Local data, centralized data analysis, and local decision making

Jeffrey Wagar writes:

Nobel-prize winning economist Oliver Williamson’s excellent book “Markets & Hierarchies” offered a set of rules for when Markets were the better method for solving a problem, or when Hierarchies [Government] was better suited. Yet could it be that some of today’s social problems persist because they are not well suited to being resolved by either of those approaches? Data Analytics offers a third choice — empower the end-user/citizen — that I think that can succeed where the other two approaches do not.

For example, in health care, the Market approach rewards drug manufacturers best when they produce drugs that treat the symptoms of illnesses, producing a recurring monthly fee, rather than produce a cure that earns just a one-time fee. Government, in turn, can provide universal health care, but that just redistributes the payments from healthy people to sick people — certainly a worthwhile objective, but an approach that does not increase innovation or productivity. In contrast, if consumers had access to Data Analytics websites displaying your excellent visualizations, perhaps of re-admission rates for doctors (did they diagnose & treat the illness correctly on the first visit), success rates for drugs (by age, race, gender, other prescriptions being taken), infection rates for hospitals, et cetera, the consumer would choose the best doctor/hospital/drug and money would flow to those sources, rewarding their excellence, providing them with more funds to innovate further.

Phil Price and I discussed similar points in our article about local data, centralized data analysis, and local decision making for home radon risk and remediation.

38 thoughts on “Local data, centralized data analysis, and local decision making

  1. That’s not a third choice. The consumer choosing the best doctor/hospital/drug and money flowing to the pocketbooks of the producers the consumer selected is making use of the market mechanism.

    • Yes, this is virtually exactly the kind of tools that people who tout a more market-based approach to health care provision have pushed for. And this is, ironically, what Medicare has begun to do in earnest, at least for quality, and what a pretty enormous number of companies have attempted to do for price (just google “health price transparency”).

      Wager manages to fit a lot of wrong things in here. Like the profitability of cures vs treatment. A cure is obviously worth more than a treatment to patients, so it’s not like they’re going to be the same price–the company can frontload your gains from not having the given disease over the life cycle.

      And even the market/government dichotomy is, of course, largely overblown too, especially in 2017.

      • The real issue is asymmetry of information. If the consumer knows virtually nothing about effectiveness, and believes whatever doctors believe, and doctors believe whatever drug companies tell them to believe… then a market mechanism becomes a massive rent-seeking opportunity. Hospitals come up with fake prices, insurance companies are forced to pay them, they transfer the cost to government subsidies, etc etc etc . it’s a disaster.

        So, more information makes markets more effective. How do we get more information? Certainly not from the makers of Tamiflu.

        It’s not clear to me that governments are the way to get the information created and disseminated, but it is clear to me that more information is better, and will make the markets work better.

        • I was responding to the combination of you and Toby I guess. For example “a cure is obviously worth more than a treatment to patients” is only true if the patients are aware of what is a cure and what isn’t, lack of information keeps a market from converging to what the consumers would really want if the information were freely available.

          How do you “fix” the market information problem? One way is to use a further market approach (ie. have providers sell information, Consumer Reports or the like) another way is to disseminate the information outside a market mechanism (say FDA or Census Bureau or something). There’s not necessarily a reason to think that companies can make enough money off selling information to make a pure market solution come into existence. For example, they may make more money by simply threatening to release the information and getting paid by pharma and hospitals to keep it under wraps!

          Particularly with healthcare data and the possibility for Hospitals and Pharma to lobby to keep people from collecting and disseminating data in the first place (regulatory capture) it’s worth considering the idea that laws should be put in place to relatively precisely and transparently allow data collection and dissemination and to have some of it be done by non-market participants to avoid the kickbacks type situation. Not that kickbacks aren’t a thing for the govt, gotta assume bribes exist etc.

        • >For example “a cure is obviously worth more than a treatment to patients” is only true if the patients are aware of what is a cure and what isn’t, lack of information keeps a market from converging to what the consumers would really want if the information were freely available.

          I disagree. It’s not like I, as a patient, know that my treatment is working, either! There’s a problem with pricing both concepts, of course, which I think is your point, but I was really pushing back at the common (and misguided) meme pushed in the OP that drug companies hold off on developing cures because it’s somehow more profitable.

          I also think regulatory capture is not really the problem here. Individually, hospitals don’t have all that much influence, certainly not at the federal level, except through something like the AHA. But a good hospital prefers transparency and public knowledge of relative quality, while a bad hospital does not, so I’d wager interests are conflicted enough so as not to make it an issue. The real barrier to data access is the law–an exciting move towards new data access was the rise of efforts towards constructing state all-payer claims databases, but the Liberty Mutual v. Gobeille case made those efforts much less useful.

          Also: As I note downthread, there are a *lot* of startups trying to enter this space right now, although none of them seem all that great to me.

  2. There is not the gigo problem for home radon risk in that you can order an assessment of radon risk of your home with fairly well know properties.

    How do you do that for side effect risk of a drug your doctor just prescribed?

    An example of a hierarchy approach to build such as source would be FDA’s Sentinel which draws on separate data providers for central data analysis and eventually sub-group analyses.

    An argument for doing such by a Market approach in epidemiology was given in https://www.researchgate.net/publication/8402267_The_Value_of_Risk-Factor_Black-Box_Epidemiology

    “Research articles can be useful even if they only report how the study was done and what associations were observed without attempting to interpret or explain these observations in terms of methodologic, biologic, or social theories. Readers can weave the observations into their own explanatory theory, one that incorporates their knowledge of related literature.”

    Not sure how successful that has been in Epidemiology or Medicine more generally.

  3. if consumers had access to Data Analytics websites displaying your excellent visualizations, perhaps of re-admission rates for doctors (did they diagnose & treat the illness correctly on the first visit), success rates for drugs (by age, race, gender, other prescriptions being taken), infection rates for hospitals, et cetera

    Does this data exist? Is it a matter of aggregating and cleaning up some data or does it need to be collected in the first place?

      • But converting the data into information that does apply to you can be very tricky – but it is possible to try (i.e. far less of a gigo problem than many other clinical areas.)

        On the other hand success rates for drugs sounds bogus to me as that’s what Sentinel is _trying_ to get to now (all that tricky causal inference from observation only high confidential personal prescription and health outcomes data).

        • Somewhat related (and possibly something Keith or Andrew might want to start a discussion on) is the white paper A Framework for Regulatory Use of Real-World Evidence (https://healthpolicy.duke.edu/sites/default/files/atoms/files/rwe_white_paper_2017.09.06.pdf) that is dated September 13, 2017, but that was announced on Sept. 6 in ASA Connect.

          It defines Real World Data (RWD) as “data relating to patient health status and/or the delivery of health care routinely collected from a variety of sources. This includes data elements captured in a patient’s electronic health record (EHR), in a hospital or insurance company’s administrative and claims data, directly from patients or providers in the course of an observational study, from sources of patient-generated information outside of clinical settings (e.g., in-home monitoring devices, wearable technologies, fitness trackers), and in registries that support various aspects of care and research. It may also include data on contextual metrics not typically considered part of a patient’s routine clinical experience, such as environmental exposures and socio-economic indicators. Importantly, this baseline definition does not preclude incorporation of routinely collected data in traditional randomized clinical trials (RCTs) from being considered RWD.”

        • I think it is a very important topic and for instance it discusses Sentinel I referred to.

          Unfortunately/fortunately I should avoid discussing it publicly.

      • ZC: Thanks for the link. I guess it is an improvement over what I’ve seen previously, but still isn’t as informative as I would hope — e.g., when comparing local hospitals, there were a lot of entries like “Average”, “Not Available”, or “No Different than the National Rate”.

  4. Isn’t the “third choice” simply the market, with an unspecified mechanism for paying for someone providing information? Would you refer to Consumer Reports magazine, or Amazon’s star ratings, as non-market solutions? (If so, that would be strange.)

    • Raghu:

      At least in the radon example, it really is a third way that’s not purely government or market. The data for the radon survey were collected by the government (a combined effort of the federal government and the individual states), and our analysis was made possible by a government research grant.

      • Andrew, is the government collecting data and then contracting out the analysis not a government solution? By the same canard, the government doesn’t build bridges because it contracts out to construction companies. Seems like a distinction without much of a difference.

        • ZC:

          The government collecting data and supporting analysis is the government part. The market part is that the decisions are made by individuals who supply data, get the information, and then decide what to do.

        • I agree with you, but then I don’t think you could define *any* fix as a “pure government” or “pure market” solution. (I happen to think the dichotomy is a waste of time, myself, so that’s fine)

        • ZC:

          Nothing is pure, but for the radon problem there were possible alternatives: (1) no government data collection or analysis, simply let people make their own guesses of their risks; or (2) government data collection and analysis followed by firm regulation (for example, requiring all houses to be measured and then remediated if the measurement exceeded a fixed threshold). One of the points of my paper with Phil is that it can work for data collection and decision making to be decentralized, especially if there is a trusted intermediary that can be involved in data quality and analysis.

        • This, I agree with this. Nothing is pure. It’s not a good idea to pretend that “the market is doing this already” or “markets fix everything” and it’s not a good idea to pretend “government fixes everything” or “government never does any good” or whatever. I happen to think the Census and the like are hugely valuable, and I am glad that everyone has access to that information not just large organizations that can pay say $35,000 a month for a special password giving them access to the data files with severe restrictions on derived information and penalties for disseminating the secrets of the guild.

          A similar data collection and dissemination procedure for healthcare could improve things a lot more than a bunch of regulations saying “hospitals do this, hospitals do that, drug makers do this…”. But I am not holding my breath for a third party market provider “Google Health” or whatever to show up and fill this niche without government intervention, at least at the level of creating laws that define and protect some specific kinds of collection/dissemination activity. The situation with Google Fiber shows what happens when you take on a well entrenched existing pseudo-monopoly with deep legally mandated protections on its way of making money.

    • I think the implied idea is that a government agency whose job is to collect and aggregate information would be a non-market solution to a lack of information which would then make the market provide better health care. I’m not convinced, but I can see that if it worked, it would certainly be a hybrid solution, with government / non-market information generation, while markets provide the health care itself.

      Information is generally a public good, so there may be a good case for non-market information providers (after all, market participants usually jealously guard their special information because it makes them money, think proprietary trading systems in finance, or rug/spice merchants in medieval bazaars or whatnot). On the other hand, I worry a lot about regulatory capture if there is just one approved government information provider. It seems to me that legal solutions that require disclosure of certain aggregated information followed by market providers like Consumer Reports or non-profit providers like The Pew Charitable Trust who slice and dice this information, much like the way people slice and dice Census ACS data or whatever would be a good hybrid approach to improving transparency.

        • Well, it sells its information. So I think it’s a market provider. It wouldn’t exist if it didn’t break even (sales + donations). I think the distinction between nonprofit and for-profit isn’t really that big. A non-profit sells some services, some of which are the services bought by individuals (say the CR magazine or website access) and some of which are the kinds of services that donors want done. If you don’t do what the donors want, you don’t get the donations. It’s just a different service.

        • A nonprofit isn’t a non-market entity, it’s just one that puts surplus income back towards its mission / operations rather than distributing the surplus to its members. The NFL was a nonprofit until 2015, for example, and it certainly raked in cash, payed high salaries, etc.

        • Ikea is a nonprofit organization, it pays no taxes, its mission is to promote interior design… its endowment is thought to be somewhat larger than the Bill Gates foundation.

          The way it works is the parent company is a charity, a subsidiary sells furniture and is a for profit subsidiary, but the profit goes to the charity… it all has the flavor of The Panama Papers but is somewhat more legal.

          http://mentalfloss.com/article/18575/ikea-worlds-largest-charity

        • It’s not exactly like that. The parent company for the Ikea group is INGKA Holding, a Dutch company, which pays taxes on its profits. The owner of this company is a foundation. It’s not Ikea who pays less taxes because of that, but its owner. By the way, there are other examples of companies owned (wholly or in part) by foundations like Bertelsmann or Bosch. I think Zuckerberg was planning to do something similar with most of his Facebook shares; that would have tax benefits for Zuckerberg, not for Facebook.

          Another issue is that Ikea, like many multi-national companies, tries to pay as little taxes as possible. This is done by funnelling money away from INGKA Holding (reducing the taxable net income). For example, paying royalties to another company (which is in turn owned by another foundation controlled by the Kamprad family) that owns the Ikea brand. But that kind of scheme is hardly unique to Ikea.

          http://paperjam.lu/sites/default/files/report_ikea_tax_avoidance_feb2016.pdf

  5. I take issue with the stance presented on the development of treatments. Pharma companies do still develop cures, rather than just treatments for symptoms (although cures are more rare, because they’re much more difficult). And they make money off of cures by pricing them exorbitantly high – take, for example, telaprevir, the first cure for hepatitis C, also the fastest drug to ever reach $1 billion in sales (priced around $189,000 for one course of treatment). Medications taken chronically for prevention or symptom relief are rarely priced at more than $10-20K per year, and then the drug goes generic after ~10 years, so the amount of money being made off of a cure is actually quite similar to the amount made off of a chronic therapy.

  6. I find most of this discussion incredibly naive. Data exists, but it is not really usable for the purposes proposed. I’ve looked through all the data that is available about different physicians, different hospitals, different treatments, etc. (for prostate cancer, but that is just an example) and, despite being a consumer with better access to understanding and resources, it is virtually impossible to make much sense out of it. It is entirely unclear what treatment is best, who is the best provider of that treatment, and don’t even get me started on trying to find out the prices (nobody knows!).

    Of course, I advocate more data being made available. I also believe the incentives of different market (and nonmarket) actors need to be considered. Humans respond to incentives and if you are paid to provide tests and treatment, you are likely to recommend these. But most physicians also care about their patients and are trained to do no harm. Exactly how these two divergent forces reconcile themselves in any individual provider is anybody’s guess.

    I don’t have answers, but I think a good beginning step would be to acknowledge the huge uncertainty about the “best” treatment for almost any condition, the large variability in the quality of different providers, and the lack of good data with which to determine quality. Amidst all this uncertainty, the discussion of market or hierarchies becomes more meaningful. There was a great article (circa 1972 I believe) by Victor Goldberg in the Bell Journal of Economics and Management (now the Rand Journal) – sidenote: it was the only article ever published in that journal that had no equations whatsoever. Anyway, Goldberg’s article was about “administered contracts” and his point was that moving something from the public sector to the private sector does not change the fundamental characteristics of the good or service. Writing contracts when quality is hard to determine is also hard to do well. His point was to stop comparing idealized “free markets” with imperfect government (and really, the reverse as well). We have imperfect markets and imperfect government and hard choices to make.

  7. Following a bit on Dale Lehman’s post, and sorry for the flurry of comments in this thread, but the solution posed in the OP is, from my experience, what every “data science”-inclined person who stumbles into healthcare seems to think is the problem and the solution. And, as a result, there is an extreme oversupply of startups that have collected data and all turn towards building some kind of price and/or quality transparency tool. (At a discussion at the NBER insurance meetings on the topic of price transparency, a discussant rattled off about 15 companies doing this, and that’s not even including public efforts such as those in New Hampshire) To boot, nobody actually uses these tools! For some numbers on this, see Whaley et al. 2014 in JAMA, Brot-Goldberg et al. 2017 in QJE, or Lieber 2017 in AEJ: Policy. Consumer shopping here is virtually impossible.

    • In a way, it reflects a most naive version of the myth of the rational actor. Consumers and suppliers as economic units are often reasonably modeled, in aggregate, as rational actors for the choices that come to be presented in markets. That does not imply that we can pull a new set of products and prices out of our hats, and expect rational decisions in aggregate.

  8. “Government, in turn, can provide universal health care, but that just redistributes the payments from healthy people to sick people” — That’s the least of it. With universal health care, doctors are paid to keep people healthy, whereas in the current system, doctors are only paid when people get sick. The current system creates perverse incentives that help enshrine the unhealth standard American diet to produce an unending supply of chronically sick people.

    • Joshua:

      Back in the 1970s, HMO’s were supposed to have the right incentives: instead of the old system where you’d pay the doctor when you got sick, you’d pay a fixed annual fee to the HMO which would then be motivated to keep you healthy. But something seems to have gone wrong, as somewhere in the tangle of billings that get sent back and forth, some people are making big money on $100 aspirins, $5000 ambulance rides, and the like.

Leave a Reply to Daniel Lakeland Cancel reply

Your email address will not be published. Required fields are marked *