Accessibility statement

Linking the effects of expenditure on mortality burden of disease to QALYs

Overview

Claxton et al (2015) evaluated the relationship between expenditure and mortality using a cross-sectional design, seeking to identify differences in mortality across health care commissioning units (at the time of this research there were 152 Primary Care Trusts, or PCTs) that could be attributed to differences in NHS spend. Empirically, the research first quantified expenditure elasticities; that is, how changes in NHS expenditure in a given year were allocated between Programme Budgeting Categories (PBCs), which reflect broad disease areas characterised by International Classification of Disease (ICDs) codes.

Secondly, the research estimated outcome elasticities; that is, how changes in expenditure by PBC (in a particular year) altered PBC-specific mortality rates (using national data on mortality reported for ICD codes or groups of ICDs, mapped onto PBCs). Analyses adjusted for important covariates (including need) and used instrumental variables to estimate causal effects overcoming the problem of endogeneity.

Results showed that the mortality effects of changes in spend could only be identified for eleven of the 23 PBCs (such as cancer and gastrointestinal disorders). For the remaining disease areas (such as mental health disorders), health care focusses primarily on improving health-related quality of life (HRQoL). Across the 11 PBCs for which mortality effects were detected, empirically-based estimates of how changes in total NHS expenditure affect mortality were generated, returning the following point estimates (using 2008 expenditure and 2008–10 mortality): £105,872 for the cost per death averted; £23,360 for the cost per life year and £28,045 for the cost per life year where life years were adjusted for HRQoL. Methods used to estimate the cost per death averted and the cost per life year are detailed in Section “From mortality to life-years” (pages 46 to 54) and methods used to adjust life years for HRQoL are reported in the section “Adjusting life-years for quality-of-life” (page 56 to 60) in the HTA report. The results presented above for 2008 expenditure are reported in Table 163 (page 438) and Table 30 of the same report (page 74).

The econometric work done in the 2015 is detailed elsewhere and has been further extended since.

Critically, the work published in 2015 required linking the estimates of effects of changes in expenditure on the life year burden of disease to the likely effect on quality-adjusted life-years (QALYs), to obtain a cost per QALY that reflects the likely impact of changes in expenditure on both mortality and morbidity. The methods and data used to do this are detailed in Section “Using estimates of the quality-adjusted life-year burden of disease” in Chapter 4 of the HTA report (pages 66 to 71). In summary, such an estimate of health opportunity costs relevant for policy needs additionally to consider:

  1. whether changes in expenditure have effects beyond the year of expenditure – this can be termed duration of effects;
  2. how the effects of changes in expenditure on mortality relate to effects on a broader measure of health that incorporates both duration and HRQoL impacts (QALYs) – this can be termed surrogacy;
  3. how changes in expenditure affect health in disease areas for which the previous work could not measure a mortality effect – this can be termed extrapolation.

In the original research, very limited data were available with which to assess each of these questions, and hence the following assumptions were made: A. effects were assumed to be restricted to the year of expenditure change; B. surrogacy was assumed to be proportionate, i.e. the effects of changes in expenditure on mortality that were empirically estimated were used as the best estimate of the effects of expenditure on quality adjusted life-years; and C. extrapolation was also assumed to be proportionate, i.e. the effects of changes in expenditure on health (quality adjusted life-years) for disease areas which previous work could measure a mortality effect were used as best estimate of the effect of expenditure on health for disease areas for which previous work could not measure a mortality effect.

Using these assumptions, a ‘central’ estimate of health opportunity costs (expressed as a cost per QALY) across all disease areas of £12,936 per QALY. An analysis of the uncertainty imposed by the empirical estimates (the expenditure elasticities estimated for each of the 23 PBCs, and the outcome elasticities estimated for 11 of these) indicated that the probability of this central estimate being less than £20,000 per QALY was 0.89.

The methods, data and assumptions required to link the effects of expenditure on the mortality burden of disease to QALYs are further discussed in Chapter 5 “Implications for a policy threshold” of the HTA report (pages 73 to 99), and have also been discussed in the following response to the OHE critique. 

Health opportunity costs: assessing the implications of uncertainty using elicitation methods with experts

The assumptions that are required to link the estimates of effects of changes in expenditure on the mortality burden of disease to the likely effect on QALYs constitute important sources of uncertainty. To inform these assumptions appropriately, the judgements of key individuals, such as those with substantive clinical or policy expertise, are important. Elicitation methods offer a systematic process for formalising and quantifying, typically in probabilistic terms, individuals’ judgements about uncertain quantities.

Further research was commissioned by EEPRU to develop an application of structured elicitation to inform the highlighted uncertainties in estimates of expected health opportunity costs in the UK NHS. This research has been conducted and further information is available in the following documents.

Report:

Additional material: