• Big data challenges

    Frontiers in Massive Data Analysis, from the National Research Council, nails some of the challenges of big data.

    But the challenges for massive data go beyond the storage, indexing, and querying that have been the province of classical database systems (and classical search engines) and, instead, hinge on the ambitious goal of inference. Inference is the problem of turning data into knowledge, where knowledge often is expressed in terms of entities that are not present in the data per se but are present in models that one uses to interpret the data. Statistical rigor is necessary to justify the inferential leap from data to knowledge, and many difficulties arise in attempting to bring statistical principles to bear on massive data. Overlooking this foundation may yield results that are not useful at best, or harmful at worst. In any discussion of massive data and inference, it is essential to be aware that it is quite possible to turn data into something resembling knowledge when actually it is not. Moreover, it can be quite difficult to know that this has happened. [...]

    Another way of characterizing the major problems of massive data analysis is to look at the major inferential challenges that must be addressed. [...]

    • Assessment of sampling biases,
    • Inference about tails,
    • Resampling inference,
    • Change point detection,
    • Reproducibility of analyses,
    • Causal inference for observational data, and
    • Efficient inference for temporal streams. [Emphasis added.]

    I worry a lot about making mistakes in inference as we leverage big data in health care. The questions I’ve been pondering are, how can we draw incorrect causal inferences, and how can we reasonably and reliably protect ourselves from doing so? I think falsification tests can help. I was curious to find out what authors of this report thought about this matter.

    The report is 176 pages long. Here is the entire section devoted to causal modeling:

    Harnessing massive data to support causal inference represents a central scientific challenge. Key application areas include climate change, health-care comparative effectiveness and safety, education, and behavioral economics. Massive data open up exciting new possibilities but present daunting challenges. For example, given electronic health-care records for 100 million people, can we ascertain which drugs cause which side effects? The literature on causal modeling has expanded greatly in recent years, but causal modeling in massive data has attracted little attention.



    Comments closed
  • Bob has diabetes

    Via Aaron on Twitter:



    Comments closed
  • A reader asks for more details on health care spending

    Reader David States wrote me:

    Re your recent TIE post [on the AcademyHealth blog], would really like to see more focus on how money is being spent. Physician salaries vs. nursing vs. therapy vs. administration vs. capital expenditures vs. insurers, etc. There still seems to be a perception that physician salaries are the major driver of health care spending when in fact that account for maybe 10% of spending. Also do not think most people appreciate the rapid growth in categories such as physical and occupational therapy.

    Would also be worth looking at facility fees over time. As hospitals have bought up physician and therapy practices these fees have exploded.

    Question: One assumes that capital expenditure today will drive other expenditures in the future. How much of a spending increment should we anticipate from the building spree that hospitals have been on for the past decade?

    With a bit of searching on TIE, I can partially answer some of this fairly easily, but not all of it:

    1. Aaron got into physician salaries here, writing that in 2006, physician salaries accounted for $138 billion in health care spending. Yeah, that’s pretty outdated, but I am not aware of any more recent data on TIE.
    2. Here’s a TIE post that puts spending on administration at something like $340 billion per year.
    3. This post has data on private construction spending by hospitals, which is part of capital expenditures.
    4. Several of the specific spending questions can be answered by looking at the National Health Accounts. (OK, that’s not on TIE.)

    I think following these links and the links in them, as well as doing some Googling, one could get pretty far on the basic spending questions. I don’t know about facility fees or about the extent to which capital expenditures drive future expenditures, but I am interested in the answers.

    Readers, can you think of other sources to address these questions? If so, drop them in the comments or email/tweet them to me. Comments open for one week from this posting and for leads only. Having said that, I’ll be off the ‘net for a few days, so comment moderation will be slow unless I can get someone else to do it in my absense.


    Comments closed
  • I need help! And an offer.

    My blogging and my research rely on rapid, electronic access to the latest publications in all the leading medical, health services research, and health economics journals. Unfortunately, for a variety of reasons, I am unable to obtain that from any of the organizations with which I am affiliated. (I have tried, really, really tried, and for years, with all the organizations. They know who they are. They are all doing the best they can, and I appreciate them all. They just have failed me in this one way.)

    Some friends and colleagues have stepped up and filled the gap, for which I am grateful. Thank you! But some of those supports are going away. Others are less than ideal. I should be able to get papers I need by myself, and not have to email others and wait for them to do that work. Frankly, I’m sick of the patchwork support I’ve stitched together. Can I do better?

    Let’s see! If you’re at an institution that can offer me what I need, let’s talk! I’m happy to exchange top notch* library access for something. It could be just good PR, but make me an offer! The marginal cost to your institution would be very low.

    * It’s got to deliver PDFs of papers for the very latest issues and early releases of all the major journals in the fields I mentioned. Not all libraries at all universities do this. Some can only delver them delayed by some months. Even weeks of delay is no good to me. I want it as it comes out. That’s what I need to do what I do.


    Comments closed
  • Some facts about Vivity

    Last month, Anthem Blue Cross in California (a division WellPoint) announced it is entering a nonexclusive partnership with seven competitive hospital systems in the Los Angeles area. The new venture is called “Anthem Blue Cross Vivity,” which will offer a network of 6,000 doctors and 14 hospitals in total.

    Some say Vivity is novel, bold, and game changing. What is it? An HMO or an ACO? Or something else? How will it work? Below are some notes (all quotes) from several sources, that partially inform these questions.

    Quotes of Reed Abelson:

    • The partnership includes such well-known medical centers as UCLA Health and Cedars-Sinai. [...] Other hospitals in the new joint venture are Good Samaritan, Huntington Memorial, MemorialCare Health, PIH Health and Torrance Memorial Health.
    • [The ambition is to provide a] level of coordinated, high-quality and efficient care that is now associated with only a handful of integrated health systems like Kaiser Permanente in California, Intermountain Healthcare in Utah and Geisinger Health System in Pennsylvania.
    • [Vivity] will be offered to large employers beginning in 2015 at what Anthem estimates could be 10 percent less than what they are already paying for coverage.
    • [A goal of Vivity is to] aggressively manage care.
    • The hospitals must meet certain quality standards to ensure that they are not stinting on care. [They are expecting to] produce significant savings and profit by reducing unnecessary tests and unneeded hospital and emergency room admissions.

    Quotes of Melanie Evans:

    • [Vivity] will seek to replicate Kaiser Permanente’s success in controlling costs and ensuring quality—but without the singular ownership Kaiser has over its insurance and provider arms.
    • Vivity seeks to compete head-to-head with Kaiser—the Oakland-based integrated delivery system that holds the largest share of the California employer market—as well as Health Net and Blue Shield of California, which also offer HMO and other narrow-network products.
    • Employers will pay a capitated payment for each enrollee [in Vivity]. There will be no deductibles or coinsurance, and copayments for medical services will be modest.
    • Vivity’s partners will pool premiums into one budget. Partners share equally in profits and losses. Partners that fail to meet quality targets won’t earn profits.
    • [Participating hospitals will] move to meld their health IT systems.

    Quotes of Bob Herman:

    • [Vivity] will blend both the past and future as its planned HMO-like structure also looks to implement elements found in today’s accountable care organizations.
    • Vivity will walk and talk like an HMO [... but] will also hold a structure similar to ACO.
    • The plan will use a capitation model, essentially assigning a fixed per-member per-month figure to each beneficiary. Administrative costs and other expenses will be subtracted from the premium risk pool, and if costs are kept in check, all the founders will see profits.
    • Reimbursement will be risk-based, meaning payment will depend on how well systems perform in certain quality measures, which have yet to be determined.
    • The California Public Employees’ Retirement System, a group that negotiates health benefits for state employees and is one of the largest purchasers of health coverage in the county, has signed on.
    •  [To pass scrutiny by antitrust regulators t]hese types of “horizontal integrations” require evidence that other payers are [not?] losing access to the providers in question.

    Quote from John Tozzi:

    • Medical providers and insurance companies in California may have another reason to collaborate to hold down costs: The state insurance commissioner may get the power to reject premium increases if voters approve a ballot measure called Prop 45 in November. A state-imposed ceiling on insurance premiums could limit how much hospitals can charge. For hospitals and insurance companies, sitting down and figuring out how to cut some costs voluntarily—and share in the savings—is a much more palatable option.


    Comments closed
  • How auto-renewal can harm competition

    The following originally appeared on The Upshot (copyright 2014, The New York Times Company).

    My colleagues Margot Sanger-Katz and Amanda Cox wrote recently that shopping around for the best price can be crucial for people renewing their coverage on the health insurance exchanges this fall. But evidence suggests that many people probably won’t do that. Not only is that bad for them, but it can also harm competition, which is bad for everyone.

    A basic truth about health insurance, as with many other things, is that people hate to shop around and change products. They have a status quo bias. That bias can be exacerbated by a large number of plan choices, as consumers in some exchanges face.

    Michael McWilliams and colleagues at Harvardfound evidence of this bias in Medicare Advantage — private plans that serve as alternatives to traditional Medicare. They found that enrollment in the program rose rapidly as the number of plans offered grew from zero to 15, but then leveled off, and declined when more than 30 plans were in the market. Too much choice can overwhelm: As the number of plans climbed above 30, Medicare beneficiaries increasingly eschewed the Medicare Advantage market completely, instead defaulting into traditional Medicare, the status quo.

    Similarly, overwhelmed with increasing choice in the new exchanges, returning consumers may not relish the idea of selecting a new plan. A feature built into the exchanges practically invites them not to do so: auto-renewal. Consumers insured by an exchange plan this year who do not actively choose a new one for next year will be automatically re-enrolled in their current plan or automatically enrolled in a similar one if their plan is discontinued. This auto-renewal is meant to help increase and maintain the size of the insured population and to promote continuous coverage. But if people rely on auto-renewal without evaluating all available options, some may end up in plans that aren’t ideal for them.

    Auto-renewal also offers insurers a way to retain customers without vigorously competing for them, counting on the fact that some consumers will stick with their plans even when, rationally, they should not.

    Next year, the premiums of the currently cheapest silver-rated plans are going up by an average of 8.4 percent. Because of that, many of those plans will no longer be the cheapest. The customers who switch to the silver plans that are the cheapest in 2015 will see their premiums rise by only 1 percent on average. So in raising their rates by that much, those plans may be assuming that status quo bias will keep many of their enrollees from switching to new, cheaper plans offered by competitors.

    There’s a paradox. Basic economic theory suggests that more choice should always improve consumers’ experience, allowing them to get better deals for less money. As choices, variety and competition increase, prices should come down, and consumers should be better able to match their preferences with product features. This theory holds up in many studies of a variety of products, but not all of them.

    In this market, there are many different product features and a lot of variation in them. A recent analysis by the Robert Wood Johnson Foundation found that among silver-rated plans, primary care co-payments ranged from zero to $75, while specialist doctor co-payments varied from $10 to $150. Premiums and the networks of doctors and hospitals in each plan vary as well, as documented in a recent report by McKinsey & Company.

    Variety in these dimensions and others offer consumers a lot of choice. But this variety can also challenge consumers’ ability to make comparisons, leading them to make poorer choices than they could.

    Here, basic economic theory is in conflict with the finding from behavioral economics that when choices become too numerous and complex,consumers resort to heuristics (or shortcuts), leading to suboptimal decisions. For instance, when we can’t fully evaluate all options, we tend to default to familiar brands. And, because it takes time and effort to re-evaluate options, we tend to stick with our initial choice of brand when making a new purchase. This is why, for example, many people stick with their parents’ brand of car or the cellphone preferred by their friends, rather than re-evaluate all makes and models with each purchase.

    A recent study by Charlene Wong and colleagues at the University of Pennsylvania sheds light on just how complex shopping for coverage can be. They closely watched and interviewed 33 young and educated adults as they attempted to navigate Healthcare.gov, the federal exchange that serves as the online marketplace for 36 states. They found that the young adults poorly understood features of insurance plans, had trouble matching plans with their preferences, and were overwhelmed and confused by the volume and types of information provided. A study by the Commonwealth Fund found that nearly one-third of Healthcare.gov users found it “very difficult or impossible” to compare benefits and out-of-pocket costs across plans.

    Auto-renewal is one solution to this problem. Otherwise, if choice is too hard, some may opt out of the market, as happened with Medicare Advantage. These are people who want coverage, but not enough to do the hard work of selecting the right plan. Auto-renewal means they don’t have to do this work each year, but at the risk of getting a worse deal than they otherwise might.

    If we want more competition, we need to induce fewer people to default to auto-renewal. Making it simpler for consumers to compare plans would help. Experts are already suggesting changes that could help consumers navigate insurance markets in 2015. For example, Ms. Wong and colleaguessuggest providing more accessible explanations of health plan terms, making it easier to sort plans by various features, among other ideas.

    Auto-renewal exists for a reason, but if consumers rely on it too much, the results will include higher premiums and greater market power for insurers.


    Comments closed
  • Methods: Falsification tests

    From “Prespecified Falsification End Points: Can They Validate True Observational Associations?” by Vinay Prasad and Anupam Jena:

    [A]nalyses in large data sets are not necessarily correct simply because they are larger. Control groups might not eliminate potential confounders, or many varying definitions of exposure to the agent may be tested (alternative thresholds for dose or duration of a drug)—a form of multiple- hypothesis testing. Just as small, true signals can be identified by these analyses, so too can small, erroneous associations. For instance, several observational studies have found an association between use of PPIs and development of pneumonia, and it is biologically plausible that elevated gastric pH may engender bacterial colonization. [...] In light of the increasing prevalence of such studies and their importance in shaping clinical decisions, it is important to know that the associations identified are true rather than spurious correlations. Prespecified falsification hypotheses may provide an intuitive and useful safeguard when observational data are used to find rare harms.

    A falsification hypothesis is a claim, distinct from the one being tested, that researchers believe is highly unlikely to be causally related to the intervention in question. For instance, a falsification hypothesis may be that PPI use increases the rate of soft tissue infection or myocardial infarction. A confirmed falsification test—in this case, a positive association between PPI use and risks of these conditions—would suggest that an association between PPI use and pneumonia initially suspected to be causal is perhaps confounded by unobserved patient or physician characteristics. Ideally, several prespecified false hypotheses can be tested and, if found not to exist, can support the main study association of interest. In the case of PPIs, falsification analyses have shown that many improbable conditions—chest pain, urinary tract infections, osteoarthritis, rheumatoid arthritis flares, and deep venous thrombosis—are also linked to PPI use, making the claim of an increased risk of pneumonia related to use of the drug unlikely. [...]

    [F]alsification analysis is not a perfect tool for validating the associations in observational studies, nor is it intended to be. The absence of implausible falsification hypotheses does not imply that the primary association of interest is causal, nor does their presence guarantee that real relations do not exist. However, when many false relationships are present, caution is warranted in the interpretation of study findings.

    More on falsification tests here.


    Comments closed
  • AcademyHealth: Decompositions of health spending growth

    What’s the “right” way to decompose health spending growth? That’s a trick question. To find out why, read my new post on the AcademyHealth blog.



    Comments closed
  • How the rise of generic drugs affects pharmaceutical innovation

    For many years, researchers and industry observers have conjectured that rising generic penetration might have an impact on the rate and direction of pharmaceutical innovation. Using a new combination of data sets, we are able to estimate the effects of rising generic penetration on early-stage pharmaceutical innovation. While the overall level of early-stage drug development has continued to increase, generics have had a statistically and economically significant impact on where that development activity is concentrated and how it is done. In the full sample, we find that, as our baseline measure of generic penetration increases by 10% within a therapeutic market, we observe a decrease of 7.9% in early-stage innovation in that market. This implies that drug development activity is moving out of markets where generic competition is increasing and into domains where it is relatively less intense.

    That’s from the conclusion of a new NBER working paper by Lee Branstetter, Chirantan Chatterjee, Matthew Higgins. Because I am not expert in this area, I have almost nothing of value to add, and, in part for that reason, I did not read the paper in full. However, what I did read was very well written and interesting. It’s worth your time if you’re seeking an introduction to pharmaceutical development and patenting, for example.

    The conclusion offers some speculation about welfare effects. They’re good and nuanced, but somewhat limited, as they are focused on the effects of shifts in pharmaceutical innovation only. They did not include the fact that cheaper generics are, by themselves, a welfare gain to consumers though, possibly, a welfare loss to producers. Do these offset? At some degree of substitution of cheaper generics for less (or different) innovation we ought to be indifferent, if the latter is a net welfare loss and the former is a net welfare gain. Maybe some losses or changes in innovation are efficient for this reason. I don’t know.

    In any case, maybe I’m not thinking about this correctly. But the fact that it entered my thoughts at all suggests that the authors might want to address it in a future draft. Or maybe they did so somewhere in the middle of the paper and I missed it. As I said, I did not read the whole thing.


    Comments closed
  • Are Medicare cuts causing private insurer cuts?

    In an economic letter from the Federal Reserve Bank of San Francisco, Jeffrey Clemens, Joshua Gottlieb, and Adam Shapiro make the case that Medicare cuts in rates paid to hospitals induce private insurer cuts. They focus on the 2% reduction in Medicare payments after April 1, 2013, as required by the Budget Control Act of 2011 (sequestration).

    Here’s the key chart:

    changes in hosp prices

    Consistent with Figure 1, we see that Medicare price inflation dropped sharply in April 2013—2.5 percentage points between March and April 2013. Over the subsequent year, private-sector price inflation declined during the months associated with substantial numbers of contract renegotiations. Specifically, private PPI inflation fell 0.6 percentage point in July 2013 and 1.6 percentage points in January 2014. These facts suggest that Medicare’s payment cuts systematically passed through into the private payment system.

    One should not be impressed with some speculative chart reading. Are there any studies, with stronger methods, that support the idea that private prices paid to hospitals fall when Medicare’s do so?

    A study of this relationship in the hospital setting by White (2013) estimates that a 10% reduction in Medicare’s hospital payments results in a 4 to 8% reduction in private payments. White and Wu (2013) further find that hospitals handle these cuts by reducing their operating costs; this and related findings are summarized in Frakt (2013).

    In the context of physician payments, Clemens and Gottlieb (2013) estimate the effects of changes in Medicare’s regional payment adjustments. They find that a $1 reduction in Medicare’s payments results, on average, in a $1 reduction in private payments. Since the average private payment exceeds the average Medicare payment by 40% in their study, the results imply that a 1% reduction in Medicare payments reduces private payments by about 0.7%.

    The letter goes on to discuss the timing of private sector price responses to Medicare payment cuts. Evidence suggests it’s spread out over many years. In this way, even a one-time cut to Medicare payments can suppress health care prices more broadly and for longer than one might expect. Click through for the details.


    Comments closed