Dark Algorithms and UX in Health Insurance | Hacker News Reader

Dark Algorithms and UX in Health Insurance

jb-rubinovitz.ghost.io

203 pointsluu65 comments

This is a classic case of attributing the results of stupidity to malice. Insurance companies are heavily regulated and subject to massive fines and public embarrassment for things like deliberately preventing the insured from filing claims. And this sort of petty detail in the UI, being directed from C-level executive management? It doesn't pass the smell test. They're managers, not supervillains.

No, this is just garden-variety incompetence. This is the result of incredibly complex specs being implemented by teams of dozens or hundreds, including offshore work and highly distributed teams. The kind of evil plots suggested here would be passing through not just executives, but architects, product owners, scrum masters, lead engineers, and the delightfully unpredictable hands of the cheapest programmers money can buy. Dozens of opportunities for sabotage of something so overtly evil.

I've worked in insurance IT quite a bit. They're just not smart enough to do this.

>This is the result of incredibly complex specs being implemented by teams of dozens or hundreds, including offshore work and highly distributed teams. The kind of evil plots suggested here would be passing through not just executives, but architects, product owners, scrum masters, lead engineers, and the delightfully unpredictable hands of the cheapest programmers money can buy.

Perhaps that is the evil plan working as intended

Yeah I mean the fact is that revenue comes from people not collecting benefits.

Nope. The vast majority of benefits are processed without incident or complexity, between the medical provider and the insurer, with no action required by the insured. Rigging some UI for hand-entry of claims in a way that suppresses only some, but not all? That's not going to make an appreciable difference in the bottom line.

I'm not sure what the "nope" is in response to: Insurance companies profit when people don't collect on their benefits. If I pay and don't use it, they win. This is why companies that self-insure take out further insurance on their pool in case someone does actually get very sick and it drains the pool, which does happen more often than you'd think.

Not really. You can find the same sort of gruesome inefficiency and bad UX in systems built by similar organizations that are supposed encourage, rather than discourage activity. The deeper problem here is a focus on internal business needs (and terminology) as the "customers", rather than end users. Heck, I've seen the same basic opaqueness and clumsy, jargon-laden UX on things as simple as credit card terminals.

The only time you see really good UX for these things is in the startup (or post-startup, ie Amazon or Google) world, where UX for end users becomes a top-level requirement, and driving clicks is measured and actively improved.

This article isn't great, it provides weak evidence for the first point and speculates on the final two points based on personal experience.

The fact that this article is garbage doesn't mean it's wrong, though. I'm very close to someone who works on healthcare UI/ux and it's true that they aren't allowed to encourage use of their healthcare (ie going to a doctor). Whether or not this goes farther to out right denying claims, I can't say, as it likely depends on the provider (UA is likely to be awful whereas non-profit orgs are probably less so). But at the very least, they put a lot of thought into minimizing costs, which can hurt patients in life threatening ways. It's pretty easy to see the incentives are all fucked and will lead to this sort of thing.

Given that insurance claims can determine if someone lives or dies I'm not very interested in parsing what the intent was in the, as you rightly say, byzantine halls of large insurers. What matters are outcomes, and these outcomes are nowhere near aligned with the patients best interests.

The sad reality is that the state must hold them to account because a few blog posts here and there is going to do nothing in the face of billions of dollars.

Edit: before anyone else bothers, I am fully aware that pointlessly complex state processes, greed, and corruption are part of why the regulations are so complicated. But, making them "simpler" doesn't change the profit motive: the only way to truly solve this problem is to provide a public option because currently patient outcomes aren't aligned with profit interests.

Most large insurers barely provide any insurance any more. Most of their customers are self insured employers. The "insurance" company just administers claims without bearing much risk. Approving a claim doesn't directly impact their profits. So their interests are better aligned with patients than you think.

Then again we have cases like the Volkswagen diesel scandal which managed to go past (I guess?) similar processes and make it into full car production.

Not similar at all. The regulatory burden on insurance companies is orders of magnitude more immense than that of auto makers. The complexity of insurance and the software products used in the industry make the auto industry seem like tinker toys.

It's not clear that the regulatory burden surrounding web page UI is all that steep. Peeking at CA's law (Title 10), for example, doesn't yield a whole lot. However, what does stand in OP's favor is that it must pass standards identified by the ADA and in secion 504 (Rehab act). This might be a way to fight back- unless under some adaptive technology everything magically appears haha.

>They're managers, not supervillains.

Those are essentially the same.

Hit the nail on the head, this is definitely the reason folks. Similar issue in the fintech space.

This post lists three examples of dark algs/UX.

However, the second and third examples are speculative. They might be a result of an algorithm, or they might be a result of stupid policies, or they might be a result of operational errors. Who knows? Alleging that it's a "dark algorithm" sounds cool, but it isn't really damning without evidence.

So we're down to one example of actual dark UX, which doesn't make for a particularly strong allegation. I wish the author had taken just a little more time before publishing to collect a few other authentic examples.

That said, I know from my own experience that health insurers' online portals tend to have horrible UX, because it is not in their best interest to have great UX. Just yesterday, I had to fill out a form indicating that none of the five members of my household had other insurance or were Medicaid-eligible. There were a total of 10 checkboxes I had to check. After checking each one, I got a loading indicator and then the page refreshed, so I had to scroll the page back down to the next checkbox. It didn't prevent me from filling in the information, but it definitely made it needlessly tedious.

It's further concerning that he appears to have written his own example code for the speculative dark patterns, presenting it alongside real code from insurer websites, which I found misleading at best and downright deceptive at worst...

... at best.

And then, like, criticize the code quality it as if it came from them for design choices that _the author_ made?

> So, my laugh cry scenario here is that they statistically set/just have variables in the code base that say claims spanning a certain amount of time are denied. E.g.

WTF, you wrote that code!

Well! That's rather ironic I'd say.

He is pointing out the arbitrary nature of the algorithm the workings of which was determined through research. The psuedo code is obviously not the point. The reader would have to either be ignorant of the discussion to make this mistake or intent on blurring the discussion.

Agreed. If my goal is to trick people into not submitting claims it would be pretty naive of me to disable or hide the button.

'oh i'll just give up and eat the $X,000 charge because I cannot find the submit button?'

Yeah right. The only thing that would do is add to your expenses tying up on-call tech support.

which fulfills the prophecy that healthcare is expensive because large call centers must be created to support all the people who aren't smart enough to submit their claims online, which in turn either raises taxes or makes the taxes go towards a call center and not care.

Im disappointed to confirm these are quite real. I've seen a particular case where a benefits enrollment page has a check box list where your selected enrollments at the top of the page _silently change as you edit fields near the bottom_. And you can bet which direction those values change. It doesnt hide a submission button like the author's example but it DOES prey on those who neglect to review their submission in full before continuing (which I guarantee is a significant and quantifiable benefit to the payer)

It's a big stretch of the imgaination to dismiss this behavior as 'designer error'. This is a great example of where class action lawsuits can and should be employed.

Insurance companies give me 90 days to submit a claim, then they do not have to pay it. I guess their rationale is that if I did not bother to submit an invoice, then why should they pay? And by the way, a lot of them will not accept paper claims any more. They want all the claims done electronically through their ancient systems.

And by the way, this push for electronic medical records(EMR) was a way for them to further decrease their overhead (thank you insurance lobbies). EMR that is CCHIT certified (= Certification Commission for Health Information Technology, pronounounced 's-h-i-t', for real) can interface with payment clearing houses directly (I as a physician have to pay to use a clearing house, btw). Then the claims are scrubbed by their scripts and rejected for various reasons, (for example, no referral from PCP, invalid gender for treatment code,etc). No humans involved. This allows them to lay off vast swaths of auditors. And effectively outsources their work onto the physician's office. And increases their earnings per share.

Now you know why the only new physicians you will meet are not homegrown anymore. It costs 400k to go to medical school. Then you have 3-7 years of postgraduate training at 35k a year. Then your debt has ballooned to 800k with deferred interest payments. BTW, new physicians get paid 90-100k.

The new physicians making $90k a year after residency are not the ones with $400,000 in student loans, and $800k in total student interest load is... atypical, to say the least. $300-400k total load post-residency is much more - I refuse to say "reasonable," but much more realistic. And the folks who have $400k total load are on a fast track to make $250-300k/yr just a few years after residency. A huge part of why doctors are paid as well as they are is because of the risk and the massive opportunity cost of getting an MD and going into practice.

Yes, it takes planning and dedication to set yourself up on solid financial footing post-medical school, and it's entirely conceivable to be in your early thirties before you're debt free. But it's pretty damn hard to be struggling financially as a physician in any field.

there is no 'reason' as to why physicians are paid what they are paid. It's simply supply and demand. A typical physician has to care for 5000 patients a year (that is the size of a full panel, btw) and gets 150k a year. Do the math - 20 patients a day, 5 days a week for 50 weeks a year is 5000 new visits. BTW, phone calls are not billable hourz. Oh, and I forgot 'emergencies' like aunt jemima who has severe burning when she pees, etc and needs to be seen ASAP or else she's going to the ER. Now add in the follow ups and emergencies and you're up to 35 patients a day. If you work for 8 hours (no lunch or breaks) that's 4 patients an hour. But wait, you have to enter patient info into an emr while you're supposed to be talking to the patient. Maybe you need to sip some water and go to the bathroom too. So that's 10 min a patient. for 150k a year.

Instead, most MDs are looking to become 'boutique' docs. Each patient pays $2k a year for as much primary care as he/she wants. The boutique doc limits his panel to 500 patients. He makes 6x the money for 1/10 the work. now the doctor can spend an hour with each patient and treat him like a human being instead of a slab of meat. And cut the expensive insurance middleman out of the picture.

So right, who needs the stress and bs that comes with a corporate existence. Why should the people who do the work (programmers and coders in your case perhaps) support the wasted busy work of people who are corporate drones and never actually interface with customers/patients/clients/aka the people who pay the money.

Not a dark UX, but another move I’ve seen that is line with the stall-for-time method that the author mentions is the requirement for a prior authorization, even for very common medicines. I have seen first-hand the amount of work this tactic creates in medical offices. It definitely allows insurance companies to stall in their approval/payment for medicines and, I’m sure in a % of cases, not pay at all (ie, the paperwork doesn’t get completed, patient gives up trying to navigate the complexity of the system).

My most hated is "The thing your doctor did was not medically necessary, therefore we won't cover it."

Never attribute to maliciousness that which can be explained with incompetence.

Since two of the three are complete speculation, without any evidence, I'll address the first example.

"suppress empty div" could mean that it is hiding an element until a condition in the form is met. You need to see the accompanying JS to understand any intent. It's entirely possible that bad programming is keeping the div both empty and suppressed.

"Hanlon's Razor" for those interested: https://en.wikipedia.org/wiki/Hanlon%27s_razor

Even if the exact issues aren't by design, I have absolutely no doubt that the competent person that hired incompetent programmers to ensure policy holders have no idea what they're signing up for or how to make a claim was very much a premeditated malicious maneuver. Bonuses for them are cheap workers and scapegoats for when they piss off the wrong people.

Having worked in equally despised and bureaucratic industries, and with HIPAA, I can assure you that the chances of that are very slim. Purposeful blocking of claim submission would mean millions of dollars in fines for an insurance company. Nobody is actively trying to bring that upon their department.

Bureaucratic systems like health insurance are about preserving your position as having the most benefit with the least amount of responsibility. They are ecosystems of mediocrity.

A VP at an insurance company generally cares about two things, balance sheets that make him/her look like they're doing a good job and tomorrow's tee time.

Directors care about their line on the balance sheet pleasing the VP. That's how you get most insurance company software written by low-wage, visa workers, who could care less about code quality or maintainability, because they'll be at a new gig within a year or two.

But developer incompetence and poor product design can be a 'dark' strategy as well, just a passive one instead of an active one. Good UX doesn't matter because users are not customers to health insurers, they are just cost centers. Their employers are the customers, and those guys care about how cheaply they can buy insurance and retain the quality level of their desired employee base. If the goal is to sell lots of insurance and pay little in claims, the strategy makes sense. So you spend as little as possible on UX.

NB: I think this is also how enterprise tech works.

Insurance companies don't think of "members" as just cost centers. They're a lot more picky around the concept of "hiring and firing" customers.

Put it in the perspective of a freelance web developer. If someone is going to pay you x amount of dollars to build a website, but cost you more than x in support time and change requests, you either have to up the price or not take on the customer.

"Never"?

When doing something malicious, use incompetence as your cover.

Nah, you just attribute it to "miscommunication".

That is hopelessly naive. When someone "accidentally" does something that furthers their own best interests, it usually isn't really an accident.

Let's forget ethics for a moment, is there any decent way to calculate the difference in bottom line between creating and providing a great service (presumably leading to a good reputation, and maybe better, more long-lasting relationships with customers), and squeezing and weaseling every penny out of your customers and doing everything in your power to keep them from using the service they bought (presumably leading to bad reputation, but more cash upfront)?

In some industries/companies/communities/what-have-you, there seems to be this consensus that exploitation, scamming and backstabbing is the obvious way to make the big bucks. Is this objectively true, or is it just some kind of oral history and tradition handed down from 1800s industrialists?

One thing they could do is assign a score to each insuree (and I guess they already do something like this) based on the number of past claims and likelihood there will be more.

Depending on that score they present more or less dark UI patterns to that user. Past a certain threshold they probably don't want to keep the insuree anyway, so if they switch it's still good for the insurer. They can calculate the likelihood the user will switch and how much the insurer will lose in premium, versus how much they'll win if they delay claims as much as possible.

I'm pretty sure "vary your evilness based on how much you think someone's worth to you" is an evil pattern. :P

That may be, but it's still incredibly common. Look up the concept of LTV.

I have personal experience in customer service scenarios where people classed by 'the computer' with an LTV score below a certain value were not to be retained if they wanted to cancel our service (c.f. for a higher level LTV customer we might offer a promotion or something) and people with a score even lower were to be actively encouraged to leave if they reported any problems - even to the point where we were supposed to suggest they try somebody else rather than spending any time fixing their problems.

Granted, these were (allegedly) people who were actively costing the company money in that it cost more to service their account than the revenue they provided, but it was an automated decision made by an algorithm that the phone and retail reps just had to enforce.

I think that's pretty heavily reliant on the rest of the industry. If it's an industry that no one feels very strongly about, there's probably something to be said for hanging out in the middle: some scummy shit, but no nickle and diming. If it's something significant where every company is hated already, a new company coming in with nothing more than "we want nothing more than to provide a good service" is likely going to thrive.

That assumes the customer is the one who chooses. With most Americans getting insurance through their employers, they don't have the option to choose a less crappy insurer even if they trusted the claims of better service.

If you don't resort to the weaseling, then you lose out to the competition. Capitalism is a system which naturally selects for the most scammy actors to survive and thrive.

Your comment makes no sense. Do you as a customer seek out the most scammy companies to do business with?

Sometimes, there are only scammy businesses left, after they have out-competed any alternatives through dirty tactics that their competitors refused to resort to.

But then the non-scammy should have a business opportunity in breaking the status quo.

I would question the incentives for an insurance company to reject claims arbitrarily or to not make the process for submitted claims at the very least adequate for two reasons:

1. If you insure a bunch of people against, in short, stuff that can kill them, then it's really in your best financial interests to be making sure that as many of those people as possible are seeing a doctor as regularly as possible (within reason) because that's how you avoid situations where someone makes shitty healthcare decisions for 3 decades then suddenly needs every imaginable treatment for every imaginable ailment that said person, who is obviously not a healthcare professional, would have never seen coming because they weren't regularly attending to their health in a low cost consistent kind of way, instead they just let their health deteriorate until it was an emergency. Which, from my understanding, especially in America, is where the vultures pounce to ensure your insurance companies pays top dollar for critical drugs and treatment.

2. On a purely accounting note, wouldn't this just fuck up your cash flow figures? I mean, if someone needs to get their insurer to pay for something because they don't have another option and they try and claim for that thing, isn't it in your best interest to make sure you pay out the amount at least reasonably quickly because, especially if you're acting in bad faith, their going to eventually break their way through the system to the "submit claim button" through sheer hours of trying, or their going to end up in court because they don't have other options. Then you've suddenly got a massive lump sum payout that probably wasn't in your forecasted expenses. To me, that just sounds like unpredictable silliness.

But then again, I live in a country with free health insurance so I genuinely could be entirely missing the mark here and it could genuinely be more profitable to pinch your pennies as an insurer.

It doesn't necessarily follow that the most profitable path will be taken by middle managers, they often have perverse incentives, real or imagined, to cut costs short term even when the company would be better off if the insured stay healthy!

I think for most insurers it's more profitable to provide less service and preventative care for two specific reasons. One, once a person is 65 they are covered by Medicare and are the government's problem. Two, if someone under 65 does get seriously sick or injured, chances are they're going to have to leave their job or get fired. With most people getting insurance through work, the insurance company isn't on the hook anymore either.

The ACA required most health plans to provide free preventive health services.

https://www.healthcare.gov/coverage/preventive-care-benefits...

Meanwhile, these insurance companies are making billions of dollars of profit (https://www.axios.com/profits-are-booming-at-health-insuranc...). There's no reason insurance companies need to be involved in health care at all.

> There's no reason insurance companies need to be involved in health care at all.

As a Brit I sort of agree, because I think free healthcare at point of delivery is the right solution. However the more continental model works perfectly well and many countries use private insurers for that sort of model. Insurance companies have a lot of expertise in risk management, analysis and costing so to claim that they've no place in healthcare is odd.

Getting far afield here, but... there's something uniquely broken about the financial model of the US system. Pretty much every industrialized nation in the world pays about half as much per capita as the US does for its health care, and pretty much all of them provide universal coverage. And this is everyone, from purely socialist systems to mostly private systems.

Which is why I have issues with both the right-wing insistence that nothing is wrong with ours that a little more free market can't cure, and the left-wing insistence that only a purely socialist single payer system will provide financial responsibility and social justice.

Billions of profit out of trillions spent on health care. Insurance company profits are not the real problem with health care costs.

Insurance profits seems to make up only about 1% of our total private healthcare expenditures. However, that doesn't take into account the amount of those healthcare expenditures that are being spent to generate those profits rather than to provide healthcare. It also doesn't account for the additional costs to non-insurers from systemic problems created and maintained insurers to increase their profits. (E.G. pushing filing/approval costs onto physician's offices or systemically fixing prices).

So while Insurance company profits are not the direct cause of the outlandish healthcare costs in the US, they are a huge contributor to our problem and a core driver for many of it's multitudinous factors.

> There's no reason insurance companies need to be involved in health care at all.

There are a very many good reasons to have insurance for health care expenses, in fact they are the perfect example of where insuring for the average cost helps those who see orders of magnitude larger costs due to black swan events that would otherwise ruin them. Another interesting stat showing how important health insurance is; 5% of policyholders are responsible for 95% of total costs.

Now you can argue that you think a national, government run, single payer insurance company would be more efficient, or more fair, or result in better health outcomes, or... but that’s an argument for a different kind of insurance company. It doesn’t change the fact that there is an insurance company very directly and intimately involved in your health care.

> Now you can argue that you think a national, government run, single payer insurance company would be more efficient, or more fair, or result in better health outcomes, or... but that’s an argument for a different kind of insurance company.

Anybody arguing for that as an alternative to our current system is... uninformed.

We already have it, and we spend more per capital on (non-universal) public healthcare via that government run insurance than many countries with public universal health care (including Canada, and the UK).

Obviously it is not our lack of "publicly run health insurance" that is the problem here.

My spidey sense in law has to ask myself, "I wonder if that's HIPAA compliant" lol. Should just call it the Dark HIPAA requirements and un-UX in Health Insurance.

Laughing aside, the real data pinch in HI is the meta data of the meta data. Once you understand the abstracted layer there, the rest is following EDI compliance contracts and creating views appropriate to the end user.

It's a slippery slope.

I worked with health insurance as a software developer (in Brazil so maybe it's all different, but still) and here are my observations.

1. Making claim submission impossible

He said he tried to submit a claim at the last minute. I believe the hidden page was just lazy work, because the page has to be visible again soon. If they hid the page at an appropiate time I don't know.

2. Routinely denying claim submissions

As far as I can tell, in Brazil a regulated health insurance plan can't deny you service based on how many claims of a certain kind you submited. They could before.

But the industry will find other ways to not lose money. Readjust prices based on number claims (from all users), prevent you from entering the plan and making a claim right away, pay for a quota of every procedure you take.

All these mechanisms already existed, but if they can't deny claims anymore they will tweak the other points.

3. Saying the customer has other insurance

This is fucked up. I can't find a explanation that isn't malicious.

Here is one similar case that happens in Brazil though. If you have health insurance and use the free health insurance (SUS). The Union will charge back your plan, not you. It's their job to sort it out not yours.

comparing brazil with the US on health insurance and worker rigths is a joke, you know that, rigth?

even with the latest illegitimate corrupt government killing most of the work laws last couple months, its still way ahead of the US.

I know but I still think the points 1 and 2 are pretty much the same for both countries.

United Healthcare... no surprised!

beat's post is correct. This is straight up incompetence and just another symptom of how bad things are in Health Care Software right now. To give more background I'll tell you that I work in a company that builds software for major state-wide US insurance companies. Our software touches hundreds of thousands of lives on a daily basis.

The problems you're seeing may not even be a part of our software even though our software will do the heavy lifting. Each Health Care customer hosts their own product, although in recent times this is changing thank god. This means they maintain their infrastructure apply patches, manage their network, do their own provisioning, etc, etc.

This also means if a manager has the bright idea to $save money$ and build their own in house customer facing claim submission portal we don't know about it. How can we open our software up to work with their shitty portal if we don't even know it exists?

It may be harsh to assume the portal is shitty and has bugs like you experienced in item 1 but it really is a safe guess. Most of these US based companies are staffed by workers in India. These workers in India have a long list of obstructions preventing them for doing their job properly: a) 8am on the US Eastern seaboard is 6:30pm in Noida, India. So they're all basically working the night shift.

b) Lack of strong labor laws or fair compensation. High turnover in a critical IT field is bad, bad, bad. I've gotten calls from people in India who are trying to find out who their manager is in the US. I don't know bro I'm just a vendor :(.

c) A handful of non-technical, non-healthcare business folks managing a team of people on the other side of the world is a recipe for pain and failure. Here's a great example of the mysterious cause behind "Rising Medical Care Costs".

d) Lack of training and proper communication leads to horrible awful undocumented software abominations. I am borderline willing to blame my balding on these troubleshooting calls.

What I'm trying to make clear is that the failure may be between some shitty interface Insurance Co. #1234 built to save money. If that's the case it'll never really get fixed as the people managing it may not even be aware of the bug. If all goes well and the interface to our Insurance software is actually correct then "Hooray!" your customer facing bug gets to someone like me!

Now we've entered the realm of a US based software company. But our service that processes this information is 12 years old, the ONE guy who built it in the 90s is long gone, and there's no documentation. So good luck getting someone to take ownership of that. That's assuming I can even reproduce the bug. Reproducing the bug may be an odyssey wholly unto itself. It might take 3 developers and 1 QA engineer that has 20 years of seniority to figure out why logging isn't working for this 12 year old service.

So that's your dark pattern. It is more of business pattern but its a pattern that has so far caught every god damn piece of Health Care software in its net.

Reply on news.ycombinator.com