We dig deeper into the pros and cons of setting the value of a cost-effectiveness threshold according to the marginal productivity of the healthcare service and the importance of deliberative decision-making. This is done in a non-technical way for those who do not have a background in health economics. Full disclosure: we are strong proponents of aiming to use the marginal productivity of the healthcare service as a threshold whenever possible, but we aim to represent the opposing arguments fairly.
Why is the cost-effectiveness threshold important?
All healthcare services require decisions about which medicines and services to fund. Some, such as those in the UK, use a cost-effectiveness threshold or threshold range to inform these decisions. In this context, the cost-effectiveness threshold can be interpreted as the maximum additional cost per unit of benefit that a healthcare service will fund for an intervention (i.e., its “willingness to pay”).
The value of the cost-effectiveness threshold heavily influences the maximum price that will be accepted for drugs and other healthcare technologies. But it is not a quantity that we can easily observe and measure. It is therefore hardly surprising that this value is contentious.
What is the ‘marginal productivity’ of the healthcare service?
To make it easier to understand the concept of productivity in the healthcare service, let’s use the metaphor of the healthcare service’s budget as a bookshelf. For a worked example, see this paper and the accompanying Excel file. Another paper extends the metaphor to the context of when there are several interventions for the same health condition.
Look at the first figure.
Each book (or bar, in light red) is a healthcare technology. The book’s height shows how productive the technology is in terms of health per monetary unit (e.g., health benefit per dollar, pound, etc.) – this is the reciprocal of the incremental cost-effectiveness ratio. The book’s width represents the cost of providing the technology to the population. The books are ranked from left to right according to height. At the far-left end of the shelf, expenditure is zero; it increases as we move along the shelf to the right.
In an ideal world, but with a limited budget for healthcare, we would choose which technologies to fund in descending order of health benefit per dollar until the budget runs out. This is shown below, where the health benefit of the least productive technology offered to the public is labelled ‘cost-effectiveness threshold’.
The health benefit per dollar of the least productive technology that was funded is the marginal productivity of the healthcare service. We call it marginal productivity because it refers to a small increase in funds relative to the overall budget.
This means that the healthcare service’s budget, together with the productivity of all the technologies funded, determines the marginal productivity of the healthcare service.
Why should we use the ‘marginal productivity’ to inform the cost-effectiveness threshold?
Let’s assume our aim is to improve population health. We should not fund technologies that are less productive than any already being funded, even if they improve health. This is because the new less productive technology will take up the budget that was being used for more productive technology. Therefore, more health is lost than gained.
If a new technology comes along that is more productive than what is already funded, it should replace less productive ones. And if we wish the healthcare service to be as efficient as it can be, we should fund all technologies whose productivity is better than or equal to the marginal productivity.
If we use the inverse of marginal productivity of the healthcare service as the cost-effectiveness threshold, we ensure that we can always identify the technologies that improve the population’s health.
What if the healthcare service does not fund the most productive interventions?
The healthcare service funds a mix of technologies. Chances are that this mix includes technologies that have a low health benefit per dollar – this is represented in the figure below.
As in our previous example, the productivity of the least productive intervention that is funded represents the marginal productivity of the healthcare service.
Here we see that one technology has lower productivity than several technologies that are currently not funded. This causes a loss of health benefit indicated by the red quadrangle. This lost health is the opportunity cost of not funding a better intervention with our existing funds.
If we could identify a non-funded technology with higher productivity, population health would rise. And if we could fund the most productive of the non-funded interventions, the red area of loss could be fully captured. By selecting technologies that have greater productivity than those currently funded, we gain more health than what is lost from those that are no longer funded.
To be efficient, therefore, any healthcare system needs reasonably believable information on the cost-effectiveness of both the interventions it funds and the most likely candidates that it doesn’t. For example, NICE has processes for stakeholders to propose possible technologies for cost-effectiveness assessments.
What are the downsides from using the marginal productivity of the healthcare service?
It is difficult to find out what the marginal productivity of the healthcare service is. Let’s suppose that the objective of the healthcare service is to improve health and we think that we have a reasonable way to measure health, say quality-adjusted life expectancy. To apply this ‘bookshelf approach’ (also known as a league table approach), we need credible quality information on the costs and health benefits of all healthcare services, which is unfeasible for many countries.
An approach increasingly used is to take advantage of variations in the healthcare service budget and in the mortality risk between regions and over time. Under some assumptions, researchers can use these variations to find out how small changes (i.e., marginal changes) in the healthcare budget affect quality-adjusted life expectancy (this has been tried in, e.g., England, Australia, Spain, South Africa). However, the reliability of this approach has been contested by other researchers (e.g., here).
Our measures of health are not, however, perfect and they certainly contain value judgements that may not be universally accepted. In cost-effectiveness analysis, we usually measure health in terms of quality-adjusted life expectancy (e.g., the QALY). But the quality weights may not capture all aspects of health or be equally sensitive to all changes. This problem also applies to approaches that seek to measure the relative health impact of healthcare technologies. One solution lies in having a deliberative decision-making procedure that allows for these aspects to be appraised on a case-by-case basis by accountable people.
In practice, the available evidence is incomplete or controversial (or may be absent). This, again, is true for all systematic ways of making decisions about the services to provide. It means that the process is necessarily fraught with uncertainties, some of which may be big ones. So who decides whether the evidence is good enough for the purpose at hand and who decides what risks are acceptable? Once again, one solution lies in a deliberative procedure that allows for these issues to be discussed and appraised on a case-by-case basis by accountable participants.
Another issue is that we care about more things than just health. For example, we care about wellbeing and fairness. We may wish to incentivize and reward the pharmaceutical industry to continue to conduct research and develop new treatments. There may be healthcare interventions that increase the costs for the healthcare service but are much more convenient, cost-saving, or time-saving for patients and their caregivers. If we use the marginal productivity of the healthcare service in terms of health as the sole criterion for selection, these factors are neglected. Most pointedly, however, to incorporate them necessarily entails some avoidable health loss (earlier deaths, reduced quality of life, etc.). This loss might be acceptable if the expected gains from greater fairness, etc., were judged to be sufficient. Hence the importance of having a procedure for deliberative decision-making where these and other factors are considered by accountable discussants.