How the U.S. Health Care System Works (And Doesn’t)
I interviewed 40 VCs, health tech entrepreneurs, health care executives, and industry experts and spent hours researching the intersections between health care, public health, and venture capital. What emerged was a multi-part report that I’m calling Preventing Prevention: Barriers to Venture Capital Investments in Upstream and Community-based Care.
Funding for this work came from the Robert Wood Johnson Foundation.
A System of Systems
This is both obvious and not: But our health care system is not so much a single system, but a system of systems.
From the publicly financed plans of Medicare (including the privately run Medicare Advantage plans) and Medicaid (with significant variances across and even between states) to the privately financed plans offered through employers or on the ACA-recently created Exchanges, there are many sub-markets segmented by geography, income levels, age, abilities, and more. Each particular insurance plan coupled with a set of providers (hospital, doctor, team, etc) is a separate thing — an individual product — to explore and understand.
This fragmented nature of our “system” makes making sweeping statements about how health care “works” (or doesn’t) difficult. But what we do know is that how a particular insurance plan pays for care matters. So let’s start there.
How Doctors Get Paid Matters: Retrospective vs Prospective Payment Models
So you walk into your doctor’s office. You pay a 20 dollar co-pay; that’s what you see. But your doctor gets money from your insurance company as well. How does your insurance company pay your doctor?
Turns out, there are a bunch of different ways. Be it private insurance given by an employer, Medicare, or Medicaid, there are a number of ways in which a hospital, physician, or provider group may get paid.
The following are a few different ways that money may flow through a system. Each payment model sets up its own game of economics, implicitly or explicitly providing incentives to do more or less of one thing or another. These are buckets, and there are certainly variations of these around the country, but the buckets do capture the big category of options.
Retrospective Payment: Fee for service (FFS)
The dominant way that health care is and always has been paid for in the U.S. is known as fee-for-service. This is straightforward and transactional: A provider — be it a doctor, nurse, care team, hospital, or clinic — does something (say they order a test) and then the insurance company pays them for doing that thing (ordering that test!).
This is simple in concept but enormously complicated in practice. There are tens of thousands of fees that a provider could bill for. And each fee, and even things not currently with a fee tied to it, can come up as a contentious policy discussion.
For example, Medicare infamously has not reimbursed providers when they just talk to their patients. This includes when they talk to them via simple technologies such as a video chat or texting. The inherent incentives here have an effect on the aggregate provider behavior: Surprise surprise! Providers don’t talk to their patients very much. On the largere business innovation side of the equation, this also dampens the market opportunity for tele-health companies. (Note: CMS updates their fee schedule every year, and actually, they recently proposed addressing this, and other technology-related fees, in their 2019 provider fee schedule.)
Note that in pure FFS, quality of the care is not factored in. It’s simply transactional: Did you do the thing? Great! Here’s money. So in FFS systems the incentives are to do more things regardless of quality. And more so, the costlier things: More dollars are made by doing more of the expensive treatments (e.g. surgery rather than physical therapy.)
The impact of these incentives are often discussed and somewhat debated, but general consensus in the policy community is that over-relying on pure FFS is problematic. It drives more of the expensive interventions without regard to whether or not the outcome was any good!
Finding an alternative way — or perhaps the various alternative ways — to pay for care is the big conversation among health policy nerds. These alternative ways of paying are conveniently called alternative payment models (APMs), and there are a few of them to talk about.
Retrospective Payment: Value-based payments (VBP)
Also known as Pay for Performance (P4P), value-based payments (VBP), like FFS, pays for each service retrospectively. However, VBP takes into account other factors that FFS doesn’t. (You may notice that I’m starting to use a lot of acronyms…)
These factors include: the quality of care given, how efficient the care was given, the patient and caregiver-centered experience, and how the cost of the care compares to historical trends.
Here, as the name of the model implies, money is made by improving the value of the service provided. Higher quality care given efficiently with favorable patient ratings will get a higher value-based payment than lower-quality care given inefficiently and with poor patient satisfaction remarks.
And theoretically there’s new business opportunity in helping hospitals figure out how to do this, how to increase the value of each service provided. I say “theoretically” because this is all fairly new and the algorithm that determines value is fairly complex.
To describe the complexity involved, here’s an example: One data source used by Medicare VBP that they make to hospitals is taken from a patient survey called the Hospital Consumer Assessment of health care Providers and Systems (HCAHPS) Survey. This survey:
“…contains 21 patient perspectives on care and patient rating items that encompass nine key topics: communication with doctors, communication with nurses, responsiveness of hospital staff, pain management, communication about medicines, discharge information, cleanliness of the hospital environment, quietness of the hospital environment, and transition of care. The survey also includes four screener questions and seven demographic items, which are used for adjusting the mix of patients across hospitals and for analytical purposes. The survey is 32 questions in length.”
These 32 questions are one of the sources of data used in this one VBP algorithm from one payer (Medicare) to one type of provider (hospitals). Again, incredibly complex.
Prospective Payments: Bundled payments
We can also move beyond the paradigm of retrospective payments (paying for the thing after the thing is done) into the world of prospective payments (paying for the thing before the thing is done).
There are many ways that a provider can be paid in-advance of the care being delivered. “Bundled payments” are lump sums given upfront to providers for a given health care “event”. For example, joint replacement surgery involves tens of small steps which in FFS would each get billed for. But these could all be rolled up into one “event”. A bundled payment for joint replacement surgery would take the expected cost of all of those individual bills and provide this in a single upfront payment.
The provider then conducts that joint replacement surgery; any funds left over they get to keep! So the financial incentive here is towards efficiency. And money is made by providing more effective and efficient care within that care deliver “event”.
The effect this has on care — whether or not a bundled payment “works” by reducing costs while improving care — actually depends on the event we’re talking about. Some bundles are more lucrative than others. There’s a lot of complex math that goes into determining, or trying to determine, what should be bundled and at what rate.
Prospective Payments: Capitation
Capitation is a bundle for a whole person for a whole year; a lump dollar amount is given to a provider to cover a given member for 12 months. So instead of paying a provider upfront before they treat a patient for a particular event, the provider gets an upfront payment to cover all events that are expected to occur for that patient over a given year.
And as with the bundled payments described above, money not spent by the end of the year is kept by the provider. (Some forms of capitation adjust for the expected costs of a patient, others don’t and instead just expect the laws of averages to work out.)
In capitated payment models, money is made by reducing the care being provided or, when care is provided, by providing the lower cost of the options. The “reducing the care being provided” part is scary for a lot of people who aren’t familiar with this space. But fear not, your doctor isn’t going to let you die (though if you think he might, well maybe you should find another doctor?).
Reducing care is best juxtaposed to the waste inherent in a FFS system. Capitated financing drives effective management of care. Managed care is a provider system that proactively oversees the care being give to a patient or group of patients in order to control costs. Care coordination, usually facilitated by information sharing across providers involved, helps reduce duplication inherent in FFS and helps manage a disease condition in its early stages from becoming worse over time and developing into an emergency. (Because sicker patients tend to cost more money.)
The risk inherent in a capitated system — and all systems carry some sort of risk — is that a patient may end up, because of an accident or sudden sickness, costing more than their capitated amount expected. This means that the cost of care given was more than allocated. But there is also the possibility of making money, if their patient group ends up healthier than expected. So for most providers operating in managed care, you play the averages: Some will likely cost you more, but some will likely cost you less. And while you can’t prevent traffic accidents and the sudden onset of sickness, you can be proactive in helping lower the risk of accidents and sickness occurring.
It’s worth noting that a capitated system carries with it strong incentives to cherry pick a population for the healthier patients: If you’re a small managed care plan and you don’t have the sick patients but only have the healthy ones, you’ll make more money! And this happens, of course, everywhere. Certainly on the aggregate as businesses, especially investor-backed business, are “supposed to” maximize profit. Read through the lens of the previous post in this series on How Venture Capital Works (And Doesn’t), a startup in this space is going to try to maximize their LTV / CAC ratio.
FFS → APMs
While there remains some debate over how much of a driver fee for service payment models have relative to other problems within the system (see: high administrative costs, high utilization prices), there is general consensus to move away from FFS towards alternative payment models (APMs).
These shifts are meant to drive success towards the three-part aim of:
- Lower costs
- Better care for individuals
- Better health for populations
The spectrum of APMs is often presented like this:
The arrows, importantly, have direction to them. CMS, across both the Obama and Trump Administrations, has worked to shift more and more financing of health care from the left towards the right.
This spectrum, as does the theory, suggests that those payment models and organization operating on the farthest right have the strongest incentives towards driving improved health across populations. These types of organizations tend to have the payer (insurance plan) and provider (doctors and hospitals) integrated. These types of integrated care providers are generally known as Accountable Care Organizations (ACOs), modeled off their sister org type: the Health Maintenance Organization (HMO).
As these organizations increase in number, there is greater demand for innovations in how these types of organizations manage their populations. And technology is a big player here.
THIS finally brings us to an important intersection of two big topics: Tech and Policy.
Where Tech & Policy Trends Merge: The Theoretical Future of Population Health Management
The trends in technology and health care financing give us one big area of growth: Population health management.
This came up repeatedly in the interviews behind this blog series: When it comes to trends in “technology and prevention”, investors, insurance plans, care providers, and entrepreneurs see opportunity in the data-fueled future of managed care.
Population health as a discipline has been around for awhile. Public health experts have long been (manually) mining the medical records of patients and clustering like-patients into sub-groups based upon typically-disease-specific characteristics. Interventions, such as a family-based weight management program, would then target this sub-group.
Technology is transforming this practice in a fairly predictable way. The goal becomes simple, and it mirrors the role of tech in other parts of our world: If an organization can gather and centralize more data in order to paint a detailed picture of individuals’ lives, that health care provider can get more sophisticated in how it provides care for that patient, knowing when to intervene and having a better idea as to what will help that patient and what doesn’t. So we enter the world of data…
The drive “understand” the patient
We live in an increasingly digitized world. Today, more and more data is able to be brought together to provide a more nuanced picture of an individual patient. Rather than a patient existing in a monolithic group (He is a white, male, healthy, in his mid-30s, non-smoker with mild allergies and asthma), this individual has its own digital fingerprint generated from data (He is Read G Holman).
This data is pulled from medical records within EHRs and electronic patient-reported histories. But data also comes from outside the health care system. This “outside” data can include data pulled from school records, credit scores, criminal justice facilities, environmental monitoring, and even the Census.
Increasingly, though still a significant challenge, patient-tracking data such as that from step-counters or digitally enabled asthma inhalers get incorporated. Further, the incorporation of genomic, proteomic, and other “omic” data is just over the horizon.
The other data source: What treatments work when and for whom.
Rudimentary systems may provide simple visualizations of this data, perhaps so as to hot-spot geographically where particular populations have a higher rate of a particular condition.
But more advanced systems, also gather data on clinical interventions from National Committee on Quality Assurance (NCQA), peer-reviewed journals, the American Medical Association (AMA), and elsewhere. The system then matches what it knows about a patient’s health with what it knows can be done to improve that patients health to suggest specific clinical interventions for particular individuals.
Algorithms then match patients to treatments (in theory)
And then the most advanced systems spit out a list of patients prioritized by which ones are most in need of intervention. Primary care providers, which currently just see patients in their office based upon who happens to be on their calendar and shows up, can better prioritize their work day by focused which patients to interact with on a day-to-day basis.
And from that: More proactive interventions with patients to prevent a condition from getting worse. Eventually perhaps preventing a disease from even occurring.
The growth in population health management is thus, paradoxically, away from populations and more toward customized, individualized, pro-active interactions.
Or so goes the theory.
The reality is this: The adoption of these population health management systems, of any form of this type of system, is as fragmented as our health care system. Many health systems have been in some form of the care management space for a while and are simply upgrading their systems. New care management systems are being created as APMs are adopted, new Medicare Advantage plans created, and new ACOs formed.
However, as the next sections layout, there are significant barriers that keep this ideal future of population health technologies, and prevention-oriented technologies more broadly, from becoming reality.
Next Up: FFS, Churn, and Clinical Efficacy: Forces Preventing Prevention in Health Care