Insurance Switching and Mismatch Between the Costs and Benefits of New Technologies
David Cutler, PhD; Michael Ciarametaro, MBA; Genia Long, MPP; Noam Kirson, PhD; and Robert Dubois, MD, PhD
Scientific progress has transformed patient outcomes in many disease areas, leading to economic gains.1 However, such therapies can challenge short-term payer budgets if benefits are not coincident with costs. Although this phenomenon is not new, recent and anticipated therapies may exacerbate these challenges. For example, sofosbuvir (Sovaldi), lauded as a breakthrough hepatitis C virus (HCV) treatment, has been restricted by some insurers concerned about short-term budget impacts, including by delaying access for patients with asymptomatic or milder disease whose costs would be paid later by Medicare.2
Disconnects between the impacts on different payers can be large in the United States, where commercial insurers, state Medicaid programs, and the federal Medicare program pay most costs. Given insurance switching by patients over time, payers covering initial costs may not benefit from all, or any, downstream cost offsets. Moreover, patients and families may highly value better health and quality of life, improved functional status and productivity, and longer life, whereas insurers may value lower costs most.
In 5 hypothetical examples, we modeled the mismatch between who pays for and who benefits from innovative therapies. Like others, we focused on patient movement over time across Medicaid, commercial insurance, and Medicare rather than contemporaneous switching among private insurers.3 Our aim was to explore a widely acknowledged feature of US healthcare, namely that the fragmented insurance system creates potential disincentives for coverage of therapies with up-front costs and long-lived or delayed benefits, and whether the situation may be exacerbated for new clinically effective therapies that are also high-priced relative to the current standard of care (SOC). Rather than precise numerical estimates for specific diseases, we explored whether substantial disconnects may be expected under credible, but by no means the only possible, assumptions (and therefore the extent to which some cost-effective therapies with potential to improve length and quality of life may face heightened coverage disincentives); whether they vary across examples; and the implications for policy. Some prior studies' results have illustrated payer disconnects for specific diseases; others have advocated specific policy responses. This study extends the literature by comparing disconnects across diseases subject to new transformative therapies and by exploring the implications for effectiveness of potential policy responses.
We examined 5 disease states for which transformative therapies have been discussed or recently launched: highly effective HCV therapy; curative gene therapy for beta-thalassemia (BT), a rare childhood genetic disorder; disease-modifying therapy for patients with mild Alzheimer disease (AD); and cardiovascular disease (CVD) therapy for patients with the rare genetic disorder familial hypercholesterolemia (FH) and those with prior CVD. Several resemble, but do not purport to be identical with, recently introduced on-market therapies (HCV and cardiovascular therapies); others reflect areas that may produce breakthrough therapies (disease-modifying therapy for AD and gene therapy). These examples, although not exhaustive, were selected to represent diverse patient types (ie, pediatric, adult, and senior populations), clinical intervention models (ie, 1-time curative and ongoing disease-modifying therapies), and disease burdens (ie, highly certain ongoing chronic health management costs, probabilistic catastrophic hospitalization costs, and custodial and other costs from function deterioration). Table 1 summarizes key characteristics across the examples.
To assess the net effect by payer type of each therapy, we constructed disease-specific analytic models and compared the present discounted value (PDV) of an individual’s expected lifetime healthcare costs under the current SOC and the hypothetical new therapy, from the age an average patient initiates the latter (eg, gene therapy at age 2 years). We adopted this analytic frame to model payers’ budget impact considerations associated with covering the new therapy for a patient of expected age. We also calculated improvements in quality-adjusted length of life associated with the new therapy and the incremental cost per quality-adjusted life year (QALY). This cost-effectiveness metric is provided as an indicator of social desirability. For US commercial payers, cost-effectiveness analysis is typically not an established coverage determination constraint, but budget impact analyses are important considerations. Therefore, we focused on budget impact in our analysis.
We first calculated the aggregate budgetary impact on all payers of the new therapy, including healthcare offsets due to morbidity improvements and additional healthcare costs due to extended life. We also incorporated the impact on elder care in the case of disease-modifying therapy for AD, including the net impact on both nursing home care and family caregiving. Second, we disaggregated these effects into those on a representative initial payer in the 3 main payer types and those on a representative downstream payer, restricting analysis to the most relevant payers (eg, 2-year-olds are generally covered by commercial insurance or Medicaid, not Medicare).
We modeled the impact of patients switching payer types as they age, rather than switching commercial insurance plans contemporaneously. Whereas approximately 1 in 8 nonelderly Americans with employer coverage switched health plans in 2010 (approximately 1 in 13 due to reasons other than job change), nearly all will transition to Medicare at age 65 years.4 If recent and expected therapy breakthroughs suggest a continuing shift toward front-loaded costs and back-loaded benefits, the implications for both commercial insurance and Medicare may be far reaching. We explored the reasonableness of this modeling choice via several anonymized interviews with medical directors at large commercial payers who confirmed that, given prohibitions on pre-existing condition exclusions and the nature of geographic competition where leading plans may tend toward similar coverage, they generally expect that short-run losses from a therapy for patients who “switch out” roughly offset gains from those patients who have “switched in” and whose therapy costs were covered by other commercial payers. However, for the effects of switching over time as patients age, similar assumptions do not apply. Respondents were not interviewed about the effects of potential Affordable Care Act repeal or about actions that could affect the balance between commercial coverage and state exchanges.
Our analyses rely on best-available Markov-type models published by others, incorporating rates of patient transition from one health state to another and healthcare costs and patient utilities for each state. In order to compare the hypothetical new treatment with the current SOC, we adapted these models by varying parameters related to efficacy, cost, and age at therapy initiation, specific to the hypothesized intervention. We applied shared assumptions across the models for the percentage distribution of insurance type by age and sex from the literature. In calculating the total impact of the new therapy, we valued a QALY at $100,000; the impact on payers excludes this value, as there is no market to monetize the value of additional QALYs.5 (Cost-effectiveness calculations exclude the value of additional QALYs, by definition.) Throughout, the value of all costs and savings was discounted at a 3% annual rate. For additional relevant disease-specific and shared assumptions, see the eAppendix (eAppendices available at ajmc.com).
Table 2 summarizes the aggregate impact of the different therapies. For the 2 CVD examples, the model relied on recently released calculations for patients aged 35 to 74 years (rather than a single age) with FH and a history of CVD.6 Without direct access to the authors’ health state-specific model, we calculated incremental cost per QALY from these figures (after adjusting for a modeled average 20% net price discount). Figures reported for the 2 CVD therapies in Tables 2 and 37and the Figure reflect these cost-effectiveness figures (rather than higher figures reflected in a PCSK9 inhibitor manufacturer’s technology appraisal submission to the United Kingdom’s National Institute for Health and Care Excellence).8 Regardless, we focused on the difference between the impacts on the initial and downstream payers rather than their absolute levels.
Under our assumptions, all 5 therapies would increase discounted net healthcare costs. The magnitude of additional QALYs and the healthcare costs in additional years of life would vary, depending on patient and disease dynamics. Under the assumptions used, 3 therapies were highly cost-effective, with an incremental cost per QALY of $55,000 or less; the 2 CVD therapies were cost-effective at a value of about $250,000 per QALY (less, under manufacturers’ estimates; translated from pounds to dollars without any other adjustment for differences in utilization or unit prices, the corresponding figures would be incremental costs per QALY of $33,703 for FH and $67,701 for prior CVD). Our focus, however, was on disconnects across payers, and the Figuredisaggregates the overall payer impact into impacts on initial and downstream payers. For HCV, BT, and AD, the financial impact on the initial payer was negative and the impact on at least 1 downstream payer type was positive. Table 3 reports these figures in dollar terms and, to allow for direct comparison, per dollar of aggregate payer impact.
Under our assumptions, treating BT costs the healthcare system nearly $180,000. The impact by payer varies depending on the initial insurer. When commercial payers are the initial insurers, they face slightly lower financial impacts relative to aggregate healthcare costs, the difference being additional downstream Medicare costs from children now surviving to age 65. That said, most costs are paid by commercial payers. Children initially covered by Medicaid, however, are covered by private insurance when older (modeled at age 21). Thus, Medicaid pays all treatment costs and commercial insurers realize a gain, the net effect of more likely survival and lower per-patient costs. For every patient whose treatment at age 2 is paid by Medicaid, commercial insurers benefit by a cumulative PDV of $120,661.
Others also have analyzed the tension between long-term cost-effectiveness and the immediate budget impact of highly effective therapies for HCV and similarly find a disincentive for commercial insurer coverage, with results borne by Medicare and other downstream payers7,9,10; 1 study estimated roughly a 15-year payback period for private payer coverage.3 Under our assumptions, initial commercial payers experience a cumulative PDV net cost of about $15,000 per patient. Medicare benefits from commercial payer coverage because patients avoid later expensive catastrophic events, such as liver cancer and transplants. These savings are greater than the additional costs incurred from patients living longer, for a gain of nearly $3000 per patient.
Slowing the progression of AD costs Medicare, both because it pays therapy costs and because patients live longer. However, it also reduces the need for nursing home care, thus saving Medicaid approximately $30,000 per patient. Families benefit from reduced needs for nursing home care, but experience additional caregiving burden during longer disease progression at home, for estimated increased net costs.
For patients initiating CVD therapy before age 65, commercial insurer and Medicaid costs are lower than the aggregate impact because many therapy costs occur after age 65. Although commercial insurers and Medicaid experience lower savings from avoided cardiovascular events, they also experience lower additional healthcare costs from longer life. For patients initiating therapy after age 65, Medicare experiences all therapy costs, healthcare cost offsets, and extra healthcare costs associated with extended life. For both populations, the net effect is negative, moreso for patients with prior CVD than those with FH. Although the magnitude reflects the assumptions used by others (which have been critiqued11), the general pattern remains under other cost and disease transition assumptions.
Our research confirms that switching between payer types over time results in financial disconnects between initial and downstream payers across multiple hypothetical examples of highly effective new therapies with front-loaded costs and back-loaded benefits. Without mechanisms to monetize the downstream benefits of health improvements to others, returns from initial payers’ investments are understated. In particular, switching from commercial to Medicare coverage at age 65 may result in systematic disincentives for some new therapies by commercial payers, depending on specifics relating to age at initial treatment, up-front therapy cost, and morbidity and mortality impacts.
Medicare may be a financial “winner” or “loser,” depending on the balance between additional therapy cost, morbidity improvement savings, and extra healthcare costs from mortality gains. For HCV, we found (as have others) that Medicare benefits from initial payer coverage.3 For BT, Medicare impacts are far in the future and somewhat negative. For disease-modifying AD therapy and the CVD therapies as modeled, Medicare would pay more. Depending on the magnitude of the effects and the numbers of patients treated, downstream Medicare impacts of commercial insurer decisions could be an important additional form of “spillover,” documented in other contexts.12 Commercial insurers face negative financial impacts across the examples when they are the initial payers, suggesting all therapies could face coverage disincentives, overlooking downstream cost offsets. Yet, from an aggregate healthcare cost point of view, the incremental cost per QALY as modeled is within standard acceptable ranges and well below for some, and investment would be socially desirable.
Absent direct social investment or subsidies, other approaches may address disconnects between privately incented and socially desirable outcomes. Two types of approaches, or a combination, may be relevant, depending on circumstances: mechanisms to align costs and benefits over time (for the same payer) and to help finance up-front therapy costs and mechanisms to share and align therapy costs and benefits across payers (eg, transfers between winners and losers). Several alternative financing proposals of the first type have been proposed, incorporating some form of cost amortizing to address challenges of high up-front costs. These include manufacturer–payer financing mechanisms (eg, manufacturer-issued debt secured by dedicated streams of contractual payments from commercial payers)13,14; changes in accounting rules and/or insurance regulations to allow payers to amortize some costs over longer time periods14; monthly annuity payments or manufacturer service fees linked to clinical milestones and/or continued efficacy, rather than single up-front or per-dosage charges15; or consumer credit or debt programs.16,17 Such arrangements would be novel in biopharmaceutical reimbursement, but they are similar in some respects to financing expensive long-lived consumer medical devices, such as insulin pumps, that are used for chronic disease. Our results suggest that alternative financing mechanisms smoothing front-loaded costs over time could be relevant for 1-time curative gene therapy and highly effective HCV therapy, but they may be only partial solutions, as benefits and costs may still accrue to different payers. Disincentives for HCV therapies occur not only because initial costs for cure are high, creating short-term budget stress, but also because substantial downstream benefits accrue to others. Such mechanisms are likely less relevant for ongoing therapies, such as disease-modifying AD therapies or cardiovascular therapies, where costs are already spread over time.
The second types, cross-payer financial transfers and burden-sharing mechanisms between winners and losers, specifically address gaps between who pays and who benefits (rather than gaps in time between costs and benefits for the same payer). Transfers can address when costs to one payer type are offset by savings to another. For instance, up-front Medicaid cost burdens and future benefits to Medicare could be recognized by enhanced state Medicaid reimbursement rates or direct federal transfers. So-called burden-sharing proposals address when therapies are cost-effective but also cost-increasing, and a gap remains after transfers.
Transfers from one payer type to another theoretically could be appropriate for therapies such as those for BT, where Medicaid bears large up-front costs and commercial insurers experience substantial downstream benefits. However, proposals to smooth out therapy costs over time for the same payer will be easier to implement than proposals to transfer costs and benefits across payer types.18 More generally, future innovative therapies may benefit from proposals tailored to their specific circumstances, including mechanisms to amortize costs over time or to transfer value from downstream winners to initial losers, or a combination (see Table 4). Although we find disconnects between initial and downstream payers in all examples, some substantial, the magnitude and reasons vary.
As with all modeling studies, different price, timing, and effectiveness assumptions yield different results for cumulative payer PDVs (see eAppendix sensitivity analyses). Moreover, not all relevant societal benefits have been included in the models relied upon for cost-effectiveness inputs. For instance, educational attainment and lifetime productivity impacts, important benefits of curing childhood genetic diseases, are not included for BT. Similarly, the value of reducing future transmission to others is not included for HCV and the benefits from sustained functioning and independence for patients and their families due to disease-modifying therapy are not included for AD. These benefits are not monetized but are real nonetheless, and including them could increase gaps between front-loaded costs and back-loaded benefits and/or improve cost-effectiveness. Second, for chronic therapies, we did not include changes in the new therapy’s net price over time. Third, given pre-existing health condition coverage exclusion prohibitions, we applied average population-level insurance coverage statistics and did not incorporate disease-specific insurance coverage or switching rates. We modeled at the aggregate payer type level, and this assumption may not hold true for all plans (eg, smaller payers may face greater temptations to free-ride on others’ prior coverage decisions) and patients (who may experience different switching rates post treatment). For simplicity, we assumed uniform therapy and medical costs across payer types; lower Medicaid prices could reduce Medicaid net PDVs relative to other payers. Fourth, our analyses reflect the limitations of the underlying Markov-type models (eg, constant age-specific mortality and transition rates over time), which may yield underestimated mortality benefits when extended over many years. To the degree that not all healthcare cost offsets from the new therapy are reflected in these underlying models, our calculations overstate net costs. For instance, cost offsets in heart failure and unstable angina are not included in the cardiovascular models and improvements in heart attack and stroke incidence may be understated, as they may reflect an assumed lower-risk treatment population than targeted.11 Because our focus is on general dynamics under plausible (but by no means the only possible) assumptions, our findings are representative rather than precise conclusions about specific disease–therapy combinations or forecasts of the impacts of specific therapies. Finally, our calculations reflect the assumption that, for chronic conditions, downstream payers also cover the therapy (this limitation is not relevant to 1-time therapies, such as gene therapy).
As scientific advances generate breakthrough therapies with varying profiles, creative thinking and flexible solutions by manufacturers, payers, and others will be needed to address barriers to realizing their benefits. Well-designed alternative financing or other mechanisms could help ensure economic incentives for future development, appropriate patient access, and sustainable payer economics for expensive but cost-effective transformative therapies. Although proposals have been suggested to address up-front cost barriers, proposed transfers from downstream winners to initial losers and burden-sharing mechanisms have received less attention. In some cases, both could be helpful, with the balance reflecting disease-specific circumstances. However, further research is needed to address practical and theoretical challenges, including defining why and under what circumstances Medicare would or would not incent private payer coverage, how to maintain incentives for private-sector coverage, and how and when to implement acceptable cross-payer transfers. The framework we used to disaggregate potential impacts on initial and downstream payers of new therapies, and to identify potential gaps between who pays and who benefits, may be a useful tool for manufacturers and others to map sources of potential coverage disincentives and develop and fine-tune such proposals before presenting them to payers. Failing to consider relevant disease-specific dynamics may mean the promise of new transformative therapies is not fully realized.
The authors acknowledge and thank C.J. Enloe for data analysis and Kimberly Westrich and Howard Birnbaum for helpful suggestions.