The following is a guest post by Michael Chernew and Aaron Schwartz. Michael is a health economist and Professor of Health Care Policy at Harvard Medical School. He has written extensively on issues of benefit design, particularly Value Based Insurance Design. Aaron is a MD/ PhD candidate at Harvard University. His research focuses on quantifying waste in the health care system and evaluating strategies to eliminate it.
Recently, there has been much discussion of innovations in benefit design, including on this blog, where there was a recent post about a split benefit design. Given the range of proposed options it is useful to revisit the connection between benefit design and theory.
The goal of optimal insurance design is to maximize societal welfare, which consists of two elements. First, an optimal plan steers beneficiaries toward high value services, minimizing moral hazard. Second, an optimal plan provides protection against risk, ensuring that beneficiaries can expect to experience relatively similar welfare across a range of possible life outcomes (i.e. in sickness and in health).
The motivation for cost sharing in standard economic models is to balance these sometimes competing objectives. Early models of optimal coinsurance were based on a single coinsurance rate. More recent innovations have more nuance. The unifying theme is that optimal cost-sharing should be targeted to situations where patients can respond by making different health care choices. For instance, a patient suffering a heart attack will almost surely exceed most deductibles. So, the cost sharing associated with a high deductible plan will have very little impact; there is no incentive for the patient to follow a more fiscally conservative treatment path or choose a less expensive facility.
One strand of new designs (e.g., reference pricing and tiered networks) focuses on choice of provider. These designs recognize the widespread variation in prices. They allow beneficiaries that seek care from low cost providers to share the savings. Reference pricing focuses on specific services. Typically a fixed price is paid by the insurer and the beneficiary must pay the difference if they get care from a higher priced provider. Tiered network plans typically identify preferred providers (physician and hospitals) based on cost, and sometimes quality and place them in a preferred “tier”.
Both reference pricing and tiered network designs will be more effective with better search tools, but they still must contend with complexities of the delivery system. For example, tiered network products sometimes place hospitals and the physicians with admitting privileges at that hospital in different tiers. Reference pricing, which is more targeted than tiered networks, may be practical for only a relatively small share of spending. Tiered networks may affect more spending, but may disadvantage patients that have conditions best treated at the high tier facilities. In both cases the effectiveness of these products depends on the variation in provider process and existence of sufficient choice. If there is only one provider these benefit designs will be ineffective.
Another strand of design, value based insurance designs (VBID), focuses on which services are used. The idea is that cost sharing should be low for high value services and higher for low value services. These designs recognize the underutilization of high value services (which may be exacerbated by across the board coinsurance increases) and the overuse of low value services (which have received increasing attention through campaigns such as Choosing Wisely and evidence on widespread geographic variation in use). Commonly, VBID designs are applied to low unit cost preventive services, but the theory is much broader. In these cases, traditional cost sharing acts like a tax, with few beneficial incentive effects. VBID allows patients who choose low cost treatment options to share in the savings.
Split benefit design applies similar principles to patients with high cost illnesses. These patients often face little cost-sharing because they have exceeded their annual out-of-pocket limits. Unlike the previous examples, split benefit design involves a cash rebate to patients who choose less-expensive treatment options. This rebate is forfeited if the patient instead chooses the more expensive treatment option.
An intriguing aspect of split benefit design is that, relative to fully covering expensive treatments, this design does not increase the financial burden of sick patients receiving expensive care, and yet it still encourages the choice of less expensive treatment alternatives. However, this feature comes at a cost of reduced income smoothing (risk protection); indeed, premiums could increase substantially under certain circumstances. Consider the extreme case in which the rebate equals the price difference between the low-price and high-price care options. This split benefit design would ensure that the low-price option effectively costs the insurance company the same amount as the high price option, and premiums would be as high as if all patients chose a fully-covered high-price option.
Chernew, Encinosa and Hirth (CEH) worked out the math in a related scenario. The insights from the CEH model is that the optimal benefit design charges patients who choose the high cost treatment a fee and pays the patients who choose the low cost option a rebate. The sum of the fee and the rebate is less than the full incremental cost (which dilutes incentives to choose the low-cost option but helps insure against the “risk” that a patient prefers the high cost treatment). This model, described in detail in the paper, does a better job at smoothing utility across different states of the world than reference pricing (in which beneficiaries may pay the incremental fee for high-cost care) or split benefit design (in which beneficiaries are paid a rebate if they choose the low cost option). Specifically, CEH is based on a utility maximizing model that recognizes the need to transfer income from the healthy to sick state of the world, and in the context of that model derives the optimal way to do that.
A key distinction among split benefit, reference pricing, and CEH is distributional. Reference pricing has low premiums and charges people who fall ill and opt for the high cost option more. Split benefit has high premiums and refunds a portion to those who get sick and choose the low cost option. The CEH model falls in between. If all beneficiaries can equally expect to become sick, then CEH maximizes patient welfare. But if risk is heterogeneous, distributional issues become important. For example, reference pricing favors relatively healthy people because premiums are low and out of pocket costs if one becomes ill could be high. Split benefit favors less heathy people because premiums are high and individuals who use care may receive a rebate. Economic efficiency criteria say nothing about these distributional issues, which depend on concepts of fairness.
As benefit designs evolve, these and other innovations are likely to get more attention. Implementation issues will be important. Currently, many insurers do not offer these more innovative designs. Over time, if insurers, employers, benefit consultants and most importantly patients, become more comfortable with these designs they will become much more common and offer mechanisms to improve the efficiency of the health care system.