HOWTO make health-care cheaper by spending more on patients who need it

Atul Gawande's New Yorker feature "The Hot Spotters" is a fascinating look at a small group of doctors and medical practitioners who are working on reducing systemic health care costs by doing data-analysis to locate the tiny numbers of chronically ill patients who consume vastly disproportionate resources because they aren't getting the care they need and so have to visit the emergency room very often (some go to the ER more than once a day!) and often end up with long ICU stays.

The approach is marvellous because it is both data-driven (data-mining is used to identify which patients aren't getting the care they need) and extremely compassionate ("super-utilizers" are voluntarily enrolled in programs where they get 24/7 guaranteed access to doctors, nurses and social workers). The programs are successful, and even though they cost a lot to administer, they still generate system-wide savings -- one patient helped with this sort of care had previously cost $3.5 million a year because of heavy ER and ICU use. In other words, providing excellent, personalized care to the small number of patients who don't fit the system's model saves far more money than making the system more stringent, with more paperwork, higher co-pays and other punitive measures. It's a win-win.

Except that it's not really catching on. Some of the doctors pioneering this approach are frustrated because they can save Medicare or an insurer millions, but they can't get funded by Medicare or the insurers -- instead, they have to fundraise from private foundations.

As he sorts through such stories, Gunn usually finds larger patterns, too. He told me about an analysis he had recently done for a big information-technology company on the East Coast. It provided health benefits to seven thousand employees and family members, and had forty million dollars in "spend." The firm had already raised the employees' insurance co-payments considerably, hoping to give employees a reason to think twice about unnecessary medical visits, tests, and procedures--make them have some "skin in the game," as they say. Indeed, almost every category of costly medical care went down: doctor visits, emergency-room and hospital visits, drug prescriptions. Yet employee health costs continued to rise--climbing almost ten per cent each year. The company was baffled.

Gunn's team took a look at the hot spots. The outliers, it turned out, were predominantly early retirees. Most had multiple chronic conditions--in particular, coronary-artery disease, asthma, and complex mental illness. One had badly worsening heart disease and diabetes, and medical bills over two years in excess of eighty thousand dollars. The man, dealing with higher co-payments on a fixed income, had cut back to filling only half his medication prescriptions for his high cholesterol and diabetes. He made few doctor visits. He avoided the E.R.--until a heart attack necessitated emergency surgery and left him disabled with chronic heart failure.

The higher co-payments had backfired, Gunn said. While medical costs for most employees flattened out, those for early retirees jumped seventeen per cent. The sickest patients became much more expensive because they put off care and prevention until it was too late.

Lower Costs and Better Care for Neediest Patients (via Kottke)

(Image: Emergency Room / Health Care, a Creative Commons Attribution (2.0) image from 41178161@N07's photostream)



  1. There is a certain irony to the fact that the work these guys are doing shares about 95% of its DNA with that of a well-baked “scientific eugenics” scheme… They’d probably be really unhappy to see their work take off in that context.

    1. Yes, taking a scientific approach to studying how to maximise our ability to help people is just like being a Nazi. Now maybe you can get on to discussing the actual content without bandying around meaningless insults by way of association fallacies.

      1. Did you fail reading comprehension? I specifically noted that they would be unhappy if their work were to be thus used. That’s approximately the opposite of attempting an association fallacy.

        I was, in total seriousness, noting the irony in the fact that, whether your motives are humanitarian or malevolent, the algorithmic exercise is one of optimizing the overall outcome by locating and addressing the outliers in the larger group.

        This is no more “insult by association” than is, say, noting structural homologies between sharks and dolphins, or the ritual similarities between two disparate religious groups.

      2. Ummm… the world, including the US, had a robust eugenics movement long before Nazis. If anything, their association only served to hasten the decline of the movement at that time.

        I think it’s about time communities have 24/7 medical clinics. Ideally they would partner with and be adjacent to the ER. They could triage for each other (maybe have one intake?) so the really sick people receive the emergency care they need while the flu case and withdrawing drug addicts can get the care they need. Just because you feel sick doesn’t mean you are dying.

        Here’s another rant. Our small town paper reports a police and fire log. This week (like most) all of fire calls were medical aid. Why do we need to haul out that kind of highly trained and specialized professional every time an elderly person is scared about their loss of mobility (yes I’m exaggerating here)? Mis-applying your life saving tools really drains your community.

      3. Nazis embraced eugenics, but so did many in the US, prior to the world wars… though not to the same degree, obviously.

        I think what phisrow is saying is that it’s a very short hop between “We can save lots of money by giving this subset of people more and better care,” a noble aspiration, to “We can save even more money by simply denying care to these people, unless they pay out of pocket,” a death sentence.

  2. Any kind of intelligent attempt an the part of the government to allocate care will be labeled a “death panel” by the right. (forget the capitalist insurance companies have real death panels)

    1. Any kind of intelligent attempt an the part of the government to allocate care will be labeled a “death panel” by the right.

      To be more precise:

      It’ll be labeled “death panel” by the corporatists in the American “health” and insurance industry. Then they’ll manufacture consent through their media outlets and it’ll be parroted by the mindless drones of the right.

  3. Wow, not many articles makes you want to become a statistician to help people…

    I wonder if you could make a huge difference in many other professions where philantropy is not usually practiced. Like a plumber or a bus driver… You just needed to find an way to help people, and you could turn your life into a force for good without even switching jobs!

  4. Preventing extremely high health care costs and improving the quality of the health/lives of people at this risk level? Yes, win win. Thanks for posting this, Cory. Also, nice reply jacobian.

  5. I fear that, no matter how successful and widespread this may become, any savings realized will simply translate to higher profit margins for insurance companies.

  6. I’m a little bit sceptic about whether this approach would scale. In other words, it may only be working now because the pilot program being run is small, and people haven’t understood the cost/benefit landscape for long enough to incorporate it into their actions.

    What if you rolled out this plan nationally, and a significant percentage of people who would not have otherwise been “super utilizers” began acting like that group so that they could get the disproportionately good healthcare?

    There’s a rule of thumb in agent-based modeling: “Bigger is Different”. It’s a rule meant to remind you how all sorts of social systems will change their behavior as they scale. I think this experiment is encouraging, but we can’t say it’s the answer based on a small pilot.

    1. Ubarch, you’re right that any good health care delivery/insurance model has to take self-interest into account (it’s a reason Massachusetts and, soon, the U.S. have a mandate to buy health insurance, because healthy people often choose to forgo insurance, despite the aggregate social cost of the 1-in-a-100 that get hurt and can’t pay).

      But in the case of super-utilizers, it’s tough to imagine (meaning, I could be wrong) that someone who spends 30 days a year in the hospital via emergency room visits wouldn’t want not to. Option 1 is bankrupting yourself (and your community), likely remaining in pain or discomfort, and sacrificing a good amount of dignity because you’re not in control of your life. Option 2 is paying less (and costing your community less), being in less pain and discomfort, and sacrificing less dignity by being visited by a dedicated doctor instead of showing up at the emergency room.

      Option 2, you’re right, probably changes as things scale up. But how wouldn’t that be someone’s preference?

      1. That (and phisrow’s) points are good ones… It may not be worth it to just decide to act like you need serious medical care as a precautionary measure. However, I can keep playing devil’s advocate. What if it’s worth it to your doctor to have as many “super utilizers” as possible under their care? Less than honest healthcare providers are incentivized to find as many of these extremely valuable individuals as they can through questionable medical decision making, and honest (but profit-driven) institutions may even base their businesses around it. What happens if you can make more money as a healthcare provider caring for three super-utilizers than you can caring for a hundred normal people?

        It’s easy to be sceptical like this in the face of what must be really hard work to solve a very hard problem. I’m not saying the work is invalid, or misguided, or not worth doing. My only point is that what these people have shown is promising but is still some very difficult milestones away from being a real candidate for an answer. I think this is especially true since this particular system involves metering out gigantic rewards to a select group of people– Something about that makes me think that it will somehow succumb to being a huge target for corruption or some other systemic failure mode.

    2. There is certainly uncertainty in scaling up; but I’d be surprised to see your scenario pan out:

      Most types of health care have an automatic limiting mechanism that throws a spanner in the econ101 “cost->0 demand->infinity” rule: They are better than what they are intended to cure; but they suck compared to just being healthy. All are time-consuming, many are painful or have unpleasant side-effects. Not a few are fairly risky.

      Aside from your hypochondriacs, Münchausen cases, or medical drug addicts(all three of which are in fact sick, just not in the way they immediately present themselves to be) or severely isolated elderly or mental health cases(who, again, could quite probably use some work) there just isn’t much incentive for healthy people to feign illness(aside from the occasional upgrading of a cold when calling in to work…). “Hey, if I feign serious illness and go to the ER N times, I get to skip the line on my N+1th instance! Hooray!”

      Plus, if this scheme is successful in keeping hardcore sickies from bouncing in and out of the ER, and either reduces their service requirements by making them healthier, or at least makes them more predictable by scheduling some of them, presumably ER waits and scheduling for low-frequency users will become incrementally less unpleasant…

    3. The “super utilizers” are people doing really harmful things to themselves, or they’re having harmful things happen to them for reasons mostly out of their control. Do you really think otherwise healthy, safe, and sane people will start harming themselves in order to get this “disproportionately good healthcare”? Multiple chronic illnesses and frequent emergency room visits in order to get a nurse to call and remind you to take your pills?

      Also, it’s not really an experiment, is it? You spend money taking care of everyone, especially the small number of really sick people, and it keeps them from getting as sick or having as many emergencies. Which brings the costs down for everyone. Pretty much everyone outside the US runs their system this way already.

    4. It doesn’t work like that. The highest cost patients are the ones who are admitted to the intensive care unit, who require dialysis, who have long stays on the ward, precisely because chronic problems that could’ve been managed in the community, weren’t.

      A good rule of thumb is that an admission to the hospital (at least where I work in Canada) costs the system $1000 if the patient goes to a bed on the ward. If they go to the ICU, it’s $10000 a day.

  7. The power of quarter-driven business: it’s vastly easier to justify spending $100,000 a quarter forever, than to spend $400,000 all at once, despite the fact that the latter option costs less over time, it costs more RIGHT NOW, and that’s bad.

    As for scaling problems- that really depends on the frequency of these “super-users” and how they’re distributed geographically.

    1. You sound like our software customers. They’d rather spend hundreds of thousands over the long run trying to get a cheap, freebie database to work than spend $50,000 or so all at once to get a license for a real database. The difference is that one shows up as a single capital expense that they have to jusify while the other method is a repeating maintence cost that they can bury.

      It is easier to keep paying ER bills to hospitals than to spend a few all at once getting care takers to check on home-care mentally ill and other chronic patients. Maybe hiding their smokes and getting them to eat something that isn’t fried would cut costs too.

  8. Many social problems demonstrate this kind of outlier effect. Education, crime, homelessness, police abuse, traffic accidents. We keep designing policies to address the majority of low-risk cases, and ignore the outliers who need a lot of help.

  9. From each according to ability, to each according to need. It’s not just kind. It’s efficient.

  10. Thanks so much for covering this – I thought the article was amazing. I hope it is scalable though, and can survive without these particular committed individuals. I’m a little worried that Jeff Brenner and Rushika Fernandopulle are linchpin characters, a la Paul Farmer with Partners In Health.

  11. The program Gawande describes seems to have a lot in common with lean production methods. Increase production by removing bottlenecks. In this case, the unit of health “production” is a disease or condition brought under control.

    The higher copayment story is something being done all around the country under the banner of “consumer-directed health care,” meaning, mostly, that the consumer takes on more risk. There’s a famous RAND study somewhere out there on the Internet that looks at the similar case of high-deductible insurance. RAND found that people with high deductibles used fewer health care services. Across the whole scale of lifesaving to frivolous, it was less use, period. Hardly an optimization of the system toward better overall well-being.

  12. “The firm had already raised the employees’ insurance co-payments considerably, hoping to give employees a reason to think twice about unnecessary medical visits, tests, and procedures–make them have some “skin in the game,” as they say.”

    Reading articles like these that make me glad to live in a civilsed country where healthcare is a basic right, not something you rely on your employer to provide.

  13. From each according to ability, to each according to need. It’s not just kind. It’s efficient.

    my grandmother’s favorite line. one of the primary difficulties with this policy heuristic is that it invokes two quantities, only one of which (the latter) can reasonably be measured. what is your ability?

  14. Newsflash: NHS-style systems are dramatically cheaper! It’s a good thing we finally have some domestic data to back this up instead of relying on the experience of every other industrialised country in the world.

  15. The big (nay, HUGE) confounder in this system are the insurance companies. They use these outlier events to simultaneously gouge the end-user (the company contracting the health plan, and therefore the workers paying their share) and to pump up their negotiations with the hospitals who are the payors. Insurance companies and hospitals/clinics are in an annual arms race to raise prices for procedures in a pell-mell, totally illogical, disproportionate manner. It might seem logical in a microsystem, but with all insurance companies and all hospitals in the country engaging this battle annually, the entire collective is haphazard.

    Here is what happens. Hospitals want to get paid, but they can only recoup a fraction of the actual cost, so they look to other high-margin procedures to offset their losses. Simultaneously, insurance companies (who know the disease profiles of their user-base) are trying to reduce the amount they pay for some procedures while “letting go” of some price-tamping on other procedures which they know will be less lucrative in the near future. The net result of this wrestling match is exponential price inflation.

    Also, the system of DRG codes allows both entities to fractionate all medical procedures into increasingly illogical components. This is so that each component has its own profit margin. Put those together for whatever could happen to a patient, and you arrive at the total cost of care. So in some regions, a knee surgery will be $10,000 total, and in other places the same surgery with the exact same components will be $25,000, and broken down into different components that have different price tags.

    Imagine if in Seattle a motherboard cost you $250 but in Miami, the same motherboard cost you $750 and there was no newegg or tigerdirect. In Seattle, the manufacturers were able to price the components lower due to a bulk purchase of some parts and only relatively few expensive parts they had to import for Uzbekistan. Whereas in Miami, all the distributors colluded and knew they could artificially keep capacitors at an unreasonably high cost, which meant a $750 motherboard to the end-user at every store in town.

    It’s not a perfect analogy, but in its essential form, THAT’S what’s happening with health care costs in the U.S.

    Without a ref, the biggest guy on the field makes up the rules as he goes.

  16. I always get a bit of a chuckle when this sort of discussion about health care comes up. It’s always a rehash of arguments that I’ve seen elsewhere in my career, albeit hidden by an extra layer of confusion confusion caused by moral ambiguity.

    Every viewpoint I’ve ever seen in relation to health care matches up with an argument that you’ll find here:

    And the results come out the same. Treat problems early, you wind up avoiding larger expenses later on. Put off treatment and things fail… sometimes catastrophically.

  17. How many above read the whole article? Seems like not many, hence the mostly self-pleasuring speculation.

    Regarding the premise that people would pretend to be sicker than they are in order to access this “disproportionally good health care”: you have to read the article to know how onerous the system actually is for the participants. They really have to want to get well. Yes, they have 24 hr. access to nurses, etc. but also they have a social worker hounding them constantly to quit smoking, lose weight, take those pills, exercise, cook their own dinners, etc. Gawande says that even some truly supersick refuse to participate because they’re suspicious of this kind of official monitoring, or they’re too far gone to really give a crap. Also, the supersick/superusers aren’t paying anything (or much) for their care as it is, so the economic incentive to joining would be minimal. Finally, it couldn’t be that hard to screen for fakers. Each of the participants had professional diagnoses – usually multiple. A well-enough person suddenly visiting the ER regularly could easily be spotted as a fraud (though I know the weeding would add some cost).

    Ubarch,the doctors in the program agreed to be paid a flat fee.

    Thought on scalability: consider Hospice – a large, successful, and money-saving program based on a similar approach. Gawande had an article about it in a previous New Yorker. One of the best things I’ve ever read. But, then, I read it.

    1. Oh, come on. What I’m saying isn’t preposterous, and it’s just silly to think that even if the program is viable on a large scale, that its current incarnation is good enough for large-scale deployment. “Scaling” doesn’t mean you clone the founders and get N experiences that are exactly like this article– It means you try to put together a system that can systematically produce teams that get as close as possible to the pilot experience.

      The author of the article even points out that the viability of a system like this is uncertain because it hasn’t been tested in a large population yet.

      I think you (and definitely lava) are confusing my skepticism with being outright against the approach. I had hoped to avoid that confusion by explicitly stating that I wasn’t. So, let us not be in such violent agreement. I hope they go on testing the approach, and I’m holding back my excitement for when it’s shown to work on a larger population.

  18. So the skeptics say that this won’t work because well people will spoof being chronically ill to get superior healthcare.

    I think I have to weigh in with the conclusion that this is as preposterous as it sounds.

  19. preventative medecine always works better than emergency surgery. This is in fact why publicly paid health care tends to cost less than the for- profit model the US uses, while still providing better care. If people know they can just go to the doctor and not have to pay anything for it they just go, simple. I expect this model can be used to help improve health care services wherever (a good example is for any treatments that may require waiting lists, need could be better determined).

  20. Until the health insurance companies have their balls chopped off, our hands are all tied. No reform is possible. They are the huge fly with one spec of ointment on its right wing. With them in the picture, no further discussion about “what should we do?” can rationally take place.

  21. This reminds me of the article in the NYer a while back about “Million Dollar Murray”.

    They talk about how the common misconception that costs follow a bell curve is not actually reality, they look at costs in common problems like homelessness, police brutality, and health care costs, and in all they notice that costs are concentrated around a few cases “it wouldn’t look like a bell curve. It would look more like a hockey stick. It would follow what statisticians call a “power law” distribution—where all the activity is not in the middle but at one extreme.” They notice that it’s a few chronically ill patients that are costing cities millions of dollars, a few “rogue” police officers that are racking up the vast majority of the complaints, and a few chronically homeless (usually sick and also addicted) that are costing cities the most in medical bills, police enforcement & other services. There’s a solution to these problems that work, but it’s also sad to note (as the doctors in the above article seem to be finding out) that these solutions are also usually politically unpopular. Anyway, it’s a long article, but I highly recommend reading it.

  22. From

    There is a lot of misinformation here, but it is a real jungle put there trying to find truth in medicine. I tend to think if there were “fakers” some would bask in the attention (independent types ….no matter how real their problems do not like surrender). This program is genuinely commendable….proving that accountability for our own actions and well being without mandating it can be fruitful. This concept has been sorely lacking from many social programs.

    The insurance companies are some of the most highly regulated companies out there…so whoever posted was misinformed because new reforms just passed that mean more money goes to the patients…like most reforms someone gets hurt amd someone gets helped.

    While the program outlaid above is truly a vision of good medicine it will eventually create a type of spoiled child in the long run when the gist of it catches on. But we will learn…refine…etc. With private funding it is harder to truly cost compare…and the level of care from these committed doctors who are willing to fundraise means once again it will be hard to duplicate. Not matter how much doctors rally against the label…the money is part of the package…and overall it is completely understandable…they worked hard to get to that level…but it is a stumbling block to making a program like this catch on nationwide for a long term (many programs work short term…the true success is measured in sustainment).

    I will say I smiled upon reading some of the rants against ranting…..sigh…..

Comments are closed.