Innovation policy often focuses on fortifying the incentives of firms that develop and sell new products by offering them lucrative rights to exclude competitors from the market. Regulators also rely on these same firms—and on similar incentives—to develop information about the effects of their products in patients, despite their obvious conflict of interest. The result may be a distorted understanding that leads to overuse of expensive new medical technologies. Recent technological advances have put healthcare payers in an excellent position to play a larger role in future innovation to improve healthcare and reduce its costs. Insurance companies and integrated healthcare providers have custody of treasure troves of data about healthcare provision and outcomes that can yield valuable insights about the effects of medical treatment without the need to conduct costly clinical trials. Some integrated healthcare systems have seized upon this advantage to make notable discoveries about the effects of particular products that have changed the standard of care. Moreover, to the extent that healthcare payers can profit from reducing costs, they will seek to avoid inappropriate use of costly technologies. Greater involvement of payers in healthcare innovation thus offers a potential counterweight to the incentives of product sellers to promote excessive use of costly new products. In recent years, the federal government has sought to promote innovation through analysis of healthcare records in a series of initiatives; some picture insurers as passive data repositories, while others provide opportunities for insurers to take a more active role in innovation. In this paper, we examine the role of health insurers in developing new knowledge about the provision and effects of healthcare—what we call ‘demand-side innovation’. We address the contours of this underexplored area of innovation and describe the behavior of participating firms. We examine the effects of current legal rules on demand-side innovation, including insurance regulation, intellectual property rules, privacy protections, and FDA regulation of new healthcare technologies. Throughout, we highlight many policy tools that government can use and is using to facilitate payer innovation outside the traditional toolkit of patents and exclusive rights.
Policy mechanisms to promote biopharmaceutical innovation often focus on fortifying incentives for firms to develop new products. Biopharmaceutical firms favor exclusionary rights that defer competition, allowing them to profit by charging higher prices prior to generic entry. In addition to providing patent term extensions for developers of new drugs,1 Congress has repeatedly provided for periods of regulatory exclusivity to encourage the same firms to collect and submit data about the effects of their products in patients.2 Providing better information about these effects is an important form of innovation that distinguishes warfarin as a human therapeutic anticoagulant from the same substance as a rat poison.3 But it is problematic to rely on product-developing firms to provide this information, because although they might profit from favorable information, they stand to lose money from disclosure of negative information about their own products. Regulatory mandates require sellers to produce data from rigorous clinical trials showing that their products are safe and effective as a condition for approval of new drugs.4 But side effects are difficult to observe in clinical trials of limited scope and duration. Often the bad news only comes to light after products have been widely used; if the news is bad enough, it may lead to the withdrawal of previously approved products from the market.5 But the adverse events reporting system6 that FDA has long relied upon as its principal source of bad news after approval is haphazard and unreliable.7
Healthcare payers,8 on the other hand, stand to profit from the bad news. Information that an expensive drug has harmful side effects, or that it does not work for many of the patients currently taking it, could lead to more sparing use of these products, reducing healthcare costs while improving quality of care. The incentives of payers to cut costs, though sometimes problematic in isolation, could be a corrective counterweight to the incentives of product sellers to maximize their own patent-protected profits.
Recent technological advances have put healthcare payers in an excellent position to play a larger role in future innovation to improve healthcare through better understanding of the effects of medical treatment. Insurance companies and integrated healthcare providers have custody of treasure troves of data about healthcare provision and outcomes that can yield valuable insights about how to improve the quality of healthcare and lower its costs. Some integrated healthcare systems have seized upon this advantage to make notable discoveries about the effects of particular products that have changed the standard of care.
Studying the consequences of past clinical care to improve healthcare practice is an important research frontier with the potential to yield valuable innovations. Although it is easier to recognize innovation when a new product is introduced than when new information leads to more sparing use or even withdrawal of existing products from the market, in both cases new knowledge is put to use to improve the quality of healthcare. Both are socially valuable forms of innovation. But the distribution of benefits from the two forms of innovation is quite different. Much of the social value of new products accrues to product sellers, at least when they are protected from competition by patents and regulatory exclusivity. On the other hand, when further knowledge leads to more parsimonious use of existing products, the benefit is captured on the demand side by payers and by patients who save money and improve health by using less of these products. These potential savings could create an incentive for innovation on the demand side by institutions that pay for healthcare.9
Healthcare payers enjoy several advantages that allow them to complement the role of product-developing firms as providers of information about the effects of healthcare products. First, payers have access to large volumes of data from administrative claims and healthcare records that reveal healthcare consequences. Although randomized, controlled clinical trials have long been considered the gold standard for studying treatment effects free of selection bias, healthcare records may provide much larger data sets and observations over longer periods of time, and can thus shed light on questions that clinical trials leave unresolved. Second, payers have an incentive to reduce healthcare costs rather than to increase them, providing a counterweight to the incentives of product-developing firms. Third, the observational studies that payers can pursue are cheaper than the controlled clinical trials that swell the R&D budgets of product-developing firms.
The standard policy toolkit for promoting biomedical innovation offers little in the way of direct benefits to these ‘demand side innovators’, although the exclusive rights that the legal system awards to developers of new products may give payers an indirect incentive to learn more about whether these products are worth their high costs. But the standard toolkit is not the only way to promote innovation. The federal government has used a variety of different mechanisms to promote the use of data from healthcare records in ongoing innovation. These mechanisms include agency initiatives to use healthcare records for regulatory purposes, such as FDA's Sentinel System, and for research purposes, such as the NIH-sponsored Precision Medicine Initiative and eMERGE network. They also include new legislation to support these initiatives10 and others, such as the establishment of the Patient Centered Outcomes Research Institute under the Patient Protection and Affordable Care Act of 201011 and agency rulemaking to address obstacles to research use of healthcare records.12 Although some of these initiatives picture payers primarily as repositories of data that others might analyse, they also provide opportunities for insurers to become more fully engaged as partners in healthcare innovation. Healthcare payers participate in medical innovation to an extent that is largely unrecognized in the legal scholarship on innovation. They could potentially do much more.
This paper proceeds in three parts. In part I, we outline the contours of this underexplored area of innovation. We describe the incentives, resources, and opportunities available to payers, as well as technical obstacles to medical innovation by payers, focusing on the challenges of making payer data useful for research. Part II considers economic and legal obstacles to payer innovation, including features of the market for health insurance and healthcare, regulation of health insurance and health technologies, limitations on intellectual property protection, and privacy protections for health information. In part III, we describe government initiatives that have helped the industry begin to address those challenges and identify further opportunities to assist this emerging area of research through law and policy. Throughout, we highlight the multipronged way that government facilitates payer innovation without relying on exclusionary rights. Although these ‘demand-side innovators’ do not directly benefit from the exclusionary rights favored by pharmaceutical firms, they have nonetheless benefited from a variety of government initiatives that have lowered the legal and technical barriers to innovation while building collaborative networks to share information and expertise.
INNOVATION BY HEALTHCARE PAYERS
While payers may lack the scientific labs of pharmaceutical companies and the front-line patient interactions of practicing physicians, they have access to valuable health data that can shed light on questions about what works in different clinical contexts and in different kinds of patients. These data give payers an advantage in innovation to improve the choice of appropriate treatments. This part describes the innovation landscape for payers. It begins by giving two examples of payer innovation efforts that fit poorly in a regulatory regime that was designed for the use of product-developing firms. Next, it briefly canvasses the innovation resources and opportunities available to payers, with a focus on research questions that payers might be better positioned to address than product-developing firms. It concludes by reviewing technical challenges to payer innovation.
Early Efforts and Regulatory Obstacles
Two extended examples highlight the incentives for and potential benefits of payer innovation, while also showing challenges payers face in implementing their innovations in a regulatory regime designed for innovation by drug-developing firms. The first involves a request by payers to FDA to switch the terms of approval for the antihistamines Allegra, Claritin, and Zyrtec from prescription (Rx) to over-the-counter (OTC) sales. The second involves the use of data from payer records rather than from drug company clinical trials to establish toxic side effects of the painkiller Vioxx, and illustrates the reluctance of FDA to rely on those data. It is no coincidence that both involve widely prescribed, patent-protected blockbuster products that were costing payers a lot of money.
Rx-to-OTC switch: non-sedating antihistamines
The first example illustrates the divergent interests of payers and drug manufacturers in the context of regulatory approval for switching drugs from prescription (Rx) to OTC sales.13 An Rx-to-OTC switch can be a significant cost-lowering innovation for at least two reasons.14 First, it permits patients to treat themselves without incurring the costs and delays associated with seeing a doctor for a prescription. Second, it often leads to a significant price reduction for the drug itself, because health insurance typically covers Rx but not OTC drugs and patients are likely to be more cost-sensitive than insurance companies.15
In 1998, Blue Cross of California (later Wellpoint) submitted a petition to FDA asking it to permit OTC sales of non-sedating antihistamines sold under the brand names Allegra, Claritin, and Zyrtec.16 Blue Cross/Wellpoint argued that non-sedating antihistamines were safer than older antihistamines, already available OTC, which had significant sedative side effects. According to the petition, the lack of OTC access to the safer non-sedating products ‘results in a greater incidence of side effects associated with the OTC alternatives adding considerable unnecessary medical costs to the health care system’. Of course, the switch would also save costs for Blue Cross/Wellpoint by allowing patients to purchase their own non-sedating antihistamines out of their own pockets in the OTC market rather than using insurance to pay for doctor visits and prescriptions.
The product manufacturers opposed the switch, arguing that Blue Cross/Wellpoint had failed to submit adequate supporting data to establish the safety and efficacy of the non-sedating products when used without the supervision of a physician.17 In warning regulators about the potential hazards of their products, the product manufacturers were also advancing their own financial interests. Drug manufacturers typically wait to seek approval for an Rx-to-OTC switch until the drug approaches the end of its patent life, when generic competition will soon erode profits. At that point, the firm may seek to mitigate the loss of revenue by claiming a statutory reward of exclusivity for conducting further clinical trials to support a change in the terms of regulatory approval.18 If further clinical trials are ‘essential’ to FDA approval of an application for the switch, the manufacturer is entitled to 3 years of exclusivity before FDA will approve a generic product for OTC sales.19 This supplemental exclusivity gives the branded product a 3-year head start in the OTC market. A switch prior to patent expiration would be less attractive to the firm because it would have to surrender more lucrative exclusivity in the Rx market in exchange for less lucrative exclusivity in the OTC market; an early switch would also hasten the arrival of full competition by allowing the OTC exclusivity period to run during the patent term. To maximize profits, drug companies would ordinarily prefer to delay the modest innovation of an Rx-to-OTC switch until the end of the patent term. Moreover, the law gives these companies a perverse incentive to persuade FDA that they must conduct costly clinical trials before the switch, because the statute authorizes further exclusivity only if new clinical trials are essential for approval, and not if FDA agrees that the product is safe for OTC sales without the need for further study. The manufacturers thus had to persuade FDA that more data were necessary to support the switch in order to get additional years of exclusivity. The Blue Cross/Wellpoint petition not only threatened to end payer coverage of non-sedating antihistamines before the manufacturers found the switch commercially advantageous, but it also undermined the case for 3 years of exclusivity in the OTC market.
The scientific question presented by the petition was more straightforward than the regulatory moves. FDA asked an advisory committee whether non-sedating antihistamines ‘could be used appropriately and safely by consumers without the intervention of a learned intermediary’,20 and the committee concluded that they could.21 But although this seemed sufficient to give FDA the authority to approve OTC sales on the petition of ‘any interested person’,22 it was unprecedented and controversial to grant such a petition over the objection of the drug manufacturer.23 The more traveled pathway was for the manufacturer itself to initiate an OTC switch by filing a supplemental new drug application at a time of its choosing. And sure enough, Schering-Plough soon filed its own application for an Rx-to-OTC switch for Claritin—the first of the non-sedating antihistamines to face patent expiration—11 months after opposing the Blue Cross/Wellpoint petition.24 FDA approved the Schering-Plough application on November 27, 2002, without ruling on the Blue Cross/Wellpoint petition.25 The patent protecting Claritin expired 3 weeks later.26
This episode shows how the interest of payers in reducing healthcare costs diverges from the interest of product manufacturers in maximizing revenues, giving payers an interest in accelerating a cost-lowering innovation (an Rx-to-OTC switch) that a manufacturer would rather defer. The statutory incentive of regulatory exclusivity may eventually motivate a manufacturer to conduct clinical trials and to pursue an Rx-to-OTC switch just prior to patent expiration, as Schering-Plough did in the case of Claritin. But payers might find it worthwhile to pursue this cost-lowering innovation more promptly and without the need for propping up prices through regulatory exclusivity. Moreover, because they do not stand to gain from persuading FDA that costly clinical trials are necessary to support a switch, payers are motivated to show safety at lower cost by consulting their own data from clinical experience with the drug without conducting potentially unnecessary clinical trials.27 But even motivated payers may need regulatory approval in order to implement cost-lowering changes in healthcare, and the FDA regulatory regime was not designed with payer participation in mind. Even with the support of an FDA advisory committee, Blue Cross/Wellpoint was unable to get FDA to approve the switch until the product manufacturer submitted its own petition in its own time. This limit on the ability of payers to implement innovations could dampen the incentive to conduct studies in the first place.
Post-approval studies: selective COX-2 inhibitors
The second example concerns exposure of a toxic side effect of the blockbuster drug Vioxx through research in health records of the integrated healthcare provider Kaiser Permanente. Vioxx is a selective Cox-2 inhibitive non-steroidal anti-inflammatory drug (NSAID) for relieving pain and inflammation without the gastric side effects of an earlier generation of NSAIDs (such as aspirin and ibuprofen).28 The manufacturer Merck voluntarily withdrew Vioxx from the market in the fall of 2004, at a time when it had sales of $2.5 billion per year, in the face of mounting evidence that Vioxx was causing fatal heart attacks.29
Data from Merck-sponsored clinical trials comparing Vioxx with naproxen (one of the older generation of NSAIDs) had previously shown more heart attacks (as well as fewer gastric side effects30) in patients taking Vioxx,31 but Merck had argued that the difference in heart attacks reflected a protective effect of naproxen rather than a toxic effect of Vioxx.32 FDA was not convinced,33 and Merck agreed to provide warnings about cardiovascular risks while it continued to monitor cardiovascular safety in additional clinical trials of Vioxx for new indications.34 Meanwhile, millions of patients took Vioxx, many of whom were at low risk of gastric side effects and could have received the same benefits at less risk and at lower cost from one of the older non-selective NSAIDs.35
While Merck's clinical trials proceeded, Dr. David Graham from the FDA Office of Drug Safety began a collaborative study with Kaiser Permanente comparing health records of patients who took Vioxx with records of patients who took other NSAIDs. That study showed significantly more heart attacks in the Vioxx patients,36 leading Kaiser Permanente to reconsider whether to provide coverage of Vioxx.37 But according to Dr. Graham's Congressional testimony, FDA sought to suppress publication of the study.38 Dr. Graham explained that FDA's primary institutional mission is approving new drugs, not re-evaluating already approved drugs. Moreover, FDA has long favored clinical trials over observational studies.39 Both of these factors favor reliance on the data submitted by drug companies over that coming from other sources with different motivations.
The same cardiovascular effects that showed up in the extensive Kaiser Permanente data eventually became too clear to overlook even in the data from the smaller number of patients enrolled in the ongoing Merck clinical trials.40 Shortly after the Kaiser Permanente data were presented at an international conference, Merck voluntarily agreed to withdraw Vioxx from the market,41 and under the intense glare of Congressional and media attention, FDA eventually allowed Dr. Graham to publish the Kaiser-Permanente study in a leading medical journal.42 Once again, FDA took no action until the drug manufacturer came around to the same conclusion as the payer.
The Vioxx episode showed the potential of large-scale observational studies to illuminate questions that were left ambiguous in data from drug company clinical trials. Healthcare payers have the necessary data for observational studies and face different incentives than drug companies. The availability of data not controlled by drug companies opens the door to analysis that is free of the possible distortions and wishful thinking of a company that is making billions of dollars a year selling a blockbuster product. FDA has long treated data from clinical trials as proprietary information belonging to the drug company that paid for the trials, and has therefore prevented public scrutiny of data that drug companies submit to it.43 But data from patient health records are free of the proprietary rights of the drug companies and could be analysed by other parties with different interests, such as Kaiser Permanente and its collaborators.
These two stories raise issues discussed in more depth throughout this paper. Payers have the ability to innovate, and crucially have different resources and different incentives than product developers. But they face substantial barriers in implementing their innovations. The regulatory and intellectual property systems are both geared to innovation by product developers, not to innovation by payers. Nevertheless, the role of payers in healthcare innovation is growing.
Payers enjoy one striking advantage as potential innovators: they possess tremendous amounts of valuable health data about treatments and consequences. So far, the longest term and most readily available form of payer data is administrative claims data used for billing and payments. Looking to the future, widespread utilization of electronic health records (EHRs) could provide payers with access to richer and more extensive data.
Administrative claims data, which include the information necessary to process payment claims, provide a view of medical encounters over time. These data could potentially answer many questions. Administrative claims data typically record diagnoses and treatments, hospital admissions and releases, laboratory test results, prescriptions filled, and professional services provided, as well as demographic information about patients (such as age, sex, and location) and the identities of providers.44
Payers also frequently have access to other data sources that can supplement administrative claims data. Prescription payment records typically reveal when patients actually pick up and pay for drugs (as opposed to when a doctor writes the prescription), and when they refill prescriptions. Payers typically know when physicians refer patients to specialists and why. They may have contracts with laboratory test providers that give them access to laboratory test results, especially for tests performed by major national providers rather than in house.45
In addition, it is increasingly common for payers to have access to patient medical records generated by doctors and other caregivers. These records can provide richer data on treatment and outcomes than administrative claims data, although analysing them can be challenging due to variability across providers in what is included and how they are written.46
Integrated health systems that combine the functions of payer, healthcare coordinator, and healthcare provider are particularly likely to have access to medical records. These integrated systems include Kaiser Permanente,47 Geisinger,48 Highmark, Intermountain Healthcare,49 and the Departments of Defense and Veterans Affairs.50 In the Kaiser system, for example, members pay premiums to Kaiser and see doctors who are Kaiser employees in Kaiser offices and hospitals.51 Integrated health systems may have centralized custody of records that are likely to be dispersed across multiple custodians in the case of patients covered by traditional health insurance plans.52 In the US health system, a relatively small fraction of patients belong to integrated health systems, but these systems have been important participants in research to date using data from EHRs.53 Single payer healthcare systems outside the USA may provide additional aggregated data sources to illuminate some research questions.54 Payers may use these data for their own research, provide them to other researchers, or enter into collaborations with others to use the data for innovation.55
Payers are not passive recipients of data; they can control the content of the data that become available to them through coverage determinations that lead caregivers to document diagnoses, treatments, and outcomes of interest. In the past, public and private insurers have often sought to exclude experimental technologies and to limit coverage to services that are ‘reasonable and necessary for the diagnosis or treatment of illness or injury’.56 Declining to cover new technologies may save costs in the short run, but could be a missed opportunity to collect data on the effects of these technologies in clinical practice. But payers are changing their behavior to capitalize on this opportunity. Medicare, for instance, has expanded coverage of some experimental technologies in order to facilitate assessment of their appropriate use and effectiveness.57 In 2000, Medicare began covering routine healthcare costs for Medicare patients enrolled in clinical trials, though it did not cover the cost of the investigational item or service itself.58 In 2006, Medicare formalized an approach called ‘coverage with evidence development’ (CED) that it had used in a handful of prior cases to provide coverage of experimental procedures for the purpose of generating data to assess their appropriateness.59 Other healthcare systems have used similar coverage determinations linked to participation in research studies.60 Studies supported by CED produced data that saved both costs and lives by preventing promising but unproven (and ultimately harmful) treatments from becoming the standard of care.61 Private payers have also sometimes chosen to cover certain technologies for the purpose of collecting data to assess clinical validity, utility, and cost-effectiveness.62
The opportunities and incentives of payers could make it profitable for them to engage in valuable forms of innovation that are underprovided by other innovators. Payers could use their data to improve the quality of care and to decrease costs, thereby potentially gaining competitive advantages.63 In particular, payers stand to benefit from identifying harmful effects of treatment and from comparing the effects of different treatment options. Their data could be useful for both traditional comparative effectiveness research and new research in precision medicine enabled by advances in genomics and information technology. Payer innovation efforts like United Health's Optum64 or Anthem's HealthCore65 include both internal research and work for other entities like pharmaceutical companies or other payers.66 This innovation offers potential benefits for patients and payers alike.
Drugs frequently have a wide range of side effects that have not yet been fully identified when they are initially approved for sale. Payers are especially well positioned to identify these side effects, which may sometimes change the determination that the drug is safe and effective.
Side effects often go unnoticed before approval because of limitations in the clinical trial process. Clinical trials typically involve only a few thousand patients, and occur over the course of a few months to a few years.67 The relatively small test population means that drug developers are unlikely to observe toxic effects that occur in only a small fraction of patients, or in a population not included in the clinical trials.68 Enrollment criteria for clinical trials often exclude patients who are pregnant, elderly, children, or taking other medications, for example, and therefore provide no information about the effects of the study drug in these excluded groups.69 Similarly, the relatively short duration of clinical trials makes it difficult for developers to observe long-term effects. As a result, for one in five approved drugs FDA eventually requires at least one new ‘black box warning’—the strongest type of warning—after approval.70 Of the drugs that add black box warnings after approval, it takes an average of 10 years before the effect is confirmed and the warning is added.71
Once a drug has been approved and is in clinical use, payers begin to accumulate longer-term observational data that permit them to observe previously unnoticed drug toxicity effects. The Vioxx example illustrates the potential of this type of payer innovation.
Payer records are not the only way to learn of post-approval drug toxicity. Side effects may become apparent in the course of further clinical trials conducted by the seller of the drug, as happened in Merck's clinical trial of Vioxx for a new indication.72 Drug manufacturers, doctors, and patients may also report side effects to FDA.73 But this passive reporting system depends on someone making a connection between the adverse event and the drug and going to the trouble of reporting it. Such reports are unlikely to provide information on increases in the frequency of otherwise common ailments, such as the cardiovascular side effects among patients who took Vioxx.74
After the Vioxx episode, Congress fortified FDA's authority to require drug manufacturers to conduct postmarket studies.75 At the same time, Congress directed FDA to establish a system for monitoring drug adverse events through use of health records, a mandate that FDA is implementing in its Sentinel program, as discussed below.76 Some foreign health agencies have similar programs.77
But while these programs give regulators access to data from a network of payers, the data can only answer the queries that someone thinks to ask. FDA continues to rely primarily on adverse event reports to identify new risks. Payers with an interest in lowering the costs and improving the quality of healthcare have an opportunity to play an active role in identifying additional appropriate queries for the Sentinel System by scrutinizing their own data for evidence of drug toxicity, either ahead of regulators or in partnership with them. Although in the past FDA showed reluctance to rely upon Kaiser-Permanente data to demonstrate the cardiovascular side effects of Vioxx, since that time Congress has given FDA a legislative mandate to use such data for safety monitoring, opening the door to a larger payer role in informing post-approval regulatory decisions.
Comparative effectiveness and cost-effectiveness
Payers are in an excellent position to study the comparative effectiveness or cost-effectiveness of different treatment interventions. Comparative effectiveness research compares health outcomes for different interventions whereas cost-effectiveness research further considers costs to determine which intervention buys more health for the money.78 Comparing the effects of different interventions is a valuable form of research that is often neglected in the premarket stage. Premarket clinical trials typically compare a new drug with a placebo rather than with another intervention, 79 unless the drug developer seeks approval to make specific marketing claims of superiority to alternative treatments.80 As a result, they provide little information about whether the new drug is better or worse than alternative treatments. Comparative effectiveness studies may involve clinical trials, in which researchers randomly assign patients to receive one drug or the other, or data-based observational studies, in which researchers observe differences in outcomes between matched populations of patients that received each course of treatment.
Payers, both public and private, are in a good position to conduct comparative effectiveness research through observational studies. As previously noted, they have access to large data sets of patient records, including information about diagnoses and drug prescriptions and purchases.81 Although administrative claims data may not indicate how well the intervention worked (beyond such crude indicators as hospital readmissions), patient health records may include richer data about outcomes.
Moreover, cost-sensitive payers have strong incentives to perform comparative effectiveness—and especially cost-effectiveness—research. Payers pay for care, and paying more for the same care, or the same for worse care, is bad for their bottom lines. Cost-effectiveness research can help make care cheaper and better. For example, Mayo Clinic researchers used Optum Labs data to determine that newer anticoagulant drugs have a higher risk of gastrointestinal bleeding among older patients.82 Despite the greater convenience of the newer—and more expensive—drugs, this risk may make these drugs less appropriate for older patients than older, safer products.83
Other performers of comparative effectiveness research face different incentives and constraints. They generally need to partner with payers for access to data. Doctors and hospitals have some access to health data, although they still may find it advantageous to partner with payers to obtain access to larger data sets that include data from different providers. But doctors and hospitals may have perverse incentives under a classical fee-for-service model, because they make more money by providing more (and more expensive) treatments.84 Drug companies could also conduct comparative effectiveness research, through both clinical trials and observational studies, and have an incentive to demonstrate that their new products are better than older drugs. However, comparative effectiveness research runs the risk of showing that a new drug is worse than existing treatments. Since placebo-controlled trials are generally enough to win regulatory approval, drug companies may decide not to take the risk of demonstrating inferiority rather than superiority for the patent-protected product. Academic institutions,85 non-profit organizations,86 and government-created comparative effectiveness institutes focus on public health goals rather than cost control. In fact, the Patient-Centered Outcomes Research Institute created by the Affordable Care Act is arguably prohibited by statute from performing certain types of cost-effectiveness research.87
Payers have different incentives which could make them an important source of comparative effectiveness research and cost effectiveness research to balance the research of other research performers with different interests. Considered in isolation, payers’ cost-cutting incentives may seem problematic; they may have hesitate to demonstrate that a new, more expensive drug is better or safer than cheaper alternatives. But considered alongside the incentives of drug manufacturers to demonstrate the superiority of costly new products, payer innovation could provide a healthy corrective. Overall, the competing incentives of different stakeholders provide counterweights that may yield a more balanced understanding than reliance on data from any one kind of innovator.
Payers can also contribute to evaluating (and perhaps supporting) off-label use of drugs. Pre-approval clinical trials often focus on relatively narrow indications to simplify the showing of efficacy and safety necessary to get regulatory approval. But once a drug becomes available, doctors are free to prescribe it for other purposes that are not indicated in the FDA-approved product label. In some fields, such as oncology, off-label use of products for indications beyond the scope of FDA approval is quite common.88 Drug companies may have little incentive to conduct costly clinical trials to provide evidence for off-label use, especially once such use enters into widespread practice; at this point, firms already benefit from increased drug sales without having to incur the costs and risks of further trials. Many off-label uses are, unsurprisingly, unsupported by rigorous evidence, even when they have become the standard of care.89
FDA has long sought to motivate drug companies to conduct further clinical trials of off-label uses by preventing firms from promoting their products for off-label use. FDA has taken the position that promotion of a product for off-label uses renders the product ‘misbranded’ in violation of the Food, Drug & Cosmetic Act.90 But recent judicial decisions have questioned this statutory interpretation and held that the First Amendment protects drug companies and their sales force from criminal prosecution for promoting off-label use (so long as they make no false statements).91 Moreover, once a generic version of the drug is available, the original sponsor has little incentive to invest in costly clinical trials of off-label uses for a product that is no longer profitable.92
Payers have an incentive to ensure that off-label uses are effective and supported by evidence, because ineffective uses waste money.93 They also have the data to observe the effectiveness of off-label uses that have already entered into practice. Observational studies in payer health records may provide a more cost-effective alternative for filling the information gap about the effects of off-label uses of drugs.
Prevention and long-term effects
Pre-approval clinical trials are necessarily limited in duration, and thus have limited value in determining long-term health effects. We noted above that clinical trials may fail to reveal toxic side effects that manifest over time.94 For some products, such as vaccines and other prophylactic measures to prevent disease or forestall its progression, long-term effects are critical for determining not just safety, but also efficacy.95 In recent decades, FDA has adapted its regulatory approach to permit approval of some products on the basis of data on ‘surrogate markers’ rather than requiring that trials continue for years to measure disease endpoints.96 This allows products to get to market that might otherwise not be approvable under a more rigorous application of standards for proving safety and efficacy prior to approval. But although it might not be commercially feasible to require that clinical trials continue for many years, the lack of data on clinical endpoints is a significant gap in the information base for determining appropriate clinical use of these products, especially since many surrogate endpoints are eventually found to be poor predictors of clinical outcomes.97
Payer data on clinical outcomes can provide a valuable and cost-effective supplement to the limited data available from clinical trials. A recent example that illustrates the potential for payer clinical data to show the long-term value of prophylactic treatment is a study of the pre-exposure prophylactic use (known as PrEP) of antiretroviral drugs using data from Kaiser-Permanente in San Francisco.98 In that study, not a single person using PrEP became infected with HIV.99 This study is notable because payer data confirmed that a potentially costly treatment is valuable, rather than indicating that a costly product should be used more sparingly.100 When payers may be on the hook for more costly future medical care, they may benefit financially from more extensive use of prophylactic treatment that forestalls the need for future care.101
Precision medicine, also known as personalized medicine and frequently touted as the future of medicine,102 aims to provide ‘the right patient with the right drug at the right dose at the right time’.103 It identifies biological variation among patients and correlates that variation to differences in the most effective and efficient treatment.104 An early success story for precision medicine was the use of a test to identify those patients that could benefit from the breast cancer drug Herceptin, a drug that is effective only against tumors that overexpress a particular gene called HER2/neu.105 A simple genetic test can measure whether a patient's tumor overexpresses the gene, allowing providers to give the drug only to patients with tumors that are likely to respond to it, while sparing other patients from exposure to unnecessary side effects.106 Precision medicine may also help determine the appropriate dose of a drug based on patient sex, weight, and genetic makeup,107 or predict which patients might benefit more or less from hospital admission.108 Research is underway to explore more complex and sophisticated precision medicine implementations.109
Much of this research uses genomic data and biomarkers.110 An individual's genome—the sum of his or her genetic information—represents a tremendous amount of biological variability; understanding genomic variations may allow individualized predictions of how an individual may metabolize certain drugs111 or how likely the individual is to develop a certain type of cancer.112 Aside from the DNA in a patient's own normal cells, the genetics of viruses, bacteria, and cancerous tumors can inform the treatment of related diseases.113 Other biomarkers, such as blood-sugar level, the amount of prostate-specific antigen, or the previously mentioned overexpression of HER2 by a tumor, can also be used to direct treatment (for diabetes, prostate cancer, and breast cancer, respectively).
Payers are using their data in precision medicine research. Optum, for example, is involved in developing predictive analytics technology to identify high-risk patients based on a combination of administrative claims data and real-time clinical data from multiple sources.114 These data may reveal patterns that suggest which drugs or treatments work best for which patients, and which patients should avoid treatment altogether in particular circumstances. Payers may have direct access to tissue samples (or analyses of those samples) to determine biomarker, genetic, and genomic status; if they do not, they may be well positioned to collaborate with other researchers or caregivers to link health records to tissue samples.115 In fact, payers are essential participants in the eMERGE network, further discussed below.116
Payers could further contribute to precision medicine research, and potentially reap further benefits, by providing coverage of genetic diagnostic tests that make their data more informative. The selection of appropriate treatments for particular patients requires the use of validated markers for distinguishing those patients who will benefit from a particular treatment. Validation of tests requires data, but many payers decline to cover genetic tests that have not yet been validated as clinically useful. This Catch-22 may stymie the development of new diagnostic tools.117 By covering the cost of promising new tests while collecting data to test their validity, payers could accelerate the development of diagnostic products that might eventually guide more appropriate use of costly therapies but that would otherwise never become available for lack of validation.118
The incentives of payers may offer a useful counterweight to the incentives of the drug companies that have become key drivers of precision medicine.119 For drug companies, precision medicine presents a trade-off between more reliable treatment and smaller market size. If research shows that a particular drug only works for a third of people taking it, and provides a mechanism for identifying those patients, the other two thirds will no longer use the product, and sales will decline.120 Precision medicine research on existing drugs may therefore be an unattractive proposition for drug sponsors.121 For payers, on the other hand, broader implementation of precision medicine could improve healthcare quality and reduce costs. A payer, for example, might save costs by demonstrating that two-thirds of patients currently taking an expensive drug would be better off taking an older generic drug or other less expensive treatment—or no treatment at all.122
Of course, payers and drug companies face the risk that observational studies will not yield the results that are best for their bottom lines. In both cases, financial incentives are likely to inform the research questions that they pursue, and perhaps to influence their analysis of results and their decisions about what results merit publication. The participation of payers as innovators in the field of precision medicine is thus likely to yield a more balanced and complete picture than would emerge from a field dominated by drug companies alone.
The use of payer data for innovation presents substantial technical challenges.123 Some challenges involving the storage and analysis of data are not unique to healthcare and therefore benefit from overall improvements in information technology.124 Particular concerns in the healthcare field revolve around data availability, data quality, data assembly, and data interoperability.
First, data must be acquired and assembled. As discussed above, payers have direct access to some data, principally administrative claims data, and prescription data, and may have indirect access to hospital admissions/releases, laboratory testing data, and provider records of clinical care.125 It takes time, money, and technical expertise to bring these data together, to link them by patient and demographic information, and to structure the assembled data to permit meaningful analyses.126 Even when firms have access to data from different sources, the fragmented nature of the healthcare system means that those different sources will cover different populations of patients. For example, although Optum's Data Warehouse has health data for over 150 million unique patients, it has the combination of claims, prescription, and clinical records for fewer than three per cent of those patients.127
For some studies, it is necessary to assemble comprehensive data not only across different patients in a population, but also across different periods in the lives of particular patients.128 Longitudinal data—that is, data that follow patients over long periods of time—are useful for measuring the effects of preventive treatments, long-term drug effects, and interactions between treatments, and for answering other important medical questions.129 But the records of any one payer frequently cover only a relatively limited span of a patient's life. Patients frequently switch their insurance coverage, whether because they change to a new job with a different set of payer options,130 change payers while staying at the same job (perhaps because the employer changes the plans it offers), change plans on the individual market, or become eligible or ineligible for Medicaid based on fluctuating income. The largest change comes when patients turn 65 and become eligible for Medicare. In any of these situations, one payer stops collecting data about that patient, and another begins. Some patients, of course, stay with the same payer for decades; in that case, the records of a single payer may provide long-term information without the need for aggregation. But this is rare; in one large data set, administrative claims data for more than 5 years were available for only about 15 per cent of patients.131 For most patients, assembling a longer-term record of information may be necessary to provide useful data for long-term studies.
Some regional efforts are already trying to overcome the challenge of fragmented data to allow caregivers to exchange patient information more readily. One example, Cal INDEX, is a non-profit health information exchange132 founded in 2014 with seed money from two major payers to store centralized, comprehensive patient information for the vast majority of patients in California.133 Providers choose whether to join the exchange,134 and their patients participate unless they opt out.135 So far Cal INDEX has had difficulty persuading providers to participate,136 perhaps because payers are reluctant to share data with competitors.137 If Cal INDEX can overcome this obstacle, it may succeed in resolving the problem of cross-provider data fragmentation, at least within California.138 A patient's single longitudinal patient record would include both clinical and administrative claims data from multiple sources even as the patient shifts providers and payers.139 Cal INDEX's stated purposes are to improve care and to increase efficiency, but it recognizes its consolidated data set could also be a useful resource for research.140 Data consolidation could arise as a side effect of industry consolidation.141 As payers merge their operations, they may also merge their data. Although industry consolidation has arguably detrimental effects for consumers,142 data consolidation could be an unexpected benefit.143
Second, and related, data from different sources must be interoperable—that is, they must be in compatible formats so they can be joined and analysed together.144 There is no standard format for EHRs or administrative claims data, and data from different systems are typically kept in different formats.145 Moreover, some payers have changed from one data system to another over time. This means that any effort to aggregate data must translate data from one proprietary format to another.146 Some pieces of information may be present in one system but not another; other information may be coded differently (eg numerical versus qualitative judgements) or using different standards (eg signifying different ranges as ‘high’ or ‘low’). Some of these barriers may arise through inadvertence, but there is also evidence that some developers of EHR systems may use proprietary formats to stymy aggregation and use of data from other systems.147 Further complicating the interoperability problem, as described below, data about different kinds of conditions may be subject to different privacy regimes, with some especially sensitive information covered by special laws.148
Third, ensuring and maintaining the quality of data is difficult.149 Especially with administrative claims data, information essential to receiving payment may be coded in ways that reflect financial incentives.150 Because insurance requires certain diagnoses or procedures to reimburse for physician services, healthcare providers may have incentives to code those data in marginal or inappropriate situations, leading to biased data.151 In addition, some health terms are inherently imprecise, such as ‘overweight’ or ‘high’ blood pressure, and may carry different meanings to different practitioners; attempting to distill imprecise categories into numerical variables can introduce errors if not done carefully and consistently. Finally, even with adequate care and effort, errors exist in all sources of data, and entities using those data for analysis need to account for that error.152
None of these obstacles are insurmountable. They do, however, suggest targets for the attention of policymakers who want to facilitate use of healthcare records in research.
ECONOMIC AND LEGAL OBSTACLES
Although payers are in a good position to play a larger role in healthcare innovation, their incentives to invest in innovation are constrained by a number of economic and regulatory features of the healthcare market. First, some quirks of healthcare markets and tax law directly reduce incentives to control costs. Second, because payers typically do not directly control care, they may fail to realize the full cost-saving benefits from their innovation. Third, intellectual property rewards are less available for the innovation opportunities available to payers than they are for new therapeutic product innovations. Fourth and finally, privacy laws restrict access to and use of health information in research.
Market Quirks and Tax Preferences
Cost sensitivity should motivate payers to invest in developing or identifying more cost-effective treatments and in curtailing overuse of costly products. However, the US market for healthcare and insurance has complexities and idiosyncrasies that blur these incentives.153 We note four features in particular: muted competition, passed-on costs, tax subsidies, and medical loss ratios (MLRs).
First, payers may face muted competition for a number of reasons. Industry consolidation may give payers some power to dictate the terms of their coverage and the rates they charge.154 Moreover, employers and individuals may tend to stick with the payer they currently use rather than shopping for competing products.155 Finally, purchasers may have difficulty understanding the differences among insurance products, further diminishing effective competition.156 The complexities of this market are beyond the scope of this paper, but it stands to reason that diminished competition in the health insurance market would decrease competitive pressure on payers to invest in cost-lowering innovation.157
Second, the combination of weak market competition and weak oversight of price increases by insurance regulators may allow payers to pass on increased costs to their customers with relative ease through increased premiums,158 although the Affordable Care Act has introduced some limits on the ability of payers to raise premiums in a deliberate attempt to make payers more cost-sensitive.159
Third, tax subsidies for health insurance may dampen incentives for frugality on the demand side of healthcare. Health insurance premiums paid by an employer are both fully deductible by the employer as a business expense and excluded from the employee's taxable income.160 In this system, the government shares the costs of healthcare, diminishing the interest of patients and their employers in cost-lowering innovation and making it easier for insurers to pass rising costs along to them in the form of higher premiums.
Fourth and finally, the complex dynamics of the Affordable Care Act's MLR provisions may reduce incentives for cost-lowering innovation. Under those provisions, payers must pay 85 cents in medical expenses for each dollar received in premiums.161 This sets a ceiling on the increase in profits to be gained by lowering costs; the total of profits plus administrative expenses can be no higher than 15 per cent of total insurance premiums. Cost-lowering innovation may still be profitable if increased efficiency in one area offsets other rising costs. Moreover, expenditures on quality improvement research count as part of the ‘medical expense’ and may thus help bring that total up to 85 per cent. On the margin, however, this cap may reduce incentives for innovation. Overall, these features of the health market likely combine to lower incentives to innovate toward efficiency.
Complexities Implementing Innovation
Another factor that makes it challenging for payers to profit from innovation is that many payers do not actually provide care—they just pay for the care that others provide. For these payers to benefit from innovations around quality, efficiency, and medical targeting, they need to influence healthcare providers to actually adopt those innovations. Integrated health systems, which both provide and pay for care, may find it easier to control the behavior of providers. But in a fee-for-service system, providers face perverse incentives to use more and costlier treatments, thereby increasing their own remuneration. These incentives are in serious tension with the goals of frugal payer innovation.162 Payers have several potential mechanisms to address this tension, including direct incentives, knowledge-sharing, contracts with product developers, and regulatory influence.
Direct control and incentives
Traditional payers might in theory require caregivers to follow prescribed procedures or use utilization review and reimbursement tiering to guide physician behavior, though these practices have had a contentious history.163 Payers can also try to align provider incentives with cost-saving goals by using financial incentives or risk sharing. When providers are compensated on a traditional fee-for-service basis, the interests of providers are opposed to those of payers: increased treatment costs mean increased provider compensation, and providers have little motivation to pursue efficiency. On the other hand, when the incentives of providers are aligned with those of payers, providers have greater incentives to adopt payer innovations, which in turn gives payers greater incentives to invest in innovation.
This may be why integrated providers such as Kaiser Permanente have been more active participants in payer innovation than traditional insurers. It may be easier to implement cost-saving innovations through caregivers who are salaried employees with nothing to gain from the provision of costly and excessive care. The Affordable Care Act aims to achieve similar alignment of incentives for frugality through Accountable Care Organizations, coordinated groups of physicians, hospitals, and other providers.164 Among other benefits, these structures allow physicians to share in the financial benefits of frugal care, rather than rewarding them for providing costly care under a traditional fee-for-service system.165 More broadly, the Affordable Care Act aims to shift a substantial fraction of care aware from fee-for-service toward value-based payments or other frugality-focused payment models, which should further align the incentives of payers and providers and enable smoother implementation of demand-side innovation.166
Payers can also influence provider behavior less directly through collaborative sharing of knowledge. They can publish the results of their studies and work to establish best practices, including clinical guidelines or ‘critical pathways’ reflecting treatment patterns that their research shows to be both effective and efficient.167 Within the medical community, expert committees—typically well-known physicians—rely on published literature to develop treatment pathways.168 Payers may influence these committees by contributing their studies to the published literature. Payers may be more effective in influencing clinical practice when they collaborate with influential clinicians to conduct and publish observational studies, before providing them to expert communities that can then establish standards of care. Such collaborations are a feature of the PCORnet and eMERGE networks.169 Nonetheless, to the extent that providers resist following new clinical guidelines, they diminish the benefit to payers from investments in innovation.170 Moreover, if payers must change the standard of care in order to realize the benefits of their investments in innovation, they will necessarily share those benefits with competitors who did not share their costs.171
Contracts with product developers
Another new strategy, increasingly popular with payers and drug companies, involves private agreements for drug companies to pay a rebate to a payer if a new product fails to meet specified performance targets.172 For example, Harvard-Pilgrim recently entered into a ‘pay-for-performance’ agreement with Amgen to provide coverage of Amgen's new cholesterol drug Repatha at an undisclosed discounted price.173 Harvard-Pilgrim agreed to exclude from its formulary (ie its list of reimbursable drugs) other drugs in the same class of PCSK9 inhibitors, while Amgen agreed to pay rebates if the drug fails to reduce cholesterol to specified target levels for different patient groups and if total payments exceed a target.
Such agreements may stimulate private innovation by leading the parties to invest resources in measuring future health outcomes and analysing the data to determine whether the targets have been met. At the same time, they could confine cost-saving benefits to the particular firms that are parties to the agreement, without changing prices to other payers and without changing the standard of care to bring about more widespread changes throughout the healthcare system.
According to a recent industry analysis, these agreements face legal obstacles under federal and state fraud and abuse laws, the federal Anti-Kickback Statute, and Medicaid rebate provisions.174 Apart from these legal obstacles, implementing these agreements may be challenging in that they require parties with competing financial interests to agree on appropriate criteria and procedures for measuring performance and to cooperate in sharing and interpreting data.175 If successfully implemented, these agreements could do a better job of motivating drug companies to optimize health outcomes rather than to maximize total sales, at least for patients who are covered by payers who have entered into such agreements.
The net effects of these agreements on payer innovation incentives are less clear, not least because they are quite new. Perhaps the prospect of recovering rebates would cause some payers to invest resources in tracking health outcomes that they would otherwise ignore, generating valuable knowledge about the effects of drugs in patients. On the other hand, payers and drug companies would both stand to benefit from keeping this knowledge and any resulting payments secret. Secrecy would reduce the social value of the knowledge relative to a system in which payers can only profit by publishing the results and changing the standard of care. Agreements with drug companies may also distort the kinds of research questions that payers ask. For example, if payers like Optum agree to put only one PCSK9 on their formulary of preferred drugs, they may be unable to use their data to do comparative effectiveness studies of different drugs in the same class. The overall impact of these agreements on incentives for payer innovation is thus a complex question.
Payers might also influence the behavior of providers by using their data to influence regulatory decisions. FDA decisions, for example, determine what products may be sold and what indications and warnings appear on product labels. These determinations in turn influence provider behavior. The Center for Medicare and Medicaid Services (CMS) further influences provider behavior through its determinations about what products and services Medicare will cover. Private payers often follow Medicare coverage determinations in their own policies. But they could play a more active role in influencing which treatments or tests Medicare covers as they accumulate data from clinical experience.
Initially, only product developing firms have information about the effects of their unapproved new products. FDA new drug approval decisions thus inevitably rely heavily on data from product sponsors to show safety and efficacy. As clinical experience accumulates, however, payers may have a larger role to play. Data from clinical experience may show product risks that call for fortified warnings or even withdrawal of product approvals, as in the Vioxx case. On the other hand, they might indicate that a product currently available by prescription only is safer than substitutes that are already available over the counter, as in the case of non-sedating antihistamines. These examples revealed a regulatory regime that gave product sellers substantial control over the information available to regulators while making it difficult for other stakeholders to influence regulatory decisions.176
In the future, however, payers might use their data to reveal risks to FDA that it should study through the Sentinel System, perhaps leading to future warnings or even withdrawal of product approvals. Legal changes after the Vioxx incident strengthened FDA's ability to take precisely this type of data into account. Payers could also provide information about off-label use of drugs that could eventually be used to win FDA approval for those uses.177 Such regulatory moves might have a greater impact on the behavior of caregivers than the exhortations of payers.
Intellectual Property Incentives
Payers also face diminished intellectual property incentives for innovation relative to the incentives of product-developing firms. In a familiar story, intellectual property provides legal excludability to solve the public goods problem that would otherwise prevent innovators from capturing the full value of their investments. By allowing innovators to exclude competitors from using their information goods, intellectual property permits them to raise prices, thereby increasing incentives to innovate. Intellectual property incentives work reasonably well for the producers of new products. But patents, trade secrecy, and regulatory exclusivity offer few direct benefits to demand-side innovators (although high prices for patent-protected products may have the indirect effect of motivating payers to invest in learning how to use these products more sparingly). The forms of payer innovation considered above are pure information goods; there is typically no new physical product that the payer can sell as a result of the knowledge gained from observational studies of patient health records, for example.178 Intellectual property is a poor fit for appropriating and monetizing the value of this knowledge; secrecy is ineffective and inappropriate, and patents are largely unavailable.179 More fundamentally, the excludability at the center of intellectual property is not a viable option for some types of payer innovations discussed above.180
The first and most obvious way to appropriate an information good is to keep it secret; if others do not have the information, they cannot use it. This strategy is ill suited to payer medical information, because payers must at a minimum share the information with caregivers before they can put it to use in a clinical setting. Caregivers in turn are required to obtain informed consent for medical treatment, which may require further disclosure of the information to patients. Broader disclosure may be necessary to bring about a change in the standard of care. For example, if payer studies indicate that caregivers should not continue to provide a form of treatment that is considered the standard of care in the medical community, caregivers may fear potential malpractice liability for withholding the treatment.181 Widespread disclosure of the study results may therefore be necessary to facilitate clinical implementation of changes in the standard of care. Secrecy may thus be a serious obstacle to effective use of payer innovations.
Patents on comparative effectiveness research results or precision medicine information are likely unavailable, unenforceable, and impractical. Judicial limitations on what sorts of inventions constitute patentable subject matter have cast considerable doubt on the patent eligibility of algorithms for selecting medical treatments for patients.182 Standard patent law rules about prior art prevent patenting the treatment options themselves. Because observational studies of health outcomes necessarily involve treatments that are already a part of current practice, those treatments could not be patented because they are already in public use, on sale, and likely disclosed in published literature and prior patents.183 Even if these innovations were patentable, the patents might be difficult to enforce for at least three reasons: first, it would be difficult to observe and police infringing behavior in light of the privacy of health records;184 second, medical practitioners performing medical activities have a statutory exemption from patent infringement remedies;185 and third, suing doctors to prevent them from practicing medicine more effectively might create a public relations problem for a healthcare payer.
Although intellectual property does not provide the same direct incentives for medical innovation by payers that it provides for product sellers, it may provide an important indirect motivation for payers by increasing the costs they incur in covering patented products. When Kaiser Permanente collaborated with FDA to study the cardiac side effects of Vioxx, payers were collectively paying $2.5 billion per year for Vioxx, creating a conspicuous opportunity to cut costs by reducing the use of Vioxx.186 Payers may be less interested in studying the effects of less costly treatments that are already off patent, except in comparative effectiveness studies that offer the prospect of lowering costs incurred for coverage of a higher-priced alternative. In this indirect sense, the law of intellectual property is likely to structure the incentives of payers toward more scrutiny of the clinical benefits of patented treatments.
Privacy of Health Information
HIPAA aimed to facilitate the flow of information for healthcare and administrative purposes (such as claims processing), while protecting patient privacy by restricting uses and disclosure for other purposes.189 The Department of Health and Human Services has elaborated upon these general statutory provisions in detailed rules, including a Privacy Rule.190 The Privacy Rule sets limits on disclosure and use of ‘protected health information’ by ‘covered entities’191 and their ‘business associates’.192 ‘Protected health information’ includes both medical and billing records.193 The baseline rule under HIPAA is that all use or disclosure of protected health information is prohibited unless it is specifically allowed.194 In addition, the Privacy Rule requires reasonable efforts to limit uses or disclosures of protected information to ‘the minimum necessary’ to accomplish the intended purpose.195
Complicating the picture, different kinds of health information are subject to different rules. The Privacy Rule itself provides additional protection for psychotherapy notes196 and allows other more stringent privacy protections under various state laws.197 Some state statutes, for example, provide additional protections against disclosure of information related to HIV status and treatment.198 Other federal statutes provide additional protection for genetic information199 and substance abuse treatment records.200 This uneven landscape of privacy restrictions, varying by state, by condition, and by information type, further complicates the challenge of assembling broad, comprehensive, longitudinal health records.
The Privacy Rule nonetheless allows some room for innovators to use health data. Some uses of protected health information are generally permitted, while specific waiver and authorization provisions may enable normally prohibited uses.
Normally permitted uses
The wording of the Privacy Rule creates considerable confusion about when the study of healthcare records to improve future patient care is allowable. The Privacy Rule permits a covered entity to use or disclose protected health information ‘for treatment, payment or health care operations’.201 Although this explicitly includes ‘quality assessment and improvement activities, including outcomes evaluation and development of clinical guidelines’, such activities may not have the primary purpose of ‘obtaining  generalizable knowledge’.202 These provisions are tough to reconcile, since it would seem irresponsible to develop clinical guidelines on the basis of anything short of generalizable knowledge.203 At a minimum, one might expect that as the analysis of health outcomes to improve clinical care becomes more scientifically rigorous (and its conclusions therefore more generalizable), it may look less like permissible ‘health care operations’ and more like restricted ‘research’.204 The 21st Century Cures Act requires the creation of a working group to study the creation of a full research exemption for the use of protected health information.205
De-identified data. De-identified data are not covered at all by the Privacy Rule, which applies only to ‘individually identifiable health information’.206 Although advances in information technology have made it increasingly easy to re-identify individuals on the basis of limited information,207 the Privacy Rule nonetheless provides a safe harbor that qualifies data as de-identified if seventeen pieces of identifying information are removed.208 De-identifying data is a key way to navigate HIPAA restrictions even for government entities.209
However, de-identification brings its own problems. The list of impermissible identifiers includes information that may be relevant to researchers, such as dates, ages, and biometric identifiers; excluding this information limits the value of the data.210 Moreover, retention of identifiers may be necessary to link data from different sources and over time.211 The most straightforward way to integrate information from different sources—a key technical challenge discussed above—is to use unique identifying information from individual records. If Miles Vorkosigan's records from different providers and payers are related only by the fact that those records all pertain to Miles Vorkosigan, the easiest way to link those records is through his name.212 Removing identifying information hinders this aggregation. There are technological ways to partially avoid this problem, but they can be complex and typically require some form of centralized infrastructure.213
Limited data sets. The Privacy Rule allows covered entities to use or disclose ‘limited data sets’ without the need for authorization ‘only for the purposes of research, public health, or health care operations’.214 A limited data set may include slightly more identifying information.215 However, a covered entity may use or disclose a limited data set only if it enters into a ‘data use agreement’ obliging the data recipient to use or disclose the protected health information only for ‘limited purposes’ permitted by the Rule.216
Public health activities. The Privacy Rule explicitly permits use or disclosure of protected health information for certain public health activities, including disclosure to a public health authority for surveillance purposes.217 This allows disclosures to FDA for postmarketing safety monitoring under the Sentinel Initiative.218 Another provision permits disclosure to sponsors of FDA-approved products for postmarketing surveillance and adverse-event reporting.219 This particularly provision may have been inartfully drafted; while it lets drug companies access information about their own products, it does not permit disclosure of data about other treatments that could serve as controls, limiting the possibility of use in comparative effectiveness research.220
The foregoing limitations221 on the Privacy Rule allow some use of healthcare information in research, although compliance with the conditions necessary to qualify for these limitations may be burdensome and may limit the scope of research.
Authorization and waivers
In addition to normally permitted uses, the Privacy Rule allows otherwise prohibited uses and disclosures in two circumstances. First, individual patients may authorize the use of their protected health information for any purpose, including research.222 However, the requirements for a valid authorization are exacting and limit possibilities for unplanned future research.223 In addition to these practical challenges, a requirement of getting individual authorizations presents a different and less tractable problem for research use. There may be significant medical differences between patients who are willing to authorize the use of their information and those who are not.224 Relying on individual authorizations could thus bias the results of observational studies, making them less informative than they could be.
Second, an Institutional Review Board or a Privacy Board225 may also waive the authorization requirement for individual research studies that meet strict specified criteria.226 This can mitigate the serious problem of selection bias, but the need for Board review imposes costs and delays. Survey data indicate that many researchers have found it very difficult to obtain Privacy Rule waivers.227 Moreover, ambiguity in the waiver criteria creates uncertainty, especially for studies that aggregate data from multiple sources and may therefore require approval from multiple Boards that may interpret the Privacy Rule differently.228
Despite these obstacles and gaps in incentives, some payers are already making notable efforts to advance the use of payer data for medical innovation. We suggest at least three reasons why this is so. First, as we noted above, for innovation related to costly patent-protected treatment, high costs give payers an incentive to reduce costs. Second, observational studies are relatively cheap, especially as compared to expensive clinical trials. Third, there are a variety of government initiatives under way that are partnering with payers and helping them to overcome obstacles and to kickstart their own research. In the next part, we explore some of these initiatives and consider additional mechanisms to encourage payer innovation.
ASSISTANCE FROM GOVERNMENT INITIATIVES
Although the challenges and obstacles to payer innovation are substantial, a number of federal legislative and regulatory initiatives are facilitating the use of health records in research. These efforts include providing incentives to promote the adoption and use of interoperable health records by caregivers and hospitals, creating a network of data sources for public health monitoring of postmarket drug safety issues, and sponsoring research initiatives in the areas of comparative effectiveness studies and precision medicine.
Electronic Health Records
The federal government has been actively promoting the use of EHRs for well over a decade in the hope of reducing medical errors, reducing costs, and improving the quality of care.229 Policymakers have also touted the potential for research use of EHRs as part of a ‘learning healthcare system’ in which caregivers continuously adapt their treatment choices in light of ever-expanding knowledge about healthcare outcomes.230
The healthcare industry has been extraordinarily slow to adopt information technology, lagging far behind the rest of the economy.231 For a variety of reasons, paper records and hard copies dominated health records well into the first decade of the 21st century.232 President George W. Bush called for computerizing health records in his 2004 State of the Union address,233 and followed up by creating a new Office of the National Coordinator for Health Information Technology (ONC) to pursue this goal, with a budget of $42 million.234 But progress remained slow.
Federal incentives to make use of EHRs were strengthened considerably in the Obama administration,235 largely as a result of the Health Information Technology for Economic and Clinical Health Act of 2009 (HITECH Act), passed as part of the American Recovery and Reinvestment Act of 2009.236 The HITECH Act charged the ONC with reviewing standards for health information exchange, coordinating the activities of the federal government concerning health information technology, certifying compliance with applicable standards on a voluntary basis, publishing reports, and disseminating financial assistance.237 The legislation also established committees to make recommendations to the ONC regarding health information technology standards and the creation of a nationwide health information technology infrastructure.238 It directed the Secretary of HHS to ‘assist health care providers to adopt, implement, and effectively use certified EHR technology that allows for the electronic exchange and use of health information’, disseminate best practices, and allow for the exchange and use of information in compliance with standards.239 And it provided $30 billion for incentive payments through Medicare and Medicaid to reward the adoption and ‘meaningful use’ of EHRs by providers and hospitals.240 Requirements to establish meaningful use increase over time, and after 2015, those who fail to make meaningful use EHRs are subject to penalties.
Use of EHRs increased significantly following the implementation of HITECH payment incentives,241 although this has hardly been an unqualified success story.242 Progress has been much slower in promoting health information exchange among providers. A major focus of the ONC in the years ahead is to achieve ‘a nationwide, interoperable health IT infrastructure’.243
HITECH-driven adoption of EHRs offers considerable potential benefits for research users. EHRs provide richer and more complete information than claims data, and are much easier to aggregate for use as research data than the paper records used by providers in the past. Although lack of interoperability is an ongoing problem,244 ONC is working to promote the development of interoperable EHR products that allow providers to share and access a common clinical data set according to common technical standards across a nationwide network.245 The networks and infrastructure that promote information exchange in the context of clinical care will also facilitate access and aggregation by researchers, as ONC recognizes.246 For research purposes, it may be possible to achieve considerable benefits without nationwide interoperability by using the records of a single large provider.247 Data quality may prove to be a more persistent problem in making research use of records that some observers claim are optimized for (or distorted by) billing purposes.248
Regulatory Use of Networked Data for Observational Studies
Payer innovators also stand to benefit from the infrastructure and technology developed to support the FDA Sentinel System, a legislatively mandated network of data sources and tools for postmarket monitoring of the safety of FDA-approved products. This initiative arose from a series of studies that convened a wide range of experts and stakeholders to come up with ideas for improving drug safety.
A series of high-profile drug safety cases (including the Vioxx episode)249 provoked members of Congress to ask the Government Accountability Office (GAO) to review FDA's organizational structure and decision-making process for postmarket drug safety.250 The GAO Report was highly critical of FDA's system of postmarket surveillance, noting that it was underfunded251 and relied too heavily on an unreliable system of adverse event reporting.252 The GAO Report concluded that ‘FDA will need to continue its efforts to develop useful observational studies and to access and use additional healthcare databases’ and recommended that ‘Congress should consider expanding FDA's authority to require drug sponsors to conduct postmarket studies, such as clinical trials or observational studies’.253
Around the same time, FDA and the Department of Health and Human Services asked the Institute of Medicine (IOM) to convene experts to assess the US drug safety system and to make recommendations to improve risk assessment, surveillance, and the safe use of drugs.254 The IOM report recommended an overhaul of FDA's outdated post-approval adverse event reporting system and an increase in ‘programs that access and study data from large automated healthcare databases’. Noting that preapproval clinical trials ‘do not provide adequate information about the balance of risks and benefits of drugs that are used by many people for many years’, the report recommended making more effective use of ‘increasingly high-quality data and scientific capacity’ of other public and private sector institutions through ‘a public-private partnership with drug sponsors, public and private insurers, for-profit and not-for-profit health care provider organizations, consumer groups, and large pharmaceutical companies to prioritize, plan, and organize funding for confirmatory drug safety and efficacy studies of public health importance’.255
The recommendations in these reports did not speak directly to the role of payers in healthcare innovation and regulation. But by highlighting the value of healthcare records and observational studies in the ongoing process of systematic learning from clinical experience, they set a course that would enlarge the role of payers as the institutions with stewardship of those records.
Congress responded by passing the Food and Drug Administration Amendments Act of 2007 (FDAAA 2007),256 a complex piece of legislation that gave FDA significant new authorities to oversee the safety of drugs after approval.257 This legislation marked a shift in the evidentiary basis for FDA decision making away from sole reliance on data from premarket clinical trials and adverse event reports submitted by drug companies258 toward new sources of data and expertise.259 It directed FDA to collaborate with ‘public, academic, and private entities’ to obtain access to ‘disparate data sources’ and to ‘develop validated methods for the establishment of a postmarket risk identification and analysis system to link and analyze safety data from multiple sources’.260 Once these methods were developed, it directed FDA to ‘establish and maintain procedures for risk identification and analysis based on electronic health data’.261
The electronic health data that these provisions direct FDA to monitor are for the most part in the custody of payers. Although the statute charges FDA with the job of developing and using the system for surveillance, it also contemplates that FDA will work in cooperation with other actors and institutions, and authorizes FDA to enter into contracts with public and private entities to achieve these goals.262 Indeed, it is difficult to imagine how FDA could carry out its new statutory directives unless it works with the payers who have custody of health data.
With these marching orders, FDA has worked with payers and others to establish its Sentinel System for monitoring the safety of drugs. It entered into a 5-year contract with the Harvard Pilgrim health plan to develop a pilot ‘Mini-Sentinel’ system, and recently entered into a new contract with Harvard Pilgrim to lead the Sentinel System in partnership with over 50 healthcare organizations and academic institutions.263
Mini-Sentinel has already facilitated innovative analysis of information using payer data. In 2010, FDA launched a study of the risk of intussusception264 in infants receiving rotavirus vaccines after ambiguous postmarketing studies conducted by the vaccine sponsors.265 FDA used Mini-Sentinel to access payer information from Aetna, Healthcore, and Humana relating to over 1.3 million vaccine administrations.266 Researchers found a small but significant increase in intussusception, enough to require labeling changes for the vaccines.267 Although this is a success story for Mini-Sentinel, it also highlights the challenge of this type of network: someone must know to ask the question, and currently, the only one asking the questions is FDA.
Sentinel is primarily an important public health initiative to develop and utilize new technology and data sources in a distributed network to monitor safety. Although the purpose of the Sentinel System is to monitor drug safety, this unique resource is currently being used for other public health purposes as well.268
More broadly, Sentinel is also a significant research initiative that leverages public resources, public health goals, and legal authorities to support the development of technology that has other uses in biomedical research.269 The statute explicitly calls for the development and validation of new analytical methods.270 More generally, it sets goals that drive the development of new capabilities. Because this initiative looks to establish a network of data sources, it creates new partnerships among institutions, including payers, that might benefit from other collaborations outside the Sentinel System. It identifies obstacles (such as data quality and interoperability and privacy) and challenges participants to develop strategies to overcome them.271 And it engages in this research effort a set of institutions that have a direct stake in the research, but might not otherwise have taken on such a significant role in health R&D.
Government Research Programs
The federal government has also used its role as research sponsor to establish new research programs that organize, subsidize, and direct research using health records. These programs provide subsidies and training and build networks across public, private, and academic institutions, providing a foundation for future research.
Comparative effectiveness research
The Affordable Care Act (2010) authorized the establishment of the Patient Centered Outcomes Research Institute (PCORI) to oversee and set guidelines for comparative effectiveness research. The legislation specifies that PCORI will be a non-profit, non-governmental research institute with an initial appropriation from Congress and subsequent funding from a new fee on health insurers until it sunsets in 2019.272
PCORI is distinctive in its focus on engagement with diverse stakeholders including clinical and patient communities to ensure the relevance and impact of research. It is governed by a 19-member board including patients, caregivers, and representatives of hospitals, insurers and product-developing firms.
The PCORI provisions of the ACA specifically target the same technical difficulties that payers would confront in their own comparative effectiveness research,273 creating communities to address these difficulties with federal funding. PCORI has awarded $100 million to establish a national patient-centered outcomes research network, PCORnet, composed of 11 large healthcare organization networks and 18 patient-group-based networks, which will generate interoperable data sets to support multinetwork studies. It is actively funding studies.
PCORI occupies a politically precarious niche in the biomedical innovation system. There have been repeated proposals to eliminate PCORI. Some critics charge that it is redundant to the ongoing efforts of other agencies. But its focus on engaging clinical caregivers and private payers in the research distinguishes it from other more academically oriented research programs, and perhaps offers the prospect of training and engaging a new set of institutions that will continue their involvement in research in the future.
The political compromises necessary to pass the ACA left PCORI with an ambiguous statutory limit on performing cost-effectiveness research.274 It reflects concerns by some opponents of comparative effectiveness research that the establishment of PCORI would lead to rationing or withholding of care from disabled people based on assessments of government bureaucrats that some lives are worth less than others. Some commentators read this provision broadly to prohibit consideration of cost-effectiveness in PCORI-funded research.275 PCORI apparently agrees.276 Whatever limits the statutory language imposes on PCORI, it does not constrain other institutions outside the government. Thus, private insurers could develop their own cost-effectiveness thresholds and use them to make coverage determinations without violating the law. Indeed, the statute explicitly states that ‘Nothing in this section shall be construed … to permit the Institute to mandate coverage, reimbursement, or other policies for any public or private payer ….’277
But the statute directly prohibits use of dollars-per-quality adjusted life years as thresholds for coverage determinations under Medicare. Since private insurers often replicate Medicare coverage determinations, the constraints on ‘the Secretary’ may effectively determine private sector moves as well. These provisions may undermine the potential of PCORI research to drive cost savings in practice, but should not significantly constrain its research mission.
Another important source of research funding that is likely to accelerate progress in overcoming technical obstacles to payer innovation is the National Institutes of Health (NIH). NIH is, of course, the largest source of funding for biomedical research, and an important driver of precision medicine research. We focus on two NIH initiatives: the Precision Medicine Initiative and the eMERGE network.
President Obama announced the Precision Medicine Initiative in his 2015 State of the Union address. The aim of the initiative is to drive precision medicine forward through public–private partnerships, including work with drug companies on cancer genomics.278 Eventually, the initiative aims to develop a cohort of at least one million Americans with full genomic and health data to be used for research.279 The President called for an initial $215 million in funding to drive this research.280 Crucially, the goals of the program included a significant focus on infrastructure for research, including developing the cohort, creating information management and analysis tools, and helping cement relationships between public and private entities.281
Another NIH-funded initiative in this area is the eMERGE (electronic MEdical Records and GEnomics) Network, a consortium of research institutions organized and funded by the National Institute for Human Genome Research that brings together researchers with wide-ranging expertise in genomics, statistics, ethics, informatics, and clinical medicine.282
The eMERGE Network aims to combine information from EHRs with genotype data from DNA biorepositories to identify relationships between genetic variations and health outcomes and to assess the utility of genotype information for clinical use.283 To facilitate this research, the eMERGE network has had to address a number of issues with legal implications, including developing standardized patient consent language and best practices for sharing patient genetic data, and has formed working groups to address these issues.284 eMERGE's sponsor, the National Human Genome Research Institute, has a long history of sponsoring research on ethical and legal issues associated with human genome research and incorporating best practices into research.285
The tools and databases created by the eMERGE network have also been used by participating institutions in collaborations outside the network, suggesting spillover benefits in promoting further research beyond the immediate scope of sponsored activity. For instance, eMERGE researchers developed the Phenotype KnowledgeBase, or PheKB, a collaborative environment used to collect, validate, and share electronic algorithms for learning about patient phenotype based on health data.286 The eMERGE model consent form can be used by any organization, including integrated systems, collecting genomic data for future analyses. Finally, eMERGE's privacy-protecting data collection framework offers a pathway for future data collection endeavors by payers or data aggregators.287 This last is a particularly relevant example of the way that federal research initiatives can facilitate payer innovation, because it bears on a large non-technological hurdle to that innovation: privacy rules protecting patient data.
The HITECH Act and amendments to the Privacy Rule
Although HIPAA privacy regulations create challenges for payer innovation, recent legislative activity has made some of those challenges slightly easier to overcome. The HITECH Act largely fortified the privacy protections of HIPAA, including applying its provisions to a broader set of entities,288 requiring notification to individuals of breaches,289 and strengthening enforcement provisions.290 It also imposed a new statutory requirement for individual authorization for the sale of protected health information,291 subject to certain exceptions, including an exception for sale for research purposes for a price that ‘reflects the cost of preparation and transmittal of the data’.292
However, in the course of amending the Privacy Rule to comply with the HITECH Act requirements,293 HHS made a number of changes and interpretations to facilitate authorizations for use of health records in research.294 HHS clarified that the receipt of grant funding to perform a research study that involves provision of protected health information is not considered a sale of protected health information.295 In another change not explicitly required by the statute, HHS amended the Privacy Rule to permit covered entities to combine authorizations for use and disclosure of health information with related permission to use biospecimens,296 and modified its interpretation of the Privacy Rule to permit use of a single authorization form for multiple future studies.297 These changes minimize bureaucratic costs by allowing a single authorization to cover multiple studies, and even to include future health information.
Overall, HIPAA still creates substantial legal barriers to innovation by payers.298 In addition to direct legal restrictions, privacy rules may exacerbate technical challenges, as when de-identification makes it harder to integrate information from different sources. Although recent legislation and modifications to the HIPAA Privacy Rule have made research uses easier in some respects, more reform may be necessary to exploit the promise of research using health records.
Yet paradoxically these obstacles may enlarge the role of payers as custodians of health records in research as research consortia use distributed networks of data rather than central repositories to avoid triggering HIPAA violations. These arrangements are an opportunity for payers to expand their involvement in health research. At the same time, increased payer participation in innovation may minimize risks to patient privacy by reducing the need to transfer health records to entities that are not bound by the protective constraints of HIPAA.
CONCLUSION: MOVING FORWARD
Demand-side innovation by healthcare payers has tremendous potential to improve healthcare quality and to lower its costs. Payers have access to health data on millions of patients, giving them the opportunity to develop new information about drug toxicity, comparative effectiveness, precision medicine, and to perform other forms of innovation. Just as important, payers have substantially different incentives than the product-developing innovators that more typically benefit from the standard tools of innovation policy such as intellectual property. As health costs continue to rise, innovation directed at frugality and efficiency becomes ever more crucial. But encouraging innovation on the demand side may require very different policy tools than those used to motivate firms to develop expensive new products.
The barriers facing payer innovators are substantial, including technical hurdles that impede aggregation and analysis of data as well as legal obstacles designed to protect patient privacy. The peculiar economic and legal landscape of the healthcare market may limit the ability of individual firms to capture the benefits of payer innovation, and the standard rewards of intellectual property are unlikely to help. However, a multipronged government approach is helping payer innovation move forward. A combination of funding and mandates for the use of EHRs, engagement of stakeholders in building research networks, and modest changes to privacy rules are helping to make new research initiatives possible.
While we applaud these efforts, there is more to be done to take advantage of the incentives and capabilities of payers as innovators. Continued support of technological enablers for payer innovation is key. HIPAA's privacy protections can also be streamlined to ease the use of data in innovative medical research. Perhaps most important, though, payer innovation needs improved pathways to implementation. Payers can share innovative information with physicians, and can create incentives to try to align the actions of physicians with that information. But payers have limited access to FDA, a central arbiter of what technology gets used and how. FDA determinations go a long way in influencing how physicians and patients use drugs or other treatments, and payer information about comparative effectiveness, toxicity, and precision medicine could aid those determinations. Unfortunately, as described in our opening examples of OTC antihistamines and Vioxx toxicity, FDA has been resistant to using information proffered by payers in the past. Congress in 2007 enacted FDAAA and invited FDA to receive input from non-traditional sources including observational studies. FDA has taken steps in that direction, partnering with payers to develop the Sentinel system—but FDA has made limited use of that system to date. Healthcare payers—both private and public—could work with FDA to use this resource more broadly to use healthcare data to improve regulatory decisions that are inadequately informed by premarket clinical trials. In a larger sense, healthcare payers should have an explicit and ongoing seat at the table when discussing the development and evaluation of health technology, both new and old. Meanwhile, scholars of innovation law and policy may find a fruitful, if largely unnoticed, target of study in demand-side innovation.
For helpful comments and conversations, we thank Kevin Haynes, Lucy Hodder, Andrea Leeb, Oskar Liivak, Michael Madison, Nina Mendelson, Robert Moses, Lisa Larrimore Ouellette, Jeanne Ryer, Jordan VanLare, and participants at the Northwestern University Conference on the Law-STEM Alliance as a Catalyst for Innovation, the Intellectual Property Scholars’ Conference at DePaul University, the University of Michigan Law School Legal Theory Workshop, the University of Michigan Law School Student Scholars Workshop, and the Health Law Professors Conference at Boston University. Cassandra Simmons provided excellent research assistance. All errors are our own.