Page 90 - Effective healthcare cost containment policies Using the Netherlands as a case study - Niek W. Stadhouders

Chapter 4
We estimated separate thresholds per gender, age and disease groups (see appendix 4.2). A higher threshold was found for males, but not significantly. This may indicate that spending on females is more beneficial, possibly due to a higher life expectancy. Thresholds over age groups are strikingly constant save the neonates and 95+ year-olds, indicating that spending on age groups largely takes into account healthy life years to be gained, and that a discount rate of 1.5% seems appropriate.
4.3.3 Results per patient group
Differences between disease categories are larger. For some disease categories high thresholds were found, i.e. diseases of the blood and pregnancy, while lower thresholds were found in other categories, specifically diseases of the nervous system and diseases of the skin. Potential explanations for these differences include inefficient allocation patterns, differences in QALY valuations, deviations from the mean in the translog estimation and measurement errors. More research is required to assess the clinical relevance of these differences. For most patient groups valid and significant thresholds were found, indicating the robustness of the estimation strategy.
The robustness checks show that the outcome was sensitive to the structural model employed, but the translog model generally was robust to different specifications (see appendix 4.3). When only patient groups with mortality were included a value of €61,100 per QALY was found, suggesting that our combined measure of disease and mortality related QALYs may not fully capture all health gains. Excluding morbidity-related QALY loss (step 1.1-1.4) raised the threshold to €89,000 per QALY. Estimating the relation between spending and mortality directly resulted in an estimated effect of €275,000 per death averted. Backwards calculations to QALYs rendered a threshold of €42,000 per QALY. Using 2-year QALY gains as outcome measure lowered the threshold slightly to €60,000 per QALY, suggesting spending affecting health outcomes primarily in the same year, but possibly also in the next years. However, estimating the effect of spending in year t on outcomes in year t+1 rendered insignificant and economically unlikely results. This may indicate that our 3- year panel dataset is too limited to estimate robust lagged effects. Neither time dummies to correct for technology shocks, nor health trends to correct for omitted variable bias influenced the threshold estimates, suggesting our OVB correction is appropriate.
4.3.4 Robustness checks
Multimorbidity corrections produced divergent results. Proportional multimorbidity corrections resulted in a threshold of €201,000 per QALY (€143,000 - €271,000 per QALY). However, a proportional distribution of deaths implies that higher spending increases the proportion of total deaths being appointed to that patient group. This increases reverse
82