Update: eliciting societal preferences for weighting QALYS according to burden of illness, size of gain and end of life

This report presents the findings of a study to elicit societal preferences and weights for Quality Adjusted Life Year (QALY) gains across three characteristics: 1) Burden of illness (BOI) from a medical condition given current health care interventions – defined as the QALY loss per patient from a condition due to premature mortality and/or reduced health-related quality of life (HRQOL) measured against life expectancy and health-related quality of life without the condition; 2) Therapeutic improvement (TI) - whether preferences for large QALY gains are disproportionately larger than the size of the gain (e.g. weight a QALY gain of 2 more than 4 times a QALY gain of 0.5) and 3) End of life (EOL) - defined by NICE as expected survival of less than 2 years and expected survival gain of 3 months or more.

Methods

A survey using a Discrete Choice Experiment (DCE) was conducted with an online general population sample using an existing panel. Respondents were asked to choose whether they thought the NHS should treat patient group A or B, who differed in terms of four attributes: life expectancy without treatment, HRQOL without treatment, survival gain from treatment and HRQOL gain from treatment. These attributes were used to derive BOI, QALY gain and EOL. The questionnaire had four variants, each with a different life expectancy without the condition (5, 20, 40 and 80 years). Each respondent answered questions for one variant and made comparisons between groups with the same life expectancy without the condition. Choices were analysed using conditional logistic regression with a range of specifications. Robustness across the four levels of life expectancies without the condition and to various exclusions was examined. Weights were estimated using the marginal rate of substitution.

Results

In total, 3669 respondents completed the survey. The sample was largely representative of the population of England for age and gender, but there were some differences in other characteristics. Regression results indicated that respondents preferred to treat patients with larger QALY gains, but at a diminishing rate meaning there was no support for TI. Respondents preferred to treat patients with a shorter life expectancy (EOL). Results suggested some support for BOI but were not robust across alternative model specifications. The coefficients varied as life expectancy without the condition varied. Regressions estimated excluding respondents who were identified as possibly misunderstanding the DCE task (remaining sample of 2247 respondents) had positive, significant and robust coefficients for BOI. Using the marginal rate of substitution to estimate weights indicated that 1 unit of BOI is equivalent to 0.04 QALYs gained, and EOL is equivalent to 3.331 QALYs gained (assuming that the value of a QALY does not change with size of QALY gain).

Discussion

This study provides the first attempt to operationalize the concept of BOI by combining the conventional notion of severity (in terms of poor health) with survival, using QALY loss attributable to the condition. The results indicate general support for maximising QALY gains, but at a diminishing rate, meaning that the evidence did not support the idea of TI. The results indicate some support for BOI as a consideration when weighting QALYs. There is robust and consistent support for EOL in general (but this conceptually overlaps with BOI and the two should not be used together). Overall the results indicate that a QALY is not a QALY regardless of the burden of the disease or life expectancy and provide a basis for determining appropriate QALY weights.