Data sources and study populationCohort of parents with depression and antidepressant initiation

To identify parents with a new episode of depression, the following Swedish registers were utilized: the National Patient Register (NPR) [30], the Prescribed Drug Register [31], the Longitudinal Integrational Database for Health Insurance and Labor Market Studies (LISA) [32], the Multi-Generation Register [33] and the Total Population Register (TPR) [34]. Individuals 18 or older who met the following criteria were included: (a) filled a prescription for an antidepressant drug (AD; Anatomical Therapeutical Chemical [ATC] code N06A) without any previous AD dispensing in the preceding 180 days in Sweden between July 1, 2006, and December 31, 2018 (hereafter referred to as the index AD fill); (b) received a diagnosis of a depressive episode (International Statistical Classification of Diseases and Related Health Problems, 10th Revision [ICD-10] code F32 or F33) in specialist care within 30 days before or up to 365 days after the index AD fill; and (c) being a parent. For a flow-chart illustrating the cohort selection and subsequent matching procedure, see Fig. 1.

Fig. 1

Study population selection flow-chart. AD = antidepressant, ECT = electroconvulsive therapy, rTMS = repetitive transcranial magnetic stimulation, TRD = treatment resistant depression

Exclusion criteria included any history of: a) an ICD-10 diagnosis of (a) dementia, (b) bipolar disorder or manic episode; or (c) psychotic disorders, as well as having filled a prescription in the 180 days preceding the index AD fill of (d) antipsychotics or lithium; or (e) anticonvulsants/mood stabilizers. For all ICD-10 codes and ATC codes used, see Suppl Table 1. Further, patients with the following measure codes in the NPR within 180 days before the index AD fill were excluded: (f) electroconvulsive therapy (ECT); (g) vagus nerve stimulation; and (h) repetitive transcranial magnetic stimulation (rTMS). Finally, i) patients who were not residents in Sweden in the two calendar years preceding the year of the index AD fill were excluded.

The commonly used definition of TRD, i.e. two adequate treatment trials without satisfactory response, was adapted for this study through an algorithm used in previous studies from our group, described in detail elsewhere [35]. Patients initiating a third treatment trial within 365 days from the index AD fill, using antidepressant monotherapy, combination therapy, add-on therapy with antipsychotics or mood stabilizers, or Electroconvulsive Therapy (ECT) or repetitive Transcranial Magnetic Stimulation (rTMS), were classified with TRD. The medication trial had to last at least 30 days according to package size and dosing texts.

Matching of offspring with general population comparators

From the cohort of parents with depression and antidepressant initiation, those parents were selected who fulfilled the TRD criteria when their oldest offspring was between 6 and 15 years of age. These parents and their oldest offspring formed TRD parent-offspring pairs. Each TRD parent-offspring pair was matched 1:1 with a parent-offspring pair from the general population in which the parent satisfied the following eligibility criteria for matching: (a) no record in the registers of an ICD-10 diagnosis of depression or any of the ICD-10 diagnoses listed as exclusion criteria for inclusion in the depression cohort above, (b) no ATC codes or measure codes corresponding to the exclusion criteria for the cohort of parents with depression and antidepressant initiation; and (c) resident in Sweden in the two calendar years preceding the year of the index AD fill of the TRD parent. The date of matching was the date of TRD criteria fulfillment. Matching criteria were: (a) age of the offspring at the matching date, (b) sex of the offspring (c) birth rank of the offspring, i.e. the oldest offspring, (d) age of the parent at the index AD fill date of the TRD parent, (e) sex of the parent, and (f) type of living area of the parent at the index AD fill date of the TRD parent (classified as cities, towns and suburbs, or rural areas).

Matching of TRD parent-offspring pairs to parent-offspring pairs with other parental depression

Additionally, each TRD parent-offspring pair was matched to a parent-offspring pair in which the parent was included in the cohort of parents with depression and antidepressant initiation but did not meet TRD criteria at the matching date, hereafter termed “other depression”. Matching criteria were: (a) age of the offspring at the matching date (± 1 year), (b) sex of the offspring, (c) birth rank of the offspring, i.e. the oldest offspring, (d) age of the parent at the index AD fill date (± 5 years), (e) sex of the parent, (f) type of living area of the parent at the index AD fill date, (g) year of the index AD fill date.

The age matching criteria were wider for depression comparator pairs due to a lower number of comparators available. All matching was exclusive, i.e. without replacement. An offspring could not be available for matching as a parent at a later date. If a parent comparator with other depression later fulfilled TRD criteria, he/she would be followed as such and matched with a new other depression comparator at that date. If both parents were available for cohort selection, only one parent was selected for each offspring. The offspring formed a pair with the parent who first fulfilled the TRD criteria. If none of the parents fulfilled the TRD criteria, the offspring formed a pair with the parent who first fulfilled the criteria for the cohort of parents with depression and antidepressant initiation. Each TRD parent-offspring pair formed a matched group together with its two 1:1 matched parent-offspring pairs.

Covariates

The analysis was adjusted for several covariates regarding parental demographics, parental clinical history, parental history of social benefits, and offspring clinical history. Parental demographics were (a) country of birth (Sweden, Europe other than Sweden, Rest of the world); (b) educational level according to the latest available information in the LISA database (< 9 years, 10–12 years, > 12 years); (c) family situation according to the latest information in the TPR at time of inclusion (Married or cohabitant without children, Married or cohabitant with children, Single household without children, Single household with children); and (d) number of children according to the latest information in the TPR at time of inclusion (1, 2, > 2). The variables describing educational level, family situation and number of children were based on register information from the end of the calendar year preceding the calendar year of the index AD fill. See Fig. 2 for a visualization of time frames for definition of covariates and outcomes.

Fig. 2

Time intervals for selection of covariates and outcomes. 1Any non-psychiatric healthcare visit. 2Any psychiatric diagnosis. 3History of substance use, depression, other psychiatric disorders or self-harm/suicide attempts. 4Number of sick leave days and/or disability pension. 5Country of birth, attained education level, family situation, number of children, latest available data at end of calendar year preceding cohort entry. 6Not having completed secondary school during the year when reaching 19 years of age and sick leave with psychiatric diagnosis and disability pension from July 1 in the calendar year when the offspring reached 19 years of age. 7Any psychiatric contact, depression diagnosis, psychiatric medication, suicide attempt, and suicide.

Covariates regarding parental clinical history were (a) depression (No or Yes) based on main- and secondary ICD-10 diagnoses; substance use disorder (No, Alcohol only, Other/combined) based on main- and secondary diagnoses and filled prescriptions of drugs; (b) other major psychiatric disorders (No, Yes), defined as anxiety disorders, obsessive compulsive disorders, eating disorders, personality disorders, autism and attention deficit hyperactivity disorder based on main- and secondary diagnoses; (c) history of self-harm/suicide attempts (No, Yes) based on external cause codes; and (d) non-psychiatric healthcare use (No, Yes) based on main diagnoses. The variable describing depression was created based on information from the 5 years preceding the index AD fill date, excluding the 180 days wash-out period preceding the index AD fill date (1645 days), and the variables describing substance use disorder, other psychiatric disorders, self-harm and non-psychiatric healthcare use were created based on information from the 5 years preceding the index AD fill date (1825 days). All codes used are listed in Suppl Table 1.

Parental history of social insurance benefits were (a) sick leave days (0, 0–90, 90–365); and (b) disability pension (No, Yes). The variables describing sick leave days and disability pension were created based on net days with benefit from the social insurance agency in the year preceding the index AD fill date (365 days).

Offspring clinical history were (a) Psychiatric disorders (No, Yes) based on main- and secondary diagnoses; (b) self-harm (No, Yes) based on external cause codes; and (c) non-psychiatric healthcare use (No, Yes) based on main diagnoses. The variables describing mental and behavioral disorders and self-harm were created based on all available information from before the matching date, and the variable describing non-psychiatric healthcare use was created based on information from the year preceding the matching date (365 days).

Outcomes

Four clinical and three socioeconomic outcomes were measured for the offspring. Offspring clinical outcomes were (a) overall contact with psychiatric care based on medical specialty codes; (b) depression diagnosis based on main- and side diagnoses; (c) psychiatric medication based on filled prescriptions of drugs; (d) suicide attempt based on external cause codes; and (e) suicide recorded as cause of death. Offspring socioeconomic outcomes were (f) not having completed secondary school in the year when reaching 19 years of age; (g) sick leave with psychiatric diagnosis; and (h) disability pension.

All clinical outcomes were measured as time to the event of interest from the matching date. Sick leave with psychiatric diagnosis and disability pension were measured as time to the event of interest from July 1 in the calendar year when the offspring reached 19 years of age. The following were censoring events: death, emigration, end of follow-up time (Dec 31 st 2018), or a parental comparator with other depression meeting TRD criteria. Not having completed secondary school in the year when aged 19 was measured as a binary variable (No, Yes) on Dec 31 st in the calendar year when the offspring reached 19 years of age. Measurement of the socioeconomic outcomes was restricted to offspring for whom none of the censoring events had occurred prior to July 1 st or Dec 31 st in the calendar year when the offspring reached 19 years of age.

Statistical analyses

Separate analyses were conducted for each outcome, not taking the other outcomes into account. All outcomes, except not having completed secondary school at 19 years of age, were analyzed using stratified Cox regression, taking the matched design into account by allowing the baseline hazards to differ across matched groups. Follow-up started at the matching date, except for sick leave with a mental diagnosis and disability pension, for which follow-up started on July 1 in the calendar year the offspring turned 19. Follow-up ended with the outcome of interest or any of the censoring events, whichever came first. Not having completed secondary school at age 19 was analyzed using conditional logistic regression. Both unadjusted and adjusted analyzes were conducted. The adjusted analyzes included all covariates regarding parental demographics, parental clinical history, parental history of social benefits, and offspring clinical history. Stratified analyzes were also conducted on subsamples based on age of the offspring: 6–10 years or 11–15 years. Offspring to parents without depression was selected as the reference group. Robust standard errors were calculated. Proportional hazards were tested and met using Schoenfeld residuals.

Additional stratified analyzes were also conducted on subsamples based on either the sex of the offspring or the sex of the parent, in combination with age of the offspring.