Health Insurance

Why don’t we rely exclusively on CEA estimates to make drug coverage determinations?

Cost effectiveness analyses (CEA) typically analyses the value of a new treatment relative to its cost.  If the value of the additional health benefits outweigh the additional cost, then they should be covered; if the additional cost outweighs the additional benefit, then it should not be covered, right?

There are a few issues with this argument.  First, typical analyses measure cost effectiveness for the average patient.  In many cases, a treatment that is not cost-effective can be cost effective for some patient subgroups but not others.  In theory, one could simply apply the same CEA principles to subgroups, but there are a number of problems with that approach.  First, oftentimes, not all the information about the patient’s likely treatment response–information likey available to the patient and provider–will be available to they payer making coverage decisions.  Secondly, making coverage restrictions (e.g., step therapy, prior authorization) is administratively expensive.

Another factor is that the value of different outcomes may vary by patient.  For instance, working age adults may place a high value on low toxicity and convenience so they can continue working; elderly individuals may focus less on convenience or mode of administration and more on efficacy. These are stereotypes, but they illustrate the point that patients will have different preferences over outcomes whereas QALYs assume that all patients value outcomes the same.

An interesting paper by Espinoza et al. (2017) examines the benefits and costs of centralized compared to individual choice decision-making in a collectively funded system.  Unsurprisingly, they find that:

the value a health care system will have to associate with offering choice to patients for its own sake (independent of its impact on population health) will have to be greater when it is compared with centralized decision‐making based on subgroup recommendations. This result is to be expected because when different decisions for different subgroups have been made, the expected health loss due to lack of understanding of the heterogeneity is lower and the expected population health gains are greater than the average case.

In short, to the extend that central planners can create precise subgroups that reflect patient clinical circumstances and preferences, then choice-based systems are less attractive.  In cases where there is signficiant unobserved clinical or preference-based heterogeneity, however, or in cases where it is expensive to administer treatment choice restrictions, then choice-based systems will be relatively more attractive.

The issue of allowing the maximum amount of patient choice while also managing finite resources is a key question that will require many years of additional research.


Leave a Reply

Your email address will not be published. Required fields are marked *