2.1 Data and study population
This is a retrospective cohort study based on longitudinal health claims data from the insurance group Helsana. Helsana is one of the largest health insurers in Switzerland. In 2021, it covered approximately 1 330 000 (15%) of the Swiss population. All Swiss residents need to contract a mandatory health insurance on the private market. Every person chooses a monthly deductible amount (franchise) between 300 to 2500 Swiss Francs per year. Health insurance can either be a standard care or managed care model. For example, in the family doctor managed care model, the insured has a physician who acts as the first provider and coordinator of care when medical care is needed. Basic health insurance covers the costs in the event of illness, accident, and maternity, including health care and pharmacy invoices for health care use and prescription drugs. For the present study, we included people aged 18 and older, who had at least one AD prescription between 2013 and 2021.
2.2 Variables
2.2.1 Antidepressants
ADs are grouped according to the Anatomical Therapeutic Chemical (ATC) classification system of the World Health Organization with the code N06A. [19] The group is subdivided according to different modes of action: 1) tricyclic ADs (TCAs, N06AA), 2) selective serotonin reuptake inhibitors (SSRIs, N06AB), 3) monoamine oxidase inhibitors (MAOIs, N06AG), and 4) other ADs (N06AX), which have unique structures and properties that target diverse receptors in the central nervous system, e.g. serotonin norepinephrine reuptake inhibitors (SNRIs) or tetracyclic ADs (TeCA). [19] For our analyses we distinguish between TCAs, SSRIs, MAOIs, other ADs, and separately between mirtazapine (SNRI), venlafaxine (TeCA), and St. John’s wort (other ADs).
2.2.2 Categories of length: Short-, medium- and long-term user
To analyse the length of prescriptions, we classified people as short-, medium-, or long-term user. To do so, we tracked the first AD prescription for each user between 2013 and 2021 without a prescription in 2012 to analyse new prescribers. Then, we counted the numbers of quarters with at least one AD prescription to analyse the length of AD prescription and distinguish between long-, medium- and short-episodes. Short-term is defined as a prescription length of one or two quarters (= max. six months), medium-term as a prescription length of three or four quarters (= max. 12 months) and long-term as a prescription length of five or more quarters (= more than 12 months). If there is a break between quarters of AD prescription, we allowed for an interruption of maximum 6 months before we labelled it as a new episode of AD prescription. Examples of episodes in each category are shown in the Supplementary Material. To analyse the numbers of people by categories of length across the years, people were allocated to one of the three categories for each year. If more than one episode in one year appeared, the longest episode was counted. We excluded all people with an AD prescription in 2013, which had a higher number of people, because it was the first observation year, making the following years comparable. We used the data from the whole year 2021 and the first quarter in 2022 to analyse categories of length for 2020, but we did not analyse the categories of lengths for the years 2021 and 2022 because long- and medium-term users would be underestimated. Therefore, long-, medium-, and short-term users were analysed for 2014 to 2020.
2.2.3 Population characteristics
Population characteristics comprise: sex, age, age groups (19-30, 31-40, 41-50, 51-60, 61-70, 71-80, >80), seven regions of Switzerland according to the Swiss Federal Office of Statistics (Zurich, Midland CH, Lemanic region, North-western Switzerland, Eastern Switzerland, Ticino, and Central Switzerland) [20], three area categories that display the level of urbanization according to the national community typology of the Swiss Federal Office of Statistics (urban, rural and intermediate area = dense urban areas and rural centres) [21], language regions (German, French, and Italian), health insurance plan (managed care vs. standard plan), franchise (low: ≤ CHF 500, high: > CHF 500), and living in a nursing home.
Treatment characteristics comprise the drug class of AD, number of prescriptions, prescription source (type of health care provider) and psychiatric or psychotherapeutic care. AD drug class includes the following categories: TCAs, SSRIs, MAOIs, mirtazapine, venlafaxine, St. John’s wort, and other ADs. Each AD prescription was coded to a prescription source that includes: 1) general practitioner (GP), 2) psychiatrists or psychiatric clinics, 3) hospital ambulatory and nursing homes, and 4) other medical specialists (e.g., gynaecologist). psychiatric or psychotherapeutic care included psychiatric or psychotherapeutic diagnostics and therapies by psychiatrists or psychotherapists in psychiatric clinics or at the doctor s office.
2.3 Statistical analysis
To analyse AD prevalence across the years, we calculated the number of people aged 18 or older who had at least one AD prescription in each year from 2013 to 2021. These numbers are provided both raw and extrapolated to the entire Swiss population using census data from the Swiss Federal Office of Statistics. [22] The procedure of extrapolation was used to adjust for age, sex and region (of residency). The prevalence rates are calculated with the extrapolated data to give an overview of the extrapolated one-year prevalence of people with at least one AD prescription in Switzerland between 2013 and 2021.
The total number of people with at least one AD prescription between 2013 and 2021 was calculated. All results were extrapolated relative to the demographic distribution of the overall Swiss population by each year. The extrapolations were based on individual weighting factors (wi), which were calculated as the inverse of the sampling probability (pi = NHelsana,i/NSwitzerland,i) of a given stratum (i): wi = 1/pi. The strata are defined by people s characteristics including age class, gender, cantons (regions areas) by year. The data for NSwitzerland,i is derived from the federal statistical office of Switzerland.[22]
Chi-square tests and ANOVA were used to test differences between short-, medium-, and long-term users in population and treatment characteristics. Associations between long-term users and population and treatment characteristics were tested using logistic regression analyses. The regression analysis model included sex, age, nursing home, insurance model, psychiatric or psychotherapeutic care, living area and language region to predict the probability to be a long-term user compared to short- and medium-term user. Results from the regression models were presented as odds ratios (ORs) with 95% confidence intervals (CI). For all tests p < 0.05 was considered statistically significant. All analyses were performed using the statistical software R, version 2022.02.3.[23]