Proposals to reform the health insurance side of our health care system usually involve a shift of the risk from insurers (public or private) to providers and individuals. The former may be under the guise of bundled payments or capitation, the latter under consumer-directed health plans (CDHPs), perhaps with value-based insurance design (VBID) elements or, more broadly and simply, cost sharing.
People will make all manner of claims about cost sharing, what it will do for spending, how it will or won’t affect health, the ways in which it will revolutionize the market. The research literature tells us a thing or two about what we should really expect from cost sharing. Where evidence is limited or nonexistent, we can’t really say what will happen with much certainty.
Here are a few things we know about high-deductible, consumer-directed health plans (CDHPs) and value-based insurance design (VBID), as conveyed in a Robert Wood Johnson Foundation Synthesis Project review paper by Katherine Swartz.
Feldman et al. (28) found that the CDHP was not able to controlmedical expenditures over time and it appears that the enrollees in the CDHP spent more on hospital care than enrollees in the traditional plans. Greene et al. (50) found that CDHP enrollees were more likely to discontinue some prescription drugs. The findings of Lo Sasso et al. (77) also suggest that CDHPs had larger relative effects on spending for prescription drugs than for outpatient or hospital spending. The findings from these three studies are consistent with expectations about deductibles — once the deductible has been met, there are no longer strong incentives for an enrollee to be concerned about further health care expenditures. […]
Health plans with high deductibles and uniformly applied co-payments or coinsurance rates are oftenreferred to as “blunt instruments” for reducing unnecessary health care expenditures because evidenceis mounting that people reduce both essential and nonessential care (100, 97). As discussed below, uniformly applied cost-sharing particularly causes people to reduce their use of prescription drugs, which in turn seems to lead to use of more expensive types of care that are indicative of adverse events andpoor health outcomes. As a result, a variation on CDHPs known as Value-Based Insurance Design (VBID) has gained traction in the last decade (29, 17, 31, 32, 20). The concept behind VBID is that CDHPs (and health plans in general) would be more effective in reducing use of care that is only marginally beneficial if cost-sharing varied according to the relative value of a service for the individual. […] To date, the handful of studies on the effects of VBID have been conducted by advocates of VBID (32, 20). The studies suggest that spending on health care could be more efficient with nonuniformly applied cost-sharing, but more studies of natural experiments along with greater variation in the cost-sharing incorporated into VBID need to be conducted.
Based on the literature, Swartz concludes the following about cost sharing,
- Cost sharing is likely to reduce patient-initiated care among the healthy, i.e., those with relatively low medical expenses to begin with.
- Recent studies tell us very little about how increased cost-sharing affects total spending.
- There has not been a study of the effects of cost-sharing on health in a general population in thirty years.
- Increased cost-sharing for prescription drugs leads to increased emergency department and inpatient hospitalization spending by the elderly and the chronically ill.
- Responses to cost-sharing differs by income. Low-income individuals are more likely to shift types of service use, not reduce overall use.
- Cost sharing can reduce use and/or expenditures for preventative services, prescription drugs, emergency department utilization, mental health care, and substance use treatment.
- Most people do not distinguish between essential and non-essential care.
- We don’t know the long-term effects of increased cost sharing on health.
I’ve left a lot out. If you want the rest, including references numbered above, see Swartz’s paper. It is ungated and not hard to read.