72 Chapter 3 characteristics with annual data from health insurance claims, between 2009 and 2014. We focus on yearly mental health care expenditures in the calendar years that individuals turn 17, 18 and 19; and define whether each individual used any basic or specialist mental health services during each calendar year. Therefore, the number of (person-year) observations included for each individual depends on the year of birth. For example, an individual born in 1990 will only contribute to the study with one person-year, when 19 years old; while and individual born in 1992 contributes with three person-years, when 17, 18 and 19 years old (see appendix C, Figure A1). In order to conduct subgroup analyses, two additional datasets are linked to characterize the study population in the calendar year they turn 17 years old (between 2007 and 2014). Based on tax records data we classify adolescents according to year and month-of-birth specific quartiles of standardised disposable household income (detailed explanation in appendix B.4). Based on medicines utilization data obtained from health insurance claims we observe whether individuals had any claims of antipsychotics (N05A), anxiolytics (N05B), antidepressants (N06A), or psychostimulants, agents used for attention-deficit/hyperactivity disorder (ADHD) and nootropics (N06B), as defined at the third level of the Anatomical Therapeutic Chemical (ATC) classification. For the subgroup analysis studying the effect of costsharing specifically on treatment discontinuation we restrict the overall study population to mental health care users at age 17. Due to lack of data to identify mental health users at 17 for the cohorts of 1990 and 1991, this subgroup analysis focuses on the cohorts from 1992 to 1997. Empirical strategy We use a difference-in-discontinuity approach [23, 24] to estimate the causal impact of increasing health insurance deductibles on mental service utilization during the transition to adulthood. This method exploits two sources of exogenous variation to disentangle the effect of increasing deductibles from disruptions in care use due to other factors. First, variation across individuals due to the exemption from the deductible when individuals are younger than 18. Second, variation in the deductibles over time, in particular the increase between 2011 and 2013. Based on this increase, we classify 2009, 2010 and 2011 as low deductible years (171, 180 and 181 euros in 2015 prices) and 2013 and 2014 as high deductible years (356 and 362 euros in 2015 prices). The deductible in the high deductible period was on average 181 euros higher than that in the low deductible period. We exclude 2012 from the analysis due to a one-year installment of co-payments for adult specialist mental health care (appendix A.3). We use ordinary least squares regression to estimate model (1), composed of the deviation between a standard RDD for the high deductible years – γ + k(Age* itm)+Eit[δ +l(Age * itm)] – and an RDD for the low deductible years – α + f (Age* itm)+Eit[β +g(Age * itm)] :
RkJQdWJsaXNoZXIy MjY0ODMw