I have previously written about the need for distributed cost-benefit analysis (DCEA), which will increase the value of treatments that can improve the health of individuals with the worst quality-adjusted life expectancy (QALE).I have written this AJMC with Health affairs In other places. A key question that many people may encounter is how to actually implement DCEA.
One way is to follow Love-Koh et al. (2019) method. There are six key steps in this method.
- Define relevant interest groups
- Measuring disease prevalence by group
- Measuring the impact of technology
- Determine preferences for fairness and efficiency
- Calculate the equivalent QALY gain of the equal distribution
- Use EDE QALY for standard CEA modeling.
The first step is to define your group. This may depend on geography, age, socioeconomic status, etc. Defining groups is analytically simple, but it can be a challenge because there is no “correct” group breakdown. In addition, the packet must have available data.
The second step is to measure the prevalence (or the incidence of disease), and the third step is to measure the effect of treatment. In the example below, treatment A increases the QALE of the relatively disadvantaged groups 1, 2, and 3; treatment B increases the QALE of the relatively advantageous groups 3, 4, and 5 (from the perspective of QALE). In this example, it may be that the disease affected by treatment A only affects groups 1, 2, and 3 (and B only affects groups 3, 4, and 5), or it may be that the disease affects all individuals, but the treatment is only applicable to Groups 1, 2 and 3 (or groups 3, 4 and 5 of treatment B). Obviously, treatment A improves the health of the relatively disadvantaged group, while treatment B improves the health of the relatively disadvantaged group.
Then a key question is, how do you calculate the cost per QALY under the DCEA method? To answer this question, you need to complete steps 4, 5, and 6 above. I do this based on a preference for relative inequality aversion (Atkinson index) or absolute inequality aversion (using Kolm index).You can find a sample calculation workbook here.
The result of this simple example is shown in the figure below. Suppose the correlation inequality aversion parameter of the Atkinson index is 10.95 and the Colm index is 0.15. However, you can download a spreadsheet and try out how different inequality aversion parameters affect the results.