on how to apply for welfare
| tells you everything you need to know. If those people
| received a basic income, public insurance, etc. etc. that
| would provide a baseline, okay existence, with the option
| to then get a job to buy luxuries and such, not a fucking
| soul would work for Walmart. Because why the fuck would you
| have a job that doesn't pay you enough to have a good life
| unless the alternative is starvation?
|
| Think of every single employer in this country that employs
| people who, despite working full time, still qualify for
| welfare. All of those companies have a direct financial
| incentive to pour shit tons of cash into ensuring our
| social safety net is as shitty as possible.
| jasonlotito wrote:
| > When an enrollee is entitled to state-administered Medicaid, it
| should not require luck, perseverance, and zealous lawyering for
| him or her to receive that healthcare coverage
|
| That's such a powerful statement. If they are entitled, they are
| entitled. Simple as that. And anything that systematically
| prevents that is absolutely wrong.
|
| What's more important is the approval process. Deloitte might
| have build this, but the State approved it's use. People signed
| off on this.
|
| Further, this isn't an easy fix. Loosing coverage like this is
| _not_ something you can just fix. This has affected people 's
| entire life going forward, and not in a good way.
|
| Yes, mistakes happen, but this isn't a simple one-of mistake.
| This is something rotten, and while Deloitte wrote it, they wrote
| it for a client who asked for it and put it out there. And that's
| who is ultimately responsible.
| mlinhares wrote:
| I doubt Deloitte did this without direct guidance and approval
| from state governments.
|
| The unemployment system here in FL was also built to spec to be
| incredibly hard to use and that was the goal of the
| administration. I have no love lost for consultancies, but if no
| government agents are penalized for this and they take the fall
| alone it will happen again with some other consultancy.
| jkaptur wrote:
| I recommend the book "Automating Inequality" by Virginia
| Eubanks for a deeper dive here.
| sitkack wrote:
| https://virginia-eubanks.com/automating-inequality/
|
| > In Automating Inequality, Virginia Eubanks systematically
| investigates the impacts of data mining, policy algorithms,
| and predictive risk models on poor and working-class people
| in America. The book is full of heart-wrenching and eye-
| opening stories, from a woman in Indiana whose benefits are
| literally cut off as she lays dying to a family in
| Pennsylvania in daily fear of losing their daughter because
| they fit a certain statistical profile. Deeply researched and
| passionately written, Automating Inequality could not be more
| timely.
|
| https://www.youtube.com/watch?v=Avxm7JYjk8M
| engineer_22 wrote:
| More likely, they lacked incentive to get it right
| maerF0x0 wrote:
| > by Deloitte and other contractors for more than $400 million--
| is supposed to analyze income and health information to
| automatically determine eligibility for benefits program
| applicants.
|
| Ultimately this needs to be weighed against 1. How accurately,
| corruption free, and expediently would $400M of humans performed
| compared to the software, and 2. How much better software could
| have been delivered by Code for America had they been given the
| $400M in resources.
|
| My gut suspicions are 1. Much worse, and 2. Much better
| justin66 wrote:
| > $400M of humans
|
| Weird.
| tomrod wrote:
| Government procurement contracts in a few ways. Only in small
| ones or DOD do you see FFP. Larger ones are structured like
| "work programs" due to laws.
| gnu8 wrote:
| How many of the software engineers, project managers, and
| sparkle-shoe consultants from Deloitte are on Medicaid?
| josefritzishere wrote:
| It's hard to ignore the fact that the prior governor of Florida
| was convicted of medicare fraud. There is something to that.
| lotsofpulp wrote:
| *prior Governor, promoted to current US Senator. You would
| think stealing from billions of dollars from taxpayers, and
| this is billions in the 1990s, would disqualify you from
| attaining one of the highest government positions.
| lenerdenator wrote:
| The last eight or so years has proven that the vast majority
| of the American system of governance is based on the idea
| that people have basic amounts of shame and are willing to
| not do something if they can't do it in good faith, and that
| such things would be enough to deter malignant parties from
| being involved in government without the need for explicit
| laws.
|
| Now you have a bunch of sociopaths realizing that, oh, wait,
| there _isn 't_ any reason, legally or technically speaking,
| that I cannot be involved in government.
| throwaway902984 wrote:
| It took four years for me to be accepted into TennCare with one
| denial. That seems average, from what I can tell. This system
| definitely contributed to some suicides.
|
| With many of those decisions as incorrect as they were, with
| people as poor as they are here... This is a tragedy that has had
| people hiding the cause of their suicides for an insurance payout
| to their families.
|
| I am guessing that the healthcare costs for those denied are
| generally larger now than they would have been had they been
| accepted; that the hospital system is still paying the price tag,
| not the destitute. - A layered tragedy in that the finances are
| worse for everyone but Deloitte.
| hsbauauvhabzb wrote:
| I don't live in your country and so I'm not familiar, is a
| denial some potentially 'fraudulent' transaction, or something
| else?
| benbayard wrote:
| Not everyone qualifies for government assistance. The
| software Deloitte built was supposed to automatically
| determine if someone was eligible. It is now being alleged
| that the software did not function correctly and was denying
| people because of bugs and issues with the software, not
| because the person was ineligible.
| throwaway902984 wrote:
| You have to apply for the social benefits here, like you
| would apply for a job. It is when the state government says
| no, you don't get to benefit from social welfare programs. In
| this case, TennCare, a health insurance system for the poor
| and disabled.
|
| This government denies those applications at around a 50%
| rate, iirc. With that decision taking around 4 months to a
| year Then an appeal is launched, with that decision generally
| taking 2+ years. A judge rules on that appeal eventually.
|
| A panel is convened with a couple of lawyers, the judge, a
| vocational expert (job fitness person), a health care
| professional, and yourself. Who you are and what you are
| capable of is summed up and deliberated upon. If you are
| deemed indigent, you are given an insurance policy with
| TennCare.
| neilv wrote:
| Imagine you get together a team of really good developers, who
| really want to help government work better for people, pay them
| semi-competitively [*], and just task them with building and
| deploying the software aspects of this system.
|
| Maybe the software is fully delivered within a year, at a cost of
| $10M? (Excluding non-software costs of this project.)
|
| Then those team members can move on to the next project(s).
|
| [*] I'm thinking: don't pay as much as FAANG, but maybe 50%
| FAANG-TC, as straight salary. Meaning FAANG mercenary types
| motivated only by money filter out themselves for you, but it's
| still doable for top-tier people who want to do beneficial work,
| with enough income and job stability that they shouldn't be
| distracted from work by those stresses. You'd really need to keep
| the hiring bar up, though, carefully hand-picking small teams,
| and not compromising if your formula is successful. (And be able
| to say no to agencies want to scale to too many projects in
| parallel, or to some politician wants to make it a jobs program,
| or stuff it with patronage/nepo hires.)
| sitkack wrote:
| I have run consulting before and was thinking of getting back
| into with a similar arrangement. Having been involved
| tangentially in government procurement for small software
| projects, you get first had visibility into how the normal
| predatory ISVs operate. Their single site licensing cost is
| from my experience 20-30% below what it would cost to build it
| yourself. But then you don't own the software and have to pay
| for ongoing maintenance. It looks cheaper but ends up costing
| more for worse software.
|
| This whole sector is ripe for getting flipped because every
| actor in it operates using the same set of greedy rules.
| neilv wrote:
| I wouldn't want to have to be playing bidding games against
| one of the big consulting firms that sometimes bill massively
| to create disasters.
|
| I'd rather just have more information systems be recognized
| as necessarily a government core competency, to in-house. And
| to pay well enough, to hire the best teams for that (without
| leaning on their altruism/patriotism so hard that they have
| to worry about money).
| sitkack wrote:
| I agree, but they often don't have the skills. I do think
| many states could benefit from something like digital
| service, where a state level dev shop could supply services
| to towns and counties.
| jahewson wrote:
| I'm with you but much in-house work relies upon consultants
| anyway. "Doing things" isn't a core competency of most
| modern governments.
| thatguymike wrote:
| This is roughly the Code For America model... there are many
| other roadblocks other than "ability to build a system" though,
| for example the pointlessly strict and bloated regulation which
| is mandatory to follow. Unfortunately it's not quite enough to
| have good engineers.
| gtvwill wrote:
| Private healthcare/medical services should be illegal.
| Nationalise it. So broken to be in 2024 and people still have to
| pay for Healthcare. Thank God I don't live in the US. Our
| healthcare ain't great but it shits on theirs (Aus).
|
| There are zero arguments that private Healthcare should exist.
| thatguymike wrote:
| A plug for everyone in this thread to read `Recoding America` by
| Jennifer Pahlka, founder of Code For America. Extremely good book
| which goes into detail about why software is such a crapshow in
| government... there are many more reasons than "bad software
| engineers", and they're all fascinating.
| tomrod wrote:
| This makes a great case for UBI.
|
| Years to get a benefit that should be dispersed in days or even
| minutes. Shame.
| cco wrote:
| If it was an algorithmic system, why did denials take months?
| melenaboija wrote:
| In banking it is mandated that a group validates any model that
| is used and reports the results to the FED. I wonder if something
| similar should exist for public institutions.
___________________________________________________________________
(page generated 2024-08-29 23:00 UTC)