Reply to the comment by Van der Deen et al.: "Possible methodological reason for the finding that 'Neither tax increase nor reimbursement reduced health disparities'".

Abstract

Corresponding Author: Eelco A. B. Over, PhD, Center for Nutrition, Prevention and Health Services, National Institute for Public Health and the Environment (RIVM), PO Box 1, 3720 BA Bilthoven, The Netherlands. Telephone: +31-30-274-2700; Fax: +31-30-274-4407; E-mail: eelco.over@rivm.nlReceived March 18, 2014; accepted March 25, 2014We would like to thank Van der Deen, Wilson, and Blakely for their thoughtful comments on our article (Over et al., 2014). They clearly explain the importance of modeling in order to investigate the long-term (cost-) effectiveness of tobacco prevention programs. We gladly take the opportunity to elaborate on our projections of health disparities and cost-effectiveness, starting with the sugges-tion that there may be a methodological reason for the finding that neither tax increase nor reimbursement reduced health disparities.We agree with Van der Deen et al. that socioeconomic status (SES) and age are not independent in the Netherlands. Also in the Chronic Disease Model, these parameters are dependent. This dependency is an additional important factor that affects SES dif-ferences in health gain and cost-effectiveness ratios, besides the five that we already mentioned: differences in baseline start and quit rates, differences in program participation, differences in pol-icy effect, and differences in all-cause mortality risks. According to the suggestion by Van der Deen et al., we reran our analyses for three age groups: 25–40, 41–60, and 61–80. Table 1 shows the stratified results. The cost-effectiveness ratios increase with age, so that the youngest age group is the most cost-effective. In addition, quality-adjusted life years gained per 1,000 smokers is lowest in the oldest group. Only for tax increases in the youngest age group, the cost-effectiveness ratio is slightly more favorable in low compared to high SES groups. However, none of the age groups showed reduced health disparities.The second comment of Van der Deen et al. refers to the importance of uncertainty analyses in cost-effectiveness mod-eling. This is certainly important, especially with regard to the price elasticity values. Because we were unable to find directly usable Dutch data on tobacco price elasticity in four-category-SES groups, we calculated best estimates based on adjusted for-eign data. These calculations obviously introduce an additional source of uncertainty. The health-disparity effect of reimburse-ment is also uncertain. For example, when we would have taken into account the uncertainty of participation for different SES groups in our reimbursement analyses, they would probably show increased as well as reduced health disparities. Van den Berg et al. did not find any significant differences in quit attempts between the SES groups (Van den Berg, Soethout, Hollinga, & Schipperen, 2009). With no significant differences in participa-tion, the smaller baseline quit rates of low SES groups drive the results toward a lower effectiveness in these groups.The final suggestion by Van der Deen et al. is to exploit the possibility of simulation modeling to evaluate “what-if” scenarios. Indeed, these possibilities are virtually infinite. However, we performed our studies in support of the Dutch Ministry of Health, Welfare, and Sport. Their main interest is in the evaluation existing or new actual policy measures that are feasible in “real life.” Evaluating continuous reimbursement is an interesting option however.Our initial expectations were that the financial incentives for smokers in the investigated interventions would provide an advantage to lower SES groups. This was mainly based on previ-ous reviews of the literature, revealing small advantages in effec-tiveness for the low SES groups for this type of interventions. The modeling results did not clearly show reduced health dispari-ties, and this may be explained by the rather unfavorable settings of the other factors mentioned previously for low SES groups: their baseline start rates are higher and quit rates are lower, they show lower participation and larger all-cause mortality, leaving less potential to profit from an effective intervention. It, therefore, seems that financial incentives alone are insufficient to reduce health disparities, and that every effort should be made to prevent tobacco use, particularly among lower SES groups. Our addi-tional analyses further show that even in younger age groups—that show most value for money for both the tax increase and reimbursement policies—these interventions did not reduce health disparities.