Benefit Design

Benefit Design through Cost Sharing, Tiers, Networks and Value-Based Insurance Design

Benefit design is the way a health plan turns a list of covered services into prices, places and conditions of access for its members. Two plans covering the same services can give very different protection and incentives, depending on their deductibles, copayments, drug tiers, provider networks and approval rules. Each design choice trades the risk protection that makes health insurance valuable against the extra use insurance encourages, and plans can use design to attract or deter particular members. Value-based insurance design (V-BID) is one approach to these choices.

The levers a plan designer controls

A plan's design combines several rule sets, each acting on a different decision by the member or the clinician. Their effects interact, so no single lever describes a plan.

Actuarial value summarises a whole design as one percentage of expected spending.

The economic problem behind benefit design

Cutler and Zeckhauser, in their survey of the anatomy of health insurance, state the central problem as a trade-off between risk spreading and appropriate incentives. More generous insurance spreads risk more widely but leads people to choose more care, which is moral hazard, and leads providers to supply more. In their analysis optimal coinsurance makes patients pay for care up to the point where the loss from less risk sharing is just offset by the benefit of less wasteful care.

The same analysis supports design that differs across services, because if moral hazard differs by disease, the optimal policy has different cost sharing by disease. Cutler and Zeckhauser also describe cost sharing based on spending as a crude control that places no direct incentive on physicians, which is why plans combine demand-side levers with supply-side measures such as utilisation review, which sits in the plan's access rules, and capitation, part of provider payment design.

Design also determines who enrols. Cutler and Zeckhauser describe adverse selection as the tendency of sick people to choose more generous insurance, and note that it encourages plans to adopt measures that deter the sick from enrolling. In the US Affordable Care Act exchanges, Geruso, Layton and Prinz found that risk adjustment and reinsurance neutralised selection incentives for most consumer types but not all. A minority, identifiable by their demand for particular classes of prescription drugs, were predictably unprofitable, and the contract features relating to those drugs were distorted in a way consistent with screening. Drug tiering can therefore act as a selection device as well as a price signal.

Tiered formularies and tiered networks

Tiering puts several prices inside one plan, charging more for non-preferred medicines in a tiered formulary and more for lower-ranked providers in a tiered network. Both aim to move use towards options the plan judges better value without excluding the alternatives.

Huskamp and colleagues compared two employer plans that moved to three-tier formularies with comparison groups covered by the same insurers. The plan with the larger change, which also raised all copayments, saw a major shift in spending from the plan to members. Among people initially taking tier-3 statins, 49 per cent switched to tier-1 or tier-2 drugs against 17 per cent of comparison-group members, but 21 per cent stopped taking statins against 11 per cent. The plan that changed only tier-3 copayments saw switching without a rise in stopping. Goldman, Joyce and Zheng, reviewing 132 articles, found that each 10 per cent increase in cost sharing was associated with a 2 to 6 per cent fall in drug spending, and that for some chronic conditions, including congestive heart failure and diabetes, higher cost sharing was associated with more use of medical services.

Tiered networks act mainly on new choices. Sinaiko and Rosenthal studied 171,581 non-elderly members of Massachusetts Group Insurance Commission plans with tiered physician networks. Physicians in the bottom tier had a lower share of new patient visits, and patients whose physician was placed in the bottom tier were more likely to switch health plans, but tier ranking had no effect on patients leaving a physician they had seen before. Narrow networks and prior authorisation save plan money by limiting access rather than by price, so part of their cost falls on members as travel time and delay.

Value-based insurance design (V-BID)

Chernew, Rosen and Fendrick argued that when everyone pays the same out-of-pocket amount for services whose benefits depend on patient characteristics, there is potential for both underuse and overuse. Value-based insurance design responds to that heterogeneity, using cost sharing to encourage services whose clinical benefits exceed their cost and to discourage those whose benefits do not justify it. It answers the RAND finding, set out below, that uniform cost sharing cut highly effective and less effective services in roughly equal proportions.

The MI FREEE trial tested the idea at random. Choudhry and colleagues randomly assigned the insurance-plan sponsors of patients discharged after myocardial infarction to full coverage or usual coverage for statins, beta-blockers, ACE inhibitors and angiotensin-receptor blockers. Adherence in the usual-coverage group ranged from 35.9 to 49.0 per cent and was 4 to 6 percentage points higher with full coverage. The primary outcome, first major vascular event or revascularisation, did not differ significantly (17.6 against 18.8 per 100 person-years; hazard ratio 0.93). Total major vascular events or revascularisation (21.5 against 23.3) and first major vascular events (11.0 against 12.8) were significantly lower. Removing copayments did not increase total spending, and patient costs fell.

What the RAND and Oregon experiments show about design

The two large US insurance experiments answer different design questions. The RAND Health Insurance Experiment varied how much members paid, and the Oregon lottery varied whether people had cover at all.

In the RAND experiment, run between 1971 and 1982, families were randomly assigned to plans ranging from free care to 95 per cent coinsurance, with out-of-pocket spending capped at a share of income or a fixed amount, whichever was lower. RAND's summary reports that cost sharing reduced the use of nearly all health services, reduced highly effective and less effective services in roughly equal proportions, and in general had no adverse effect on health, although free care improved hypertension, dental health, vision and selected serious symptoms, mainly among the sickest and poorest participants. Aron-Dine, Einav and Finkelstein note that the results are still widely held to be the gold standard for designing insurance policies.

The Oregon Health Insurance Experiment studied OHP Standard, a Medicaid programme offered by lottery in 2008, which Finkelstein and colleagues describe as relatively comprehensive with no consumer cost sharing. In the first year the selected group had higher use of care, including primary and preventive care and hospital stays, lower out-of-pocket spending and medical debt, and better self-reported physical and mental health. After about two years Baicker and colleagues found no significant effect on hypertension, high cholesterol or average glycated haemoglobin, a 9.15 percentage point fall in positive depression screens, and near elimination of catastrophic out-of-pocket medical spending. For design, Oregon shows the value of cover itself as financial protection, while RAND shows the cost of making members pay at the margin; neither tests targeted pricing, which is the question V-BID trials address.

Worked example of who pays when a copayment is removed

The figures in this example are illustrative, not trial results. A plan covers 1,000 members after a heart attack. Each member either takes a preventive drug all year or does not take it. The drug's allowed cost is USD 50 a month, or USD 600 a year, and the current copayment is USD 30 a month, or USD 360 a year. Before the change 500 members are adherent; after the copayment is removed, 550 are.

1. Spending before. The plan pays the USD 240 left after the copayment for each adherent member, 500 x 240 = USD 120,000, and members pay 500 x 360 = USD 180,000, a total of USD 300,000.
2. Spending after. The plan pays 550 x 600 = USD 330,000 and members pay nothing, a total of USD 330,000.
3. Decomposing the change in plan spending. The plan's drug spending rises by USD 210,000 (330,000 minus 120,000), which splits into two parts as follows.

$$\Delta S = N_0 k + (N_1 - N_0) p = 500 \times 360 + 50 \times 600 = 180{,}000 + 30{,}000 = 210{,}000$$

where $\Delta S$ is the change in annual plan drug spending, $N_0 = 500$ and $N_1 = 550$ are the adherent members before and after the change, $k = 360$ is the annual copayment per adherent member before the change and $p = 600$ is the annual allowed drug cost per adherent member.

1. Break-even. If each avoided cardiovascular event saves USD 25,000 in treatment costs (illustrative), the plan needs 210,000 / 25,000 = 8.4 avoided events a year to cover its extra drug spending. Counting member and plan payments together, only the new use is an extra resource cost, so 30,000 / 25,000 = 1.2 avoided events cover it.

The USD 180,000 is spending shifted from members to the plan, not new resource use. A V-BID change can therefore look expensive to the payer while being close to cost-neutral for health care spending as a whole, and the choice of perspective in a cost-effectiveness analysis of a design change decides which figure counts.

Benefit design in economic evaluation and HTA

Health technology assessment usually asks whether a technology should be covered; benefit design then sets what members pay for it and on what conditions. Those terms change uptake and adherence, so effectiveness in practice depends on the design. A model of a design change needs the price elasticity of demand for the service, any offsetting use of other services, and a clear statement of whose spending counts.

Boundaries with nearby concepts

Benefit Package is the list of services covered; benefit design adds the prices, networks and conditions attached to that list. Cost sharing is one lever within the design, a formulary is its drug component, and actuarial value summarises its generosity for a standard population without showing how costs fall on particular members. Provider payment acts on the supply side and sits outside benefit design, although the two interact.

Common misreadings

Four readings of benefit design fail against the evidence above. Each is given with the reason it fails.

• Uniform cost sharing removes mainly low-value care. The RAND experiment found highly effective and less effective services cut in roughly equal proportions.
• Lower cost sharing for high-value care pays for itself. MI FREEE found no significant change in total spending, with a confidence interval wide enough to include savings or extra cost, and the worked example shows the payer's cost can be large even when the resource cost is small.
• Two plans with the same covered services give the same protection. Tiers, networks and approval rules change both the price paid and the care used, as the formulary studies show.
• Higher claims in a generous plan prove moral hazard. Sicker people may choose generous plans, and risk adjustment does not remove every selection incentive.

Sources

• Aron-Dine A, Einav L, Finkelstein A. The RAND Health Insurance Experiment, three decades later. Journal of Economic Perspectives. 2013;27(1):197-222. doi:10.1257/jep.27.1.197
• Baicker K, Taubman SL, Allen HL, et al. The Oregon experiment: effects of Medicaid on clinical outcomes. New England Journal of Medicine. 2013;368(18):1713-1722. doi:10.1056/NEJMsa1212321
• Chernew ME, Rosen AB, Fendrick AM. Value-based insurance design. Health Affairs. 2007;26(2):w195-w203. doi:10.1377/hlthaff.26.2.w195
• Choudhry NK, Avorn J, Glynn RJ, et al. Full coverage for preventive medications after myocardial infarction. New England Journal of Medicine. 2011;365(22):2088-2097. doi:10.1056/NEJMsa1107913
• Cutler DM, Zeckhauser RJ. The anatomy of health insurance. In: Culyer AJ, Newhouse JP, eds. Handbook of Health Economics. Vol 1A. Amsterdam: Elsevier; 2000:563-643. Working paper version: NBER Working Paper 7176; 1999.
• Finkelstein A, Taubman S, Wright B, et al. The Oregon Health Insurance Experiment: evidence from the first year. Quarterly Journal of Economics. 2012;127(3):1057-1106. doi:10.1093/qje/qjs020
• Geruso M, Layton T, Prinz D. Screening in contract design: evidence from the ACA health insurance exchanges. American Economic Journal: Economic Policy. 2019;11(2):64-107. doi:10.1257/pol.20170014
• Goldman DP, Joyce GF, Zheng Y. Prescription drug cost sharing: associations with medication and medical utilization and spending and health. JAMA. 2007;298(1):61-69. doi:10.1001/jama.298.1.61
• Huskamp HA, Deverka PA, Epstein AM, et al. The effect of incentive-based formularies on prescription-drug utilization and spending. New England Journal of Medicine. 2003;349(23):2224-2232. doi:10.1056/NEJMsa030954
• RAND Corporation. The Health Insurance Experiment: a classic RAND study speaks to the current health care reform debate. Research Brief RB-9174-HHS. Santa Monica, CA: RAND; 2006.
• Sinaiko AD, Rosenthal MB. The impact of tiered physician networks on patient choices. Health Services Research. 2014;49(4):1348-1363. doi:10.1111/1475-6773.12165