Proposal: Psychiatric Crisis Bonds
A design for a new kind of financial instrument in the mental healthcare industry to make paying for someone to recover from psychosis *profitable*
(This is an idea I had last November, but am only publishing now. Accordingly, some of my views have shifted since I wrote it, and it certainly could use improvements! Posting here as a way of starting a conversation, and also documenting the changes in my thinking.)
After someone experiences a first episode of psychosis, their expected lifetime treatment costs massively increase through subsequent hospitalizations, disability costs, and direct treatment costs. Helping them recover well and prevent future episodes can both save a lot of money in lifetime treatment costs, and massively improve their future quality of life.
However, none of the current payers are incentivised to make sure this recovery effort happens, and the patients themselves are generally inexperienced and simultaneously suffering from executive function problems related to the psychosis itself. Currently, public insurers like Medicaid are the payers of last resort who end up footing the bill for treatment of long-term mental health issues after a first episode of psychosis, and Medicaid’s cost structure lacks an incentive to resolve this problem for individual patients.
We propose creating a new financial instrument, a bond issued upon evidence of a person’s first episode of psychosis, which are then sold to a bond-buyer who agrees to pay all the person’s mental healthcare costs for a fixed period of several years, in exchange for an income stream equal to what Medicaid (or equivalent) estimates it would have paid in treatment costs for that person over the same period. The bond-buyer’s profit is equal to how much less they can pay over that time period than the income stream. This justifies making careful up-front investments and idiosyncratic, non-medical investments, that the bond-buyer expects will minimise the cost of expensive psychiatric re-hospitalizations over the lifetime of the bond.
The Problem:
Paying for the treatment costs of people who experience serious mental health challenges is a perfect storm of terrible incentives:
Currently, most providers are still paid based on the ‘units of treatment’ they provide, and so are incentivised to treat more rather than less.
Private insurers are paid based on the difference between what they charge in premiums versus what they pay for treatments, so they are incentivised to continually minimise the amount of treatment they pay for.
Public insurers like Medicaid are forced to play by rules that treat everyone as equally as possible, patients and providers, leading them to pay for lots of small portions of low-quality (or at least low-paid) treatment. Also, administrators are often rewarded for managing programs with large budgets, which leads them to try to increase their budgets.
In many serious mental health crises, the person loses a lot of their ability to plan, organise and execute complex plans, and their motivation to do things that will resolve the mental health crisis. So as it gets harder to access treatment, the people with the worst executive function challenges are the ones who stop being able to get good treatment first.
Sometimes, there is a family member who has the motivation and ability to pursue good treatment on the patient’s behalf, but this person is unpaid, unfamiliar with the system, and often experiencing burnout with little support themselves.
From patients and providers there is strong political pressure to increase the amount of treatment provided, and none to decrease it, and very little to improve the quality.
If what you are purchasing with ‘treatment’ is ‘some chance that this condition will improve’ then that second factor is mostly hidden from most purchasers of treatment, whether they are patients, insurers, or public insurers like Medicaid.
(Note–these incentives are definitely present in the much-maligned US healthcare system, but they are all at least somewhat present in the other healthcare system I’m familiar with, the Australian one.)
This is a (very humble, tentative) proposal for a project that aims to point some or all of these incentives in the right direction–towards the patient becoming healthier, more resilient, needing less ‘treatment’, and staying alive.
We know that someone who has a first episode of, e.g. psychosis has some chance of:
Future A: recovering from that episode and never having one again
Future B: having more episodes, in a somewhat predictable pattern
Future C: experiencing a serious spiral into homelessness, serious drug addiction, jail time etc
Future D: dying, of suicide
In terms of how much they suck for the person experiencing them, they rank:
(Sucks least) A < B < C < D (Sucks most)
In terms of healthcare and disability costs, they rank:
(Least expensive) A < D < B < C (Most expensive)
So, the first thing you notice is that for anyone whose primary concern is minimising healthcare costs, preventing death isn’t a super good way to succeed. Now, most people have inbuilt values that stop them from actively killing someone to earn money, but it’s common that the people who are aiming to maximise profits here simply don’t know how whatever choices they make impact the death rate, and changes in the death rate are sort of an unfortunate side-effect of attempting to maximise profit.
The second thing you notice is that aside from that fuckery, the preferences of the person having mental health challenges and the person paying for their care actually line up! So, is there a person who is trying to minimise healthcare costs?
In aggregate, kinda. On an individual level, no. And here is where those preferences between the payer and the patient start to diverge. Even in situations where there is someone sort of central whose job it is to minimise healthcare costs (like, say, someone running the healthcare department of an entire county, whose job it is to provide healthcare for everyone who happens to live inside their district’s boundaries), they have no specific incentive to minimise the cost for any individual person.
So they may be happy to hang out in a world where 90% of their population is chill and barely costs anything to look after at all, and 10% have occasional monster healthcare costs. They have so much to pay attention to that the need to ‘reduce costs specifically for someone who is at risk of spiralling into a really bad life situation or dying’ is only one of many things they care about.
This also doesn’t make sense for a private insurer, because it’s almost always easier for a private insurer to just wait until the person loses their insurance because they lost their job than to try and stop the spiral in the first place. Mostly, by the time someone is in a situation where their psychiatric care costs are that fucking expensive for whoever is paying for them, they no longer can afford health insurance anyways.
Also, it’s extremely illegal for an insurer to make insurance more expensive for someone once they get a worse prognosis (thankfully).
(I’m only familiar with the insurance systems in the US and Australia and this case is roughly true in both countries–in the US if someone loses their job they generally lose their insurance, meaning they’re then either paying out of pocket or on Medicaid, and in Australia, going to a public psych ward is paid for by Medicaid for anyone, and if you want to go to a nicer private place you have to have private insurance or pay a fuckton anyways).
There are people who run programs with the explicit aim of stopping this spiral–they are called early psychosis programs–but the program directors and employees get paid based on how much ‘treatment’ they give and not based on their success at preventing a spiral at all. These programs are still much better than not having them, in part because therapists et al get intrinsic rewards from helping people, but there’s no direct financial incentive here.
So, we’re within reach of a world where whoever pays is working towards the same future as the person with the mental health challenges wants, but we’re not quite there. How do we get there?
Even though there are things that could help prevent a psychotic episode, it’s extremely hard to find the people who are more susceptible to such crises, and expensive to give good-enough preventative care to everyone if the crisis will only happen to a minority of people without the prevention. Using weed heavily every day is a big contributor to psychosis for the people who are already susceptible, but most people aren’t susceptible, so persuading everyone to reduce their weed use in the hopes of reaching the people it will help doesn’t have great ROI. And yes, this is a good, compassionate thing to do, but we’re looking for an incentive that doesn’t only rely on compassion (which gets weaker the less connected you are to the relevant person).
So, although as a society we generally want people to get preventative help, there isn’t anyone in particular who profits from providing it.
However! Once someone has given some strong signal that they’re in the group of people who might either B) need frequent care for the rest of their lives, or C) need super intensive, random use of expensive services because they have spiralled into an awful, precarious situation, suddenly preventing those outcomes is a much surer investment.
Some governments already do something similar–the county of Alameda did an 80/20 and identified the 800 people who were using like 50% of their emergency services capacity, and then made a separate program to help them that both probably makes their lives better and saves the county money compared to them ending up in the ER 3 days a week. But, this is identifying people who are mostly already in situation C, AND, it’s doing so in a place where there already happen to be tons of people in situation C so it makes financial sense to do something specific for them because that will save the county enough money to justify the effort of creating the program. Plus, it seems to be pretty hard to get people OUT of situation C, particularly compared to preventing them from falling into it in the first place.
But ending up in a psych ward for the first time for any reason, but particularly psychosis, is a pretty good signal that you have a higher chance of ending up in situations B, C or D. I don’t know how much your chance increases compared to the rest of the population, but this data does exist–you could take the healthcare data for some group of people who had had one hospitalisation and compare it to the data for similar people with no hospitalisation. This study does something similar and concludes that if you know that someone with a schizophrenia diagnosis ended up in hospital for a suicide attempt recently you can predict that their healthcare costs will be $20K more per year in the near future than someone with the diagnosis but no recently crisis hospitalisation.
So, the proposal:
If someone ends up in a psych ward with some sort of psychosis-related problem for the first time, someone (probably the government, but perhaps a private organisation) creates a bond that goes up for sale. The bond will pay out (say, per year) an amount slightly less than the government would expect to pay in healthcare costs for this person, by comparing their basic data to the basic data of other people the government has treated. So, say they had a similar life situation to some other people who on average had $100K over ten years in healthcare costs paid for by the government, then the bond would pay out slightly less than $100K over ten years.
Then, companies can buy these bonds. When they buy them, the deal is:
They pay for all mental health and related disability costs for this person for the duration of the bond
If the person dies, they forfeit some or (probably all) of the bond value to the person’s family (a bit like life insurance)
They’re allowed to pay for anything legal to help the person recover
They earn the bond’s value over its lifetime (so, the amount the government expects it would have paid to treat the person otherwise)
At the end of the ten-year bond period, the bond must be offered up for sale to any other parties who want to buy it (which could, theoretically, include the patient themselves, but might include any other similar companies). The original seller (the government in this case) uses the information from the ten years of healthcare costs to re-price the bond. If the company was successful at beating the government’s prediction, the price would go down to reflect that. If it was unsuccessful, the price would go up, making it more lucrative for anyone who thought they could help the person recover or get sufficient treatment more cheaply.
You could imagine that at some point, the person is cheap enough to look after that the bond just isn’t that attractive as an investment, so maybe the government buys it back.
My belief is that this short-term profit-based competition will over time lead to the discovery of more and more ways to help people recover quickly and cheaply and thus to the decrease in the prices of the bonds themselves.
Risks and structural considerations:
This set-up should create actual pressure towards recovery and not get goodharted sideways in any *extremely* obvious ways, and there are adversarial goodharting risks that I haven’t satisfactorily accounted for in this short, untested explanation. Some of them:
The bond-buying companies find ways to get out of paying for expenses that are legitimate (like hospitalisations) and so the strategy of deflecting expenses becomes more profitable than the strategy of working on the person’s recovery.
In a worse extension of that, it becomes more profitable to assert that the person died by non-psychiatric causes if they died, than to forfeit the remainder of the bond legitimately.
The bond-buying companies inherit a culture that lacks creativity from e.g. modern insurance companies, and limit themselves to a small set of treatment strategies for cultural and not financial reasons.
The bond-buying companies find it more profitable to inflate the treatment costs in a way that increases the price of the bond at the end of the bond period than to take profits during the bond period.
The bond-buying companies find it more profitable to keep people involuntarily on high doses of drugs that massively tank their quality of life but keep them alive and satisfying the criteria that let the bond-buying company receive the bond, rather than helping them recover to become healthy and resilient.
One great thing about this set-up is that unlike clinical services, you don’t need to implement them at a big-enough scale to justify e.g. expensive clinicians or expensive new technologies whose costs are spread across a large number of patients. You could start by creating bonds just for the highest-expected-cost new patients, or even for patients with the financial resources to fund them privately, and expand to include new people gradually.
However, it does mean that the bond-buyers can get returns to scale on solving non-service-based problems for a large number of patients simultaneously, such as using automated user interfaces to make booking appointments easier. They also get returns to scale on negotiation, such as negotiating on-demand service agreements with psychiatrists. In this way their constraints are similar to a provider-insurer like Kaiser Permanente, but with a focus on a specific mental health issue.
I imagine that the companies that would want to buy these bonds would mostly be normal insurance companies, but there is no reason they couldn’t be the businesses of individual psychiatrists, for example, or mental health clinics. There also isn’t anything constraining what kind of business buys the bonds except whatever is legally allowed and what is acceptable to the patient.
One of the advantages of this system is that, instead of the patient or provider having to justify an expenditure as ‘healthcare related’, the bond-buying company can pay for anything that will reduce their potential healthcare costs, on an individual level. So, if someone is at risk of homelessness, maybe paying for a month of rent is going to be cheaper than the potential psychiatric hospitalisation that might result.
Also, they don’t need to justify their expenditures to providers based on the licensing of the provider, but rather, spend their budget the way they think will be most cost-effective. So if they can keep someone out of hospital in a high-stress period by paying a grandma to go check on them three times a week they can do that, rather than having to choose between a limited set of licensed-provider options. Probably, someone at the organisation would need to take legal responsibility for making medical decisions, so a psychiatrist would need to be employed, but there wouldn’t need to be many of them, and if they were taking a share in the company’s profits, they wouldn’t need to prioritise maximising their billable hours.
(A limited subset of) questions that need to be worked out in more detail:
How is the patient involved in choosing (if at all) which organisation buys their bond? What happens if they don’t want one to be issued at all, or don’t understand finance enough to consent to it?
How would the money be paid out over the lifetime of the bond? Should it be split equally over the years, or track the expected change of the costs it is based on, or weighted to provide a large payout at the end?
How can this design be used for people with e.g. suicidal depression, disordered personality diagnoses, and what needs to be adapted?
The role of a dedicated nonprofit in running the scheme:
define the rules of the market, including helping to define data use practices for pricing the bonds
jumpstart the market by taking on some initial risk of buying bonds from jurisdictions and selling them to buyers
Influence the culture of bond-buyers through the selection of initial customers to prioritise adaptability and respect for patients
expand the market by recruiting new bond-buyers, payer jurisdictions and patients
protect the instrument from goodharting with ombudsman efforts, design changes, and advocating for legal protections
protect the transparency of the market
encourage an ecosystem of supportive tools such as analytics software, new products to help fill gaps in need, education to spread strategies
Expanding the scope of the bond type to cover people with a variety of serious mental health challenges
Why this might work:
Some reasons to suspect this financial instrument might work to prevent people who have a single psychiatric crisis from experiencing a spiral that ends in some kind of permanent disability:
Our current popular conception of psychiatric crises is as illnesses which, once diagnosed, follow a fixed path throughout someone’s life, which treatment mitigates in a sort of linear way. If you assume this, the most cost-effective thing to do might be to prescribe treatments that mitigate symptoms or resolve issues in a dose dependent way–big problem, big treatment, and avoid going ‘all in’ to resolve smaller problems early.
However, the best psychiatrists, successfully recovered patients, and many academics agree that one initial psychotic crisis only creates a small direct increase in the chance of a crisis occurring again, and that much more of the suffering and later mental health challenge that occurs after a crisis occurs because their temporary disability from the initial crisis causes a cascade of practical problems or losses of trust in their life that become too overwhelming to handle and create a permanent loss of capacity which in turn makes further crises more likely.
This model implies that the response to the first crisis, both allowing enough time for recovery, and creating resilience in the person’s life to protect against damage from future crises, has an outsized importance when it comes to helping someone at risk of serious psychiatric disability avoid that outcome.
If this model works for people who experience an episode of psychosis, it could potentially be expanded to people who make a suicide attempt, or have an anxiety- or drug-use-based crisis that leads to hospitalization.
Size of the problem (for psychosis only):
3% of people in the US experience at least one psychotic episode throughout their lives (almost 10 million people currently alive)
Every year about 100,000 young people experience their first psychotic episode
Estimates of the additional lifetime cost of Serious Mental Illness (a category of mental illness that most people who have recurrent psychotic episodes would fall into) is $1.8 million per patient (although this mostly emphasises lost earnings).
What’s your take? Who do you think should read this? How could it be improved? How could it get to e.g. Medicaid administrators? Questions and critiques always appreciated.
Some thoughts and questions:
1. A lot of the potential goodhart stuff like arguing that hospital stay isn't justified (and paying doctors to make this point for them) happens already. It is demonstrably a very cost-effective thing for insurance companies and HMOs to do. If the interventions don't work well, aren't a lot of these systems going to just continue the tradition of trying not to pay for stuff whenever possible?
2. Seems like a bond isn't that different from just getting premiums and from and paying for a patient that you can't just drop when they become a poor investment. Are bonds going to change things enough that now it makes sense to invest a lot in R&D to test and implement exactly what works, vs. just be as big a cheapskate you can get away with?
3. Not all first episode psychosis the same. A high strung executive who has an income and a supportive spouse but has an episode of mania while dealing with insomnia is probably going to cost much less than the kid who develops severe psychosis before age 15. How do you avoid the problem that various people will have bonds of clearly different value? ACA already means you can't refuse for poor prognosis, but if the whole point is abandoning public insurer fairness requirements wouldn't fairness be a problem? And presumably if insurers or other buyers refuse the bad investment, then medicaid is just stuck with the bills for the tough cases anyway.
You could go for a kind of auction system which could be interesting, but might take a lot to get off the ground. If the solution is a big company buying bonds for a bunch of patients together, that's already basically an HMO or managed care medicaid. Then the bet is basically can a company do better than medicaid at managing these patients. This is basically medicare advantage, right?
4. Medicaid footing the bill for almost all the severe patients in the end means that they already capitated and have incentives to minimize costs and do so through a number of instruments like paying caregivers, paying social workers for house calls and home check-ins, PHPs after a crisis and wrap-around services for heavy utilizers. They seem to be trying a lot of things, why not just pitch any effective and cost-saving measures directly to them? The fee-for-service isn't a great model especially for hospitalization, but the payments are already low compared to other areas of medicine and it's not like the mental health industry needs to create artificial demand except in edge cases, there is plenty as is!
5. Physician experts and legal issues. Are these bonds supposed to get exempt from existing standards of care and that's how they save costs? Presumably any new company would have to provide the same resources as available existing ones and failure to follow the expert recommendation could cost them in court in case of a negative outcome. Forfeiting all the bond income in case of death is an incentive to keep patients healthy but would place a lot of risk on the bondholder and wouldn't save you from getting sued regardless. Then it's trying to change the whole system again, and because it's hard to tell what works here are going to be a lot more hurdles in your way including professional organizations.
6. If no normal standards of care, what is the replacement? Since from what I was reading this plan looks to save on cost and preserve life, but doesn't specify anything else. For some patients, it would undoubtably be cost-saving and life-preserving to be very paternalistic and involuntary institutionalize. Once institutionalized, you wouldn't have any incentives to keep the experience nice beyond whatever prevents the next hospitalization. Are we okay with going back to this? What about other kinds of potentially coercive stuff like offering money to patients or family members to eg get on medications or off substances and all the problems that that creates? And even an org eg paying for rent could be incentive for patients to purposely get themselves into this system similar how people game the disability system or VA.
I think a lot of the reason you don't see more change is:
1. it's really hard to tell what works and most interventions don't work great
2. A lot of the stuff that doesn't work well is there because of law
3. The things that may work well may only work in certain circumstance and often very difficult to scale cost-efficiently
4. The patients are broke
5. the systems are basically broke too and often short-staffed and undertrained
So often we end up defaulting to standard of care and then don't fund it well because it doesn't work very well so why would we?
National health systems in other countries, The VA, medicaid, and big HMOs the US are already capitated and incentivized to bring down costs while ensuring customer health and satisfaction while working within the standards of care. They are already doing a lot more than most people probably realize to reduce hospitalizations and bring down costs. And because they are big healthcare systems with bigger budgets, more risk tolerance, longitudinal relationships, they are much better positioned to try and innovate than the smaller organizations. They do research and test the efficacy of their interventions, although I can see an argument that they are slowed down by their bloat and the administrative incentives you mentioned.
The aspect I like the best is getting people with FEP into a comprehensive care system right away. It's too easy for people to fall through the cracks and end up in scenario C more often than if they had great followup and more assistance at getting other help. Better surveillance for FEP and followup is likely something that a lot of these companies could save resources by investing in in the long run.
> However, the best psychiatrists, successfully recovered patients, and many academics agree that one initial psychotic crisis only creates a small direct increase in the chance of a crisis occurring again, and that much more of the suffering and later mental health challenge that occurs after a crisis occurs because their temporary disability from the initial crisis causes a cascade of practical problems or losses of trust in their life that become too overwhelming to handle and create a permanent loss of capacity which in turn makes further crises more likely.
> This model implies that the response to the first crisis, both allowing enough time for recovery, and creating resilience in the person’s life to protect against damage from future crises, has an outsized importance when it comes to helping someone at risk of serious psychiatric disability avoid that outcome.
Catching after the first crisis is the best, but there are lots of second best options too. Early psychosis programs have inconsistent eligibility criteria across states and countries. And the boundary is often discrete. If you are under 25 you are in, if you are over 25 you are out. Yet that person on the borderline is just as likely to benefit at say 27 or 30 and improve lifetime costs and outcomes.