80 Chapter 3 Table 4. Differences in mental health care cessation between low and high deductible periods (panel A) by subgroups defined according to previous treatment intensity (panel B), number (panel C) and type (panel D) of psychotropic drug classes used (continued) High deductible (pp) Baseline use Relative change Observations Antidepressants -4.6 (-15.6 , 6.3) 49.4% -9.4% (-31.5%, 12.7%) 1,616 Agents ADHD* -10.1 (-19.4 , -0.7)** 31.6% -31.9% (-61.5%, -2.3%)** 2,232 Male Antipsychotics -9.6 (-22.9 , 3.6) 52.3% -18.4% (-43.8%, 6.9%) 957 Anxiolytics -11.9 (-48.5 , 24.7) .. .. .. 110 Antidepressants -11.6 (-29.7 , 6.4) 40.8% -28.5% (-72.9%, 15.8%) 697 Agents ADHD* -15.1 (-20.9 , -9.3)*** 26.9% -56.1% (-77.7%, -34.5%)*** 5,074 Notes: pp - percentage points, *p-value<0.05, **p-value<0.01, ***p-value<0.001, (Confidence Intervals at 95% level). a First quartile corresponds to the lowest mental health care expenditure group. Baseline use refers to the average mental health care use across study years at 228 months of age, corresponding to the first moment in which young adults pay the full-year deductible. The best fitting age polynomial is linear for both genders. Model estimated based on birth cohorts 1992 to 1992 and study years from 2011 to 2014. Estimates display main coefficients for each quartile/subgroup, obtained by repeating the estimation of the interacted model with each quartile/subgroup as reference category. Hence, statistical significance reported in the table concerns differences of the main term from zero. Within each model the interaction terms for the remaining quartiles are not significantly different from the main term (p-value for the coefficient > 0.05). All models control for migratory background, postcode and income quartiles at the year of 17th birthday. Relative size is not estimated for anxiolytics due to the small number of observation available to estimate the baseline use. Observations are in person-years. Sensitivity analyses The sensitivity analyses in Appendix G confirm the robustness of our estimates, both in terms of direction and size. We find consistent results when using larger age bandwidths, with estimates for males becoming more precise, and small and imprecise placebo discontinuities at age 16 (Table A3). Estimates obtained from a logistic model lead to similar conclusions, as well as those allowing for postcodeyear variation or not accounting for any of the controls (Table A4). DISCUSSION In this study we use an innovative quasi-experimental design to show that increasing health care deductibles led young adults in the Netherlands – mostly low-income females – to reduce the use of mental health services at the transition to adulthood. We find smaller and imprecisely estimated effects for males, which might be explained by several factors. First, male and female subpopulations have distinct mental health needs. Differences in the type of youth mental health disorders by gender are welldocumented in observational studies [2, 30], and are supported by the gender differences in psychotropic class consumption we observe. Second, young males disconnect more regularly from health care services during adolescence and are more often underserved [31, 32]. Reasonably, if factors as self-stigma, shame and masculinity [31] would already lead males to use treatment less ahead of the deductible increase, they might respond
RkJQdWJsaXNoZXIy MjY0ODMw