Big Data in Healthcare: Good or Evil? Depends on the Dollars

flying cadeuciiAn organization’s “business model” means: How does it make a living? What revenue streams sustain it? How it does that makes all the difference in the world.

Saturday, Natasha Singer wrote in the New York Times about health plans and healthcare providers using “big data,” including your shopping patterns, car ownership and Internet usage, to segment their markets.

The beginning of the article featured the University of Pittsburgh Medical Center (UPMC) using “predictive health analytics” to target people who would benefit the most from intervention so that they would not need expensive emergency services and surgery. The later part of the article mentioned organizations that used big data to find their best customers among the worried well and get them in for more tests and procedures. The article quoted experts fretting that this would just lead to more unnecessary and unhelpful care just to fatten the providers’ bottom lines.

The article missed the real news here: Why is one organization (UPMC) using big data so that people end up using fewer expensive healthcare resources, while others use it to get people to use more healthcare, even if they don’t really need it?

Because they are paid differently. They have different business models.

UPMC is an integrated system with its own insurance arm covering 2.4 million people. As a system it has largely found a way out of the fee-for-service model. It has a healthier bottom line if its customers are healthier and so need fewer acute and emergency services. The other organizations are fee-for-service. Getting people in for more tests and biopsies is a revenue stream. For UPMC it would just be a cost.

The evil here is not using predictive modeling to segment the market. The evil here is the fee-for-service system that rewards waste and profiteering in medicine.

25 replies »

  1. “Do more stuff, do more complex stuff, do more over-valued stuff.”

    I was thinking about primary care docs and others who essentially bill only for office visits (i.e., E&M codes). I don’t see how FFS incentivizes them to over-treat.

  2. “This discussion (or at least my part of it) is not about capitation. It is about direct fee-for-service to the doctor” … ” among which are the various forms of capitation.” …and previously you said “The evil here is the fee-for-service system…”

    Joe, I know you think otherwise, but you certainly did make your discussion include capitation and fee for service. Look at what you are promoting. Look at what you are demonizing. In the process you totally skip over incentives and third party payer that change the behaviors of all systems of payment.

    Economic incentives might vary, but one of the biggest incentives of a capitated program is the denial of necessary treatment. That brings big profits into capitated plans. That type of incentive doesn’t change just because the letters change in the name. That incentive strongly exists in both the HMO and the ACO.

    Bundled payments: I did a word search for bundled payments and only found one other mention of them by Barry C. so this is new. What are you trying to tell us about bundled payments? I know what they are, but your statement only adds confusion. Bundled payments can both add to and decrease costs; decrease quality and increase quality. It depends upon where and how they are used along with the circumstances that exist at the time. Your conclusion on this subject has no proof and I believe is erroneous.

    I agree, reputation is desirable both economically and for one’s self esteem. Reputation becomes less of a necessity when the process of monopolization exists and the ACA has certainly created that scenario. Third party payer leads to the purchase of insurance with less than a stellar reputation because the third party is more interested in price than in reputation.

    Yes, I hope you enjoy your party. This is a party day/weekend and I will be doing the same. As soon as you are able see the excellent documentary film “America: Imagine the world without her”. You will find it quite interesting and enlightening even if you might have some disagreement with the authors politics. It’s about America, the country we are celebrating this weekend, so it probably should be seen by everyone. July 4 is not just about firecrackers and parties. It is about the birth of a great nation.

  3. Both, really. Do more stuff, do more complex stuff, do more over-valued stuff. That’s the “volume” side of things.

  4. This discussion (or at least my part of it) is not about capitation. It is about direct fee-for-service to the doctor (in which the doctor gets paid more for doing more and more complex things) and all the other ways of paying for healthcare, among which are the various forms of capitation. The economic incentives vary tremendously among all these different ways. Bundled payments for cases from out of the region, for instance, is fee-for-service. The patient and the patient’s insurer pay a fee for the service. But they don’t pay each doctor a fee for each service. So the doctors have no incentive to lard in lots of little extras. They have an incentive to do each one efficiently and effectively, so that they can do more of them and their reputation will bring them more cases.

    And yes, all doctors care about their reputation. But here I am talking about reputation as an economic incentive, not just a matter of professional pride. Perhaps “branding” might be a better word for it — the general high reputation that brings a steady flow of patients to places like Mayo and MD Anderson.

    I will look at the studies. And have a happy Independence Day. I have to go now and prepare for our party.

  5. Joe, of course incentives change if the method of payment changes, but if two things are capitated in the same way, they will have very similar incentives. The discussion was about capitation. But I also touched on hybrids so I am not sure what you are trying to demonstrate.

    Reputation is very important, but in a third party payer system, which you have yet to mention, reputation becomes much less important. Do you see what happens when you attempt to control the marketplace? Reputation is very important to most physicians no matter what type of practice they have. It isn’t as important to the HMO as those that run them generally live in a different area. Many levels of responsibility develop.

    ” is looking at cost in the last six months or year of life”

    That is what I addressed in my comments while at the same time warning you that one cannot adequately predict the last 6 months of life. You cannot compare the Mayo to a local hospital. Patients fly out from their home town to the Mayo and then return home where they die. You have to assimilate that into your calculations, but that is often overlooked and I think you are overlooking that here as well. You are in good company for the Dartmouth Atlas frequently makes the same mistake. In fact I believe the first study that put them on the map lacked an easily recognizable variable that changes the conclusions, but ideology seems to trump science.

    Try reading that paper I gave you the citation for. Its commentary is very valuable in showing how mistakes can be made and a lot of other things about this type of study. Also start thinking about that paper trail I have mentioned and see if that affects the way you think.

  6. Does FFS really incentivize over- treatment by physicians paid based on E&M codes/cognitive services? Or does it only incentivize over-treatment by physicians who are paid based on procedures that are grotesquely over-valued by the RUC and CMS?

  7. I wouldn’t say that all the incentives are exactly the same in anything labeled “HMO” or “ACO.” Roughly speaking, a FFS payment system incentivizes over treatment while a risk situation incentivizes under treatment. But the behavior of a particular medical group can depend on what they are truly at risk for.

    For instance, the Alternative Quality Contracts under BCBS Massachusetts put primary care providers at risk for the costs of ED visits, hospitalizations and surgeries without giving them any “gatekeeper” power to keep patients from accessing those services. The only way they can make more money/do better is by providing better after hours, preventive and chronic care that the patients don’t need to go to the ED or get hospitalized. Those are very different incentives.

    Incentives for medical groups definitely include their reputation in the market place. Kaiser’s reputation suffered a lot in the 1990s while trying to radically keep costs down to compete with the HMOs that were popping up everywhere. What seemed to turn that around was a rebellion by the doctors (who, as you note, take half of the margin for themselves). A new overall management structure gave the docs more power, and their quality seemed to improve. Reputation management is a serious incentive for groups like Mayo, Cleveland Clinics, and Scott & White, as well.

    As for the “last 6 months of life” study, you are misinterpreting what it means. It had nothing to do with predicting who would die, much less hastening their deaths. It was simply looking at the medical costs of patients who had died in the care various centers after being involved with them for more than 6 months. It was a Dartmouth study. At the extremes, the costs at Cedars Sinai in LA, UCLA Medical Center, and NYU Langone were twice what they were at Mayo and Cleveland Clinics.

  8. Joe, this study is not one done today. It was done awhile back, but the important things, THE INCENTIVES, haven’t changed. When one burns themselves with boiling water from an old Tea maker one doesn’t try the experiment again with a new Tea maker under a different name unless the problem has been solved.

    Unless one deals with the incentives one will fall into the identical trap. The incentives of HMO’s and ACO’s remain essentially the same except the present day ACO can have more power and has been aided by the change in some laws protecting the ACO from suit. That is not patient friendly.

    The GAO study was done for the original HMO where today’s MA didn’t apply. It took the GAO at least 10 years to recognize what was apparent to those that understood what was happening and understood incentives. I think the exact statement from the GAO might have been along this line… if risk was included the HMO’s should have on average been paid 68% of the per capita rate. [totally from memory]

    Oct 2 JAMA. 1996;276:1039-1047

    Differences in 4-Year Health Outcomes for Elderly and Poor, Chronically Ill Patients Treated in HMO and Fee-for-Service Systems

    Results From the Medical Outcomes Study

    John E. Ware, Jr, PhD; Martha S. Bayliss, MSc; William H. Rogers, PhD; Mark Kosinski, MA; Alvin R. Tarlov, MD

    I want you to understand, Joe, that though I do not personally like HMO’s I believe they should be among the free choices available as long as the tax deduction is offered to the patient in the same fashion as it is offered to the employer. If this occurs then some of the problems with incentives might be corrected.

    I don’t want to get into Kaiser in this already long reply, but Kaiser has deep problems and is protected politically and by MICRA. It is considered a staff model HMO, but the physician *partners* get to share the profits 50:50 with the hospital portion. This was called the Tahoe agreement that was made a long time ago. Thus it isn’t true that physician partners have nothing to gain by denying necessary treatment. They have a lot to gain.

    Hybrid functioning will be different based upon how much the hybrid affects the basic incentives. Denial of care frequently doesn’t leave a paper trail that can prove negligence in court. Fee for service most frequently leaves large paper trails and a whole bunch of people in the know.

    Be careful about the assumption of the last 6 months of life for you are looking at known data. We don’t know which patient has only 6 months of life left while the patient is still alive. We don’t know which patient will have useful life. We can only guess and perhaps tilt the odds in one direction or the other, but I don’t think the American public likes the idea of letting people die based upon odds that aren’t clear and definitive.

    If one wishes to put the finger on and solve one thing that has caused the American health are system to cost too much and have the difficulties it has one has to look at and abandon the third party payer system. This does not mean that people don’t carry insurance, nor does it mean that government cannot be involved.

    The extraordinary thing about Mayo and other great centers of medical care is that they all function differently and likely cannot be reproduced in a massive scale. Locations are different, patients are different and doctors are different. We need a multiplicity of free market solutions that compete with one another. (Again this does not mean that government cannot be involved.)

    I cannot resist one last statement on Kaiser. I have heard that they have ended lives of patients quicker than the patient or family wanted under circumstances that do not look favorably on Kaiser. The claim is that this is one way Kaiser saved money.

  9. Interesting, Allan. I certainly agree that no one model is a panacea, and certainly not the rather heterogenous group lumped under labels such as “HMO” or “ACO.”

    Do you have a cite or an exact title for either the GAO study or the Ware study so that I can look at them and add them to my database? I would imagine the 30% overpayment estimate might also include the fact that Medicare Advantage plans, after being established on a claim that they would save money, have in fact been paid 15% or so more than FFS plans.

    One of the things I would wonder about is to how those numbers differ between large, traditional HMOs (such as Kaiser) and the “HMOs” that sprouted so vigorously in the ’90s “managed care” rage, which were not staff-model HMOs, and did not involve doctors who had chosen for their careers to be in such models. And whether they differed between “pure” HMOs (in which, like Kaiser and Group Health of Puget Sound, there is one risk model for essentially all patients) and HMOs that involve medical groups operating in a more hybrid environment — especially ones where they might have some choice whether to steer a long-term patient into an risk payment situation or a FFS situation.

    An interesting part of this discussion is looking at cost in the last six months or year of life, in which the patient typically engages heavily with all parts of the system, from pharma to surgery to hospital care. There is a clear differentiation between the most expensive and the least expensive, and it is not based on risk vs. FFS. The real stand-outs for least expensive on this basis are all medical groups working together in large facilities, such as Mayo, Cleveland Clinics, and Scott & White in Texas. They seem to make better, more efficient and appropriate decisions while being mindful that they survive on their reputation for great care.

  10. Perry, I see no solution to this. All systems have trade offs.

    The vilification of FFS, largely justified IMO, has not been accompanied by an honest assessment of its benefits in the taking on of risky patients.

    Which would be fine if the culture of the country was one where the shepherd was praised for ignoring the lost sheep & attending to the flock of 99.

    It’s not. Except in the Gaussian distributions produced by academia.

  11. Big data can be very poorly used by all types of payment methods. The one payment method with the worst record for those that are sick and need care is the capitation model. HMO’s (same incentives as ACO’s) typically used all sorts of methods to cherry pick the healthy. The GAO finally released a report many years ago demonstrating that based upon risk HMO’s of the past were paid about ~30% more than the per capita rate would have been if risk were included in the payment plan proving that cherry picking existed as a basic method of making a profit.

    The suits that are on record demonstrate that those signed up with these capitated plans were denied necessary care on a regular basis. The best study of the time written by Ware et. al. concluded “During the study period, elderly and poor chronically ill patients had worse physical health outcomes in HMOs than in FFS systems”

    This does not mean there aren’t problems with the fee for service model though it has been proven safer for the sick and poor patient. I am just balancing the rhetoric of this latest piece. Remember denial of treatment need not contain medical records that show what was denied and therefore protect the criminal from prosecution., however fee for service records do leave paper trails along with multiple observers. That is how criminality was proven with some physicians that deserve to remain in jail.

    If Kaiser wants transparency then it would release the records of all the settlements that have been sealed. It would also not prevent Kaiser physician owners from speaking out. I have been told that when a Kaiser physician speaks out they jeopardize their retirement benefits. I cannot prove that contention, but this information comes from prior Kaiser physicians.

  12. Big data might give us some insight into that, though it is hard to tell people’s intentions from, for instance, sales data.

    My own guess would be more along the lines of this: These days you can sign up for any insurance you want, no matter what pre-existing condition you may have. If you knew you had some kind of cancer and had not signed up for insurance yet, would you be likely to sign up with Kaiser, which would want to care for you within their system, or would you see what health plan might give you access to a place that specializes in cancer, such as MD Anderson or Fred Hutchinson?

  13. I was thinking more along the lines that, if they have a reputation for carefully monitoring care for meeting their internal standards of appropriateness (not necessarily something I disagree with), would patients who know they will need lots of expensive care choose another insurer when presented with a choice? Not that they’re run away, but that they avoid KP to begin with. I guess big data would give us some insight into this.

  14. No, it’s not a way of encouraging the sickest patients to go elsewhere. Because we are not talking about the sickest patients who really need help. I know more than my share of people who went through the Kaiser system with cancer and have high praise for the care they received.

    No we are talking about, for instance, me and my arthritic knees. I went in demanding a new knee some time back. They never said no, but put me through a series of tests and discussed my alternatives and their likely outcomes — which convinced me that I was better off with medical management. They did the medical management skillfully and continued to check back with me about it even years later. I remain convinced that was the right decision. Sometimes more (and more expensive) medical care is not better.

  15. “One area you mention which Kaiser has done well at is convincing their patients that more is not always better”

    And isn’t that a way of encouraging the sickest patients to go elsewhere? Glad to hear that they’re increasingly vulnerable.

  16. Saurabh,

    This is precisely the question to be answered when we look at performance issues. Who wants to take care of the sickest patients when the outcomes are going to be loaded against a good recovery? We do need the cowboy types because that’s how we learn to do miraculous things with shattered bodies, but who will want to if you’re going to be penalized for a bad outcome?

  17. I used to work for a surgeon who was highly astute at predicting the operative risk of patients. Another was a cowboy. He took all the train wrecks. They had differing mortality rates. The former would be delighted to hear that we are moving to pay for performance. His numbers performed. But it was the cowboy who truly performed.

    Predictive analytics can predict the low risk for skimming profits and the high risk to avoid for “achieving bonuses by better outcomes.”

    Insurers do this all the time. We call it risk classification. It’s frowned upon and rightly so.

    Predictive analytics are value-neutral. Not least because the line between “I’m avoiding the Gomer because she’s unlikely to survive a ruptured aneurysm” and “I’m avoiding the Gomer because if she’s going to die on the table that’s less in the Christmas bonus packet” is a thin, blurry and dark one.

    You might find that integrated systems are sending their patients to traditional fee for service models.

    Ah, that’s where predictive analytics come in handy!

  18. Maybe hospitals and other healthcare providers can improve patient outcomes by having access to patient daily spending habits. But I think another reason they are doing this, and a more important reason at that in turns of saving their bottom line, is to find evidence against patients in cases involving readmission penalties. That way if a hospital has concrete evidence that one of their heart failure patients bought a Big Mac, Biggie fries and a super-sized soft drink at McDonald’s prior to showing up at the ER in need of being readmitted for CHF exacerbation, the hospital can ask Medicare to reverse their penalty for having a readmission within the 30-day allotment window. I don’t know for sure if Medicare would agree to this, but to be fair, they should.

    Apparently, the overseers at Medicare didn’t foresee that the Rube Goldberg set of readmission penalties, which they have imposed on hospitals, is causing hospitals to take healthcare dollars that would normally be spent on patient care and use them to buy data-mining software programs that are usually just reserved for credit card companies and law enforcement agencies. Hospitals already have this overly costly problem of having too many employees in the back office working as data-miners and the like and not enough employees on the front lines working as direct-patient caregivers. And Medicare only wants to makes this problem worse?

    How in the world could the folks at Medicare possibly think that imposing these types of penalties on hospitals will result in better patient care! Don’t they understand that this sort of bureaucratic overreach is doing nothing but causing the overhead costs among hospitals to go up? These folks are beyond stupid. They are total idiots!

  19. Thanks for putting a little more meat on the bones of my little 300-word post, Barry.

    Clearly we are talking about two different ways to use market segmentation — giving more care to those who would benefit the most, and giving more care to those who don’t necessarily need it, but who have the most comprehensive insurance. The NYTimes article discusses UPMC’s market segmentation only in terms of its insurance members, as a method of keeping its clinical costs down on those insured members. How it acts in its market for other insurers is a different question. Its prices will only be forced down by competition of one form or another, including many of the devices you mentioned, especially reference pricing and competitive medical tourism.

    Whether this model works, and how easily, has added importance because it is part of the model UPMC is exporting through Evolent, its joint venture with the Conference Board.

    There is little doubt that market dominance strongly affects prices, and Western Pennsylvania (like southwest Georgia) is a prime example.

    Kaiser is another interesting example. One area you mention which Kaiser has done well at is convincing their patients that more is not always better, as I have experienced first hand as a Kaiser member. For the most part, they have been strongly developing the tools, but have struggled so far to bring them to bear to lower their internal costs so that they can be more competitive in their rapidly evolving markets. They have a strong market positions in their regions, but they cannot be said to dominate, and in the new environment are actually quite vulnerable.

  20. My impression is that for UPMC, the largest healthcare provider in its region, its own insured members account for a comparatively small percentage of its patient base and due to its high profile name and its significant regional market power, it’s the most expensive provider in the area for other insurers. I don’t know whether it charges less for health insurance than Highmark Blue Cross and other insurers in the market or how extensive its network is beyond UPMC owned facilities. As for getting people in for testing to prevent expensive hospitalizations down the road, we don’t know how many people need to be tested and at what cost to prevent one expensive surgery. The usual complaints about UPMC are that it uses its market power to drive up costs for other insurers and patients. That’s why Highmark bought West Penn Allegheny.

    A lot of people think the future of healthcare is in managing health and healthcare costs at the population level through alternative business models like capitation and shared savings, coupled with bundled payments for surgical procedures, reference pricing, tiered networks and narrow networks among other approaches. Kaiser is the largest example of a provider that is also a health insurer. It also makes extensive use of technology. It should have significantly lower costs than its insurance industry competitors but I don’t think it does or at least its cost advantage is not nearly as great as it once was. For example, Kaiser allows patients to contact their doctors with questions via e-mail. Patients love this but somebody at Kaiser must read and respond to the e-mails. Kaiser gets 15 million of these e-mails per year but the weird thing is that it hasn’t reduced the number of in person doctor visits according to former CEO, George Halvorson.

    If we really want to lower healthcare costs or at least reduce the growth rate, we need to convince patients that more care isn’t always better care and more expensive care isn’t always better care either. Patients also need to be more accepting of death when the time comes as they already are in some parts of our country and in most other developed countries. Sensible tort reform could help to reduce defensive medicine and good user friendly price transparency tools should make it easier for referring doctors to send their patients to the most cost-effective high quality providers.

  21. Your conclusions far exceed your data and the level of proof you provide which is nil.

    (Take note I am not talking about the potentialialities of predictive modeling nor UPMC’s role in healthcare.)

    Additionally you made the statement ” it has largely found a way out of the fee-for-service model.” Tell us how and give us some numbers.

  22. Thanks for the reference to the report, Ceci. We really can’t say anything helpful or insightful about how healthcare uses these new technologies without reference to what organizations are trying to do with them, which will be dictated by how they make their money.

  23. Nice post Joe. Pithy.

    I think this is a great exercise in understanding that what’s really going on in healthcare is a little harder to come to terms with than it might at first seem.

    Potential conflicts do not always translate themselves into real conflicts.

    Great point on the sources the report relies on. Much of the time, going to the “for and against” quote guys frames the story in a simplistic way that buries the nuance. Sure, as a journalist, I got a great quote – but am I serving my readers? I’m not so sure ..