Page 92 - Effective healthcare cost containment policies Using the Netherlands as a case study - Niek W. Stadhouders

Chapter 4
thresholds. For example, cross-country research from 2013 finds marginal effects of €20,000 to €30,000 per life year gained for the Netherlands5 (Heijink et al., 2013). One of the most extensive lines of research up to date has been performed by the Centre of Health Economics in England. This line of research, utilizing regional variation in spending, finds threshold values for England of £13,000 per QALY6 (Claxton et al., 2015a; Drummond et al., 2015). Extrapolating this finding to the Netherlands, accounting for income elasticity, renders threshold values of €21,000 to €29,000 (Woods et al., 2016). Possibly, the UK is more efficient or uses cost-effectiveness thresholds more strictly than the Netherlands. Also, thresholds may increase over time, as diminishing marginal returns make it more and more difficult to increase population health by one QALY (Barro, 1996; Murphy and Topel, 2003). Research from 2001 shows that the marginal cost per life year gained for a 65-year old increased from $121,000 in 1985 to $141,0007 in 1995 (Cutler and McClellan, 2001).
We distinguish between factors increasing uncertainty and factors that could potentially bias the estimators. Claims data may not represent unit costs due to internal cost shifting between departments of the hospital, which may increase uncertainty. The mortality dataset could contain measurement and classification errors. For example, in 2013 primary cause of death classification was altered (Harteloh, 2014). This may increase uncertainty of our estimations, but time dummies do not indicate any bias. Questionnaires used to infer quality of life might contain sampling uncertainty, interrater variation and framing issues, amongst others. For example, very ill patients may be underrepresented. Although quality of sampling by the Dutch Statistical Bureau was excellent, some inaccuracies may be expected. These data limitations primarily increase model uncertainty, but may also lead to small sampling bias in unknown direction.
4.4.2 Factors influencing threshold estimation
Increases in model uncertainty by data transformations are captured by the Monte Carlo simulations. However, some extrapolations were necessary, requiring additional assumptions. For example, we had to assume that disease-specific DALYs were stable over age, which may not be valid. Differences in estimations between age groups could in part reflect differences in burden of disease. Morbidity-related QALY loss for ages under 50 was extrapolated, assuming smooth trends. However, nonlinear trends may be present, e.g. when the very young patients are healthier than the linear extrapolation predicts. This could
5 A ratio of ~0.6 QALY per life year would imply thresholds of € 37,785 -€ 56,700 per QALY in 2014 euros
6 Or €19,200 in 2014 Euros
7 This is 203,724.08 US Dollars of 2014, or 153,176 Euros of 2014 per life year saved. Using a QALY per year ratio of ~0.6 this would be €255,293 per QALY. In comparison, we find for this age category a marginal value of € 65,000 per QALY.
84