Page 86 - Effective healthcare cost containment policies Using the Netherlands as a case study - Niek W. Stadhouders

Chapter 4
Next, the elasticity was evaluated at the mean to obtain the marginal effect of spending for
(6)
the mean patient group:
(5)
Uncertainty with respect to the construction of the outcome variable was incorporated into the estimation by running 10,000 Monte Carlo simulations for all transformations combined, and separately for each individual transformation (Claxton, 2008). We incorporated uncertainty regarding the values for healthy life expectancy, quality of life gains, burden of disease and cost in last year of life (see section 2.1).
As robustness check, we tested differences in elasticity with respect to gender, age category and main disease category (appendix 4.2). However, these estimates should be treated with caution, as digression from the population mean reduces the accuracy and the validity of the Taylor approximation (Boisverf, 1982). As marginal values may depend on outcome variable and model specification used (Gallet and Doucouliagos, 2017), alternative outcome variables and model specifications were explored (appendix 4.3). We separately estimated mortality, death-related QALY loss and morbidity-related QALY loss as outcome measures. Furthermore, we estimated alternative model specifications; linear models and Cobb- Douglas (per patient) specifications. We included health trends and health shocks, specifically the percentage of (heavy) smokers, the percentage of obesity and the percentage of heavy drinkers.
4.2.5 Robustness checks
As patients could have been part of more than one patient group in the case of multimorbidity, in theory spending on one disease-specific patient group may influence mortality of another. We corrected for multimorbidity by defining unique patient groups based on primary diagnosis. After appointing deaths to the unique patient groups based on spending patterns on secondary diagnoses (proportionally or through OLS estimation), data transformations and estimations were performed according to figure 4.1.
4.3 Results
4.3.1 Summary statistics
Summary statistics of the data are presented in tables 4.1-4.3. Real hospital spending was relatively stable around €21 billion between 2012 and 2014. The number of patients declined slightly from 7.4 million to 7.1 million, as did the total number of deaths from 141,000 to 139,000. Per patient hospital spending was highest between the ages of 76 and
78