Jeroen van de Pol

100 Chapter 4 of the CMRs performed in this study. The CMRs in this study focused mainly on the patient's preferences related to their medication and health, and could thereby improve their self-experienced quality of life, whereas other studies in medication review often focused on optimizing treatment according to guidelines [14,33,34]. The goals in the CMRs could be solved by drug changes performed during CMRs as shown by an in depth-analysis of the DREAMeR study [36]. Health utility values did not change significantly. This may be explained by the fact that the EQ-5D is less responsive compared to the EQ-VAS, especially when baseline values are high [37,38]. However, VAS is not a generally accepted way to measure utilities, due to the risk of end aversion bias. To negate end aversion bias, the possibility of conversion of VAS scores has been explored [39]. However, it was chosen not to convert VAS scores due to the fact that utilities measured with EQ-5D and EQ- VAS do not differ substantially, so the presence of end aversion bias seems very limited. Also, this study was conducted among patients aged ≥ 70 years and using at least seven drugs, which also reduced the chance of giving a VAS score of 100, and therefore introducing the risk of end aversion bias. VAS does, however, give some additional information on the (improved) health status experienced by the patient themselves and was therefore used to calculate utilities as well. A previous study conducted in Spain illustrated that their medication review decreased costs, increased HR-QoL measured with both EQ-5D and EQ-VAS and was also seen as the dominant strategy over usual care [19,40]. The effects on HR- QoL were even higher than the effects in our study. Although this Spanish study was not explicitly designed as a patient-centered intervention, CMR in this study was accompanied with many follow-up contacts, which probably contributed to the patient-centeredness of the study. Costs in the Spanish study were not directly comparable to the Dutch situation as these were not calculated from a societal perspective. A decrease in drug costs and hospital admissions was also demonstrated by Desborough et al., but they did not show effects on HR-QoL measured with EQ-5D [41]. The average healthcare costs of the patients in this study are representative of the current Dutch situation for this age group [42]. A CMR could lead to small cost savings in healthcare compared to usual care and an average reduction of 0.5 in the number of drugs in use after 6 months [32]. Although the variation for each cost category was high in both groups, the results are strengthened by the sensitivity analyses, which show that the analysis is robust to variations in variables. The probability of cost savings in healthcare consumption is high ( > 90%) according to the cost-effectiveness planes of the ICERs. The costs with the highest influence on the variability of the estimated cost savings were the intervention costs and the costs of institutional care and secondary care.