Page 83 - Effective healthcare cost containment policies Using the Netherlands as a case study - Niek W. Stadhouders

The marginal benefits of healthcare spending in the Netherlands
decline. This would reduce the size of the patient group, while mean patient health may be unaffected. Therefore, changes in patient QALY scores could be fully attributed to the health sector. Extra spending may increase quality of life of patients who would not have died, but may also avert deaths of patients who would have, rendering the effect on average quality of life of all patients ambiguous (Ochalek et al., 2015). While the health questionnaires were used to measure the primary effect of increases in quality of life of patients (step 1.1 to 1.4), the effect of lower mean quality of life due to increases in survival was introduced in step 2.3 (described below).
Mortality statistics were collected by Statistics Netherlands and contained all nationwide deaths in a given year including information on age, gender and primary cause of death according to the 3-digit ICD-10 codes. The ICD-10 codes allowed appointment of deaths to the same patient groups as defined by claims data (step 2.1). In 3713 patient groups, at least one death was recorded. In total 94% of all deaths were appointed to a patient group with positive spending.
From mortality data to death-related QALY loss
To transform the number of deaths to death-related QALY loss per patient group, we followed Claxton et al. (2015). Contrary to the UK, estimates of healthy life expectancy were readily available in the Netherlands (CBS, 2018). This allowed us to compute healthy years of life lost for deaths in all age groups (step 2.2). Some of the benefits of averted deaths are in the future, requiring discounting to calculate the net actuarial benefit of averted deaths. Following Dutch guidelines, in step 2.3 we apply a discount rate of 1.5% (Zorginstituut Nederland, 2015). As literature provides no consensus on the appropriate discount rate (Claxton et al., 2011), we allowed discount rates to vary between 0 and 5 percent in our sensitivity analysis. If a death is averted, a patient may not fully return to the average health status of the population. Therefore, in step 2.4 we used Dutch disease-specific disability- adjusted life years (DALY) estimates from Hoeymans et al. (2014) to correct for burden of disease (see table 4.2). DALY values ranged between 0 and 1 and included utility losses due to premature death and lower health in life (Hoeymans et al., 2014). In our research, healthy life years were reduced by the relative DALY burden (step 2.4), e.g. a DALY of 0.1 resulted in a 10% reduction in disease specific healthy life years relative to the healthy population. Steps 2.2 to 2.4, rendering the number of QALYs lost due to mortality (Gafni and Birch, 1993), introduced extra uncertainty in the estimates, which was evaluated using Monte- Carlo analysis. By adding the number of QALYs lost due to mortality to the number of QALYs lost due to morbidity (step 2.5), total QALY loss for each of the 11,000 patient groups was obtained.
75