OUP user menu

Socio-economic determinants of early discontinuation of anti-depressant treatment in young adults

Karolina Andersson Sundell, Margda Waern, Max Petzold, Mika Gissler
DOI: http://dx.doi.org/10.1093/eurpub/ckr137 433-440 First published online: 27 September 2011


Background: Early discontinuation of anti-depressant treatment is common. This study analysed whether socio-economic factors influence early discontinuation among new anti-depressant users aged 20–34 years. Methods: Our study population included all Swedes aged 20–34 years who purchased anti-depressants in 2006 and had not purchased such drugs in the preceding 6 months (n = 25 003). We obtained prescription data from the Swedish Prescribed Drug Register. Information about demographic and socio-economic factors (country of birth, marital status, household size, education level, occupation, income and social assistance) was collected from Statistics Sweden by record linkage. We defined early discontinuation as filling only one anti-depressant prescription within a 6-month period. We used multiple logistic regression analysis to analyse the socio-economic factors associated with early discontinuation. Results: We identified 6536 individuals (26.1%) as early discontinuers. Early discontinuation was less common among women [odds ratio (OR) = 0.82; 95% confidence intervals (CI) 0.75–0.87] and in those with at least two years of higher education (OR = 0.71; 95% CI 0.61–0.83), whereas it was more common among those born outside Sweden (OR = 1.76; 95% CI 1.48–2.10) and those who received social assistance (OR = 1.26; 95% CI 1.11–1.44). Compared with selective serotonin re-uptake inhibitors, SSRI, early discontinuation was more common among individuals who started treatment with a tri-cyclic anti-depressant, TCA, (OR = 2.58; 95% CI 2.24–2.98) or an anti-depressant other than SSRIs, TCAs or selective serotonin-norepinephrine re-uptake inhibitors/norepinephrine (noradrenaline) re-uptake inhibitors (OR = 2.90; 95% CI 2.05–4.10). Conclusion: Early discontinuation occurred more commonly among social assistance recipients and those with immigrant background, suggesting that those groups might require greater support when initiating anti-depressant therapy.


Anti-depressants are used for several indications although most commonly for depression.1,2 To increase the likelihood of successful treatment outcomes and reduce risk for relapse, current treatment guidelines recommend a treatment period that continues at least 6 months after determining that the treatment is effective.3 Early discontinuation of anti-depressants indicates suboptimal use and might affect the duration and recurrence of depression11,13,4

However, early discontinuation is common, ranging from 15% to over 55%.2,5–19 Indeed, many earlier studies reported that over one-fourth of all new users discontinue anti-depressants early.2,7,10,12–18,20

Previous international studies suggested that groups characterized by low education level, low income, unemployment, disability pensioners or born abroad or with other citizenship than the country where the study was conducted are more likely to discontinue treatment early.4,5,12,13 Compared to clinical characteristics, the relative impact of socio-economic factors differs among studies.5,12,14 However, several studies identified foreign background as a strong predictor of early discontinuation, sometimes as strong as clinical predictors.4,14 Few studies have examined how prescriber characteristics, e.g. whether the prescription was issued by a general practitioner (GP) or a psychiatrist, influence early discontinuation. However, a Danish study reported that prescriber characteristics are less important than clinical and socio-economic characteristics.12 However, some studies have only included anti-depressants prescribed by GPs.5,10,18 Furthermore, previous reports about differences in early discontinuation rates by age are inconsistent. Some studies noted age differences, but others did not.5,12,14

Sick-listings and disability pensions related to mental health problems, especially depression, have increased among Swedish adults aged ≤35 years.21 The age span from late teens to late 20s or early 30s is sometimes termed ‘emerging adulthood’, a relatively new phenomenon resulting from increased incidence of postsecondary education and postponement of family formation and childbearing.22 Some commonly used socio-economic indicators, e.g. income level, might be less sensitive during emerging adulthood, a period characterized by transition.22 Parental background likely impacts the health and health behaviour of young adults and increases the need for information about their parents’ history. Most previous studies investigating early discontinuation assessed only whether individuals were born outside the country where the study was conducted or had another citizenship.4,5,12,13 Therefore, we sought to examine whether early discontinuation of anti-depressants in Swedish adults <35 years of age linked with demographic and socio-economic factors accounting also for parental country of birth.

Our population-based national register study analysed whether early discontinuation among new anti-depressant users aged 20–34 years associated with socio-economic factors, prescriber characteristics, anti-depressant type and concurrent use of other medicines in Sweden. We also investigated whether socio-economic factors differed among anti-depressant users aged 20–34 years and a sample-based reference population taken from the general population in the same age group.


Study population and study period

The study population encompassed all Swedes aged 20–34 years who filled at least one anti-depressant prescription between 1 January and 30 June 2006 and had not filled an anti-depressant prescription in the preceding 6 months (n = 25 003). The first filled prescription in 2006 represents the index prescription and the index date for each individual. The study period began on 1 January 2006 and ended on 31 December 2006.

To compare background characteristics, we randomly selected a reference population of 500 000 individuals aged 20–34 years who filled no anti-depressant prescriptions in 2006 and lived in Sweden on 1 January 2006.


The Swedish Prescribed Drug Register, SPDR, provided information about all prescription drugs purchased by the study population between 1 July 2005 and 31 December 2006. The SPDR includes information about the purchased drug and the unit that issued the prescription.23 Medicines were classified using the Anatomical Therapeutical Classification system (ATC).24 Patients can purchase a maximum 90-day supply of continuous use medicines included in the Swedish Pharmaceutical Benefits Scheme per fill.25 When initiating treatment, individuals can purchase either a 90-day supply or a start package that includes a maximum supply of 30 days. Around half of our study population purchased a start package on the index date.26

We defined anti-depressants as substances included in ATC-group N06A. Anti-depressant types included tricyclic anti-depressants (TCA) (ATC-group N06AA); selective serotonin re-uptake inhibitors (SSRI) (ATC-group N06AB); serotonin-norepinephrine re-uptake inhibitors and norepinephrine (noradrenaline) reuptake inhibitors (SNRI/NRI) (ATC-codes N06AX21, N06AX16, N06AX18) and all other anti-depressant medications. Characteristics of the healthcare facility that issued the index prescription included type of care facility (primary/ambulatory care, specialized or hospital care and other) and sphere of activity (primary care, general psychiatry, other psychiatry, occupational health services, internal medicine, neurology, pain care, oncology and others).

We defined concurrent use of other prescription medications as at least one non-anti-depressant prescription filled within the 180 days following the index date. We calculated the total number of concurrent drugs purchased during that time period. In addition, we classified concurrent use of specific medicines such as anti-diabetic drugs (A10); drugs for acid-related disorders (A02B); anti-hypertensive drugs (C03, C07-C09); beta-blockers (C07); anti-acne preparations (D10A); oral contraceptives (G03A); thyroid therapy (H03A); antibiotics (J01); analgesics (M01 and N02); anti-migraine drugs (N02C); opioids (N02A); codeine (N02AA59); tramadole (N02AX02); anti-epileptics (N03A); anti-psychotics (N05A); sedatives (N05B); hypnotics (N05C) and anti-asthma drugs (R03A and R03B).

We collected socio-economic data from both the longitudinal integration database for health insurance and labour market studies (LISA) and the Total Population Register. We grouped marital status (defined as the marital status recorded on 31 December 2006) into four categories: single, married (including registered partnership for same-sex couples), divorced and widow/widower. We also grouped country of birth into four categories: (i) Sweden, (ii) other western industrial countries (the Nordic countries, the 25 European Union countries [EU25]), North America and Oceania), (iii) eastern European countries (including those outside EU25) and the former Union of Soviet Socialist Republics and (iv) other countries (Africa, South America and Asia). A composite variable, denoted ‘background,’ reflected the national origin of participants and their parents: (i) Swedish (born in Sweden, two parents born in Sweden); (ii) born in Sweden, one parent born outside Sweden; (iii) born in Sweden, both parents born outside Sweden; (iv) born outside Sweden, at least one parent born in Sweden and (v) born outside Sweden, two parents born outside Sweden.

The number of household members and information about whether an individual had received study allowances or unemployment benefits were assessed in 2005. We defined social assistance (a cash benefit received in the absence of other income) at the household level, i.e. had the individual or any other household member received social assistance in 2005 (yes/no). Social assistance only includes a cash benefit given to those with an absence of other income. It does not include other benefits, e.g. study allowances or unemployment benefits. We also recorded the highest level of education attained at the end of the 2006 spring term. We defined occupational status as employment/unemployment in November 2005 and also included an individual's occupation type (upper white-collar; lower white-collar; blue-collar and others including armed forces, farmers, market gardeners and forestry and related workers. Income/household income data referred to year 2005. We used the median annual income and an annual income <60% of the median income as income indicators in the study population. We collected date of death and migration data from the LISA-database at the end of the study period.

Ethics approval, a prerequisite for data access, was obtained from the regional ethics board in Gothenburg.

Outcome measures

The outcome was early discontinuation of anti-depressant use defined as no anti-depressant refills within 6 months after index date.

Statistical analyses

We used Pearson's χ2-test to analyse differences between groups for categorical variables and also between the study population and the reference population. Means and standard deviations (SDs) were presented for continuous variables as well as median values and inter-quartile range if these differed. We used the Student's t-test to analyse differences between groups for continuous variables. Differences where P < 0.05 were considered to be statistically significant.

We used multiple logistic regression analysis to identify factors associated with early discontinuation, and included variables selected a priori with a P < 0.2 in the univariate logistic regression analyses in the multiple logistic regression models. Independent socio-economic variables included sex, age (grouped in 5-year classes), marital status, background, level of education, student status, unemployment, socio-economic status, having received social assistance and income <60% of the median income. We also considered the type of anti-depressant purchased on the index date and the sphere of activity of the health care facility where the index prescription was issued. We analysed the data for both the entire study population and the groups stratified by 5-year age groups. Data analyses were carried out in SAS version 9.1 (SAS Institute Inc, NC, USA, 2005).


The study population comprised 25 003 individuals aged 20–34 years who initiated anti-depressant use during the first 6 months of 2006. In total, 6536 (26.1%) individuals filled one prescription only once (table 1). A larger proportion of the early discontinuers had completed only mandatory education or had received social assistance, whereas a smaller proportion of the early discontinuers had a Swedish background and higher education (≥2 years) compared to those who purchased refills (table 1). Furthermore, a smaller proportion of early discontinuers purchased an SSRI on the index date. A smaller proportion of early discontinuers received their index prescription in a general psychiatry setting compared to those who filled at least two prescriptions (table 2).

View this table:
Table 1

Distribution of socio-economic characteristics and type of antidepressant at baseline among individuals who discontinued anti-depressant use early (n = 6536) and individuals who did not (n = 18 467)

CharacteristicNo early discontinuationEarly discontinuationP-value
n (%)n (%)
Woman11 771 (63.8)3930 (60.2)<0.001
Age (years)
    20–244856 (26.3)1663 (25.4)0.849
    25–296140 (33.3)2170 (33.2)
    30–347471 (40.5)2703 (41.4)
Marital status
    Unmarried14 423 (78.6)4839 (75.2)<0.001
    Married3086 (16.8)1222 (19.0)
    Divorced or widow/widower832 (4.5)378 (5.9)
Number of people in the household
    19258 (50.6)3098 (47.9)<0.001
    22117 (11.6)826 (11.6)
    3–56668 (36.4)2397 (36.4)
    6 or more261 (1.4)142 (2.2)
    Born in Sweden, two parents born in Sweden13 159 (79.3)4126 (74.3)<0.0001
    Born in Sweden, one parent born in Sweden1840 (11.1)600 (10.8)
    Born in Sweden, two parents born outside Sweden863 (5.2)362 (6.5)
    Born outside Sweden, one or both parents born in Sweden63 (0.4)25 (0.5)
    Born outside Sweden, two parents born outside Sweden671 (4.0)441 (7.9)
Highest attained education
    Mandatory education (0–10 years)3415 (19.0)1649 (26.0)<0.0001
    Upper secondary school8792 (48.8)3118 (49.2)
    Higher education < 2 years1574 (8.7)395 (6.2)
    Higher education ≥ 2 years4244 (23.6)1171 (18.5)
Activity on the labour market
    Employed7998 (43.7)3019 (46.7)<0.001
    Student5006 (27.4)1504 (23.3)0.001
    Received social assistance3103 (17.0)1532 (23.7)<0.001
Socio-economic status
    Upper white collar worker3327 (24.9)944 (21.2)<0.001
    Lower white collar worker4839 (36.2)1593 (35.8)
    Blue collar worker5085 (38.0)1866 (41.9)
    Other134 (1.0)46 (1.0)
    Disposable income < 60% of the median value4171 (22.6)1554 (23.8)0.049
Type of anti-depressant purchased at baseline
    SSRI14 519 (78.6)4613 (70.6)<0.0001
    TCA997 (5.4)822 (12.6)
    SNRI/NRI2805 (15.2)991 (15.2)
    Other anti-depressant146 (0.6)110 (1.7)
  • Percentages of missing value are given as per cent of total for each category where available.

View this table:
Table 2

Early discontinuation by characteristics of prescribing unit issuing the index prescription for individuals who discontinued early (n = 6536) and individuals who did not (n = 18 467). Percentages of missing value are given as percent of total for each category, where available

CharacteristicNo early discontinuation (n = 18 467)Early discontinuation (n = 6536)P-value
n (%)n (%)
By type of care
    Primary care/ambulatory care8640 (44.5)2502 (44.7)0.397
    Specialized care/Inpatient care8588 (44.2)2430 (43.5)
    Other2183 (11.3)2183 (11.8)
By sphere of activity
    Primary care9215 (56.7)3293 (58.1)<0.001
    Psychiatry—general5399 (33.2)1546 (27.3)
    Other psychiatry341 (2.1)122 (2.2)
    Occupational health service450 (2.8)129 (2.3)
    Internal medicine382 (2.4)216 (3.8)
    Neurology145 (0.9)73 (1.3)
    Pain care38 (0.2)13 (0.2)
    Oncology14 (0.1)9 (0.2)
    Other281 (1.7)265 (4.7)

On average, each member of the study population used 2.0 (SD 2.1) concurrent prescription medicines (median 1). The number was slightly lower among early discontinuers [mean 1.7 (2.0), median 1] compared to those with refills [mean 2.0 (2.2), median 1], P < 0.001. The most commonly used medicines were antibiotics (41.7%), sedatives (33.6%), hypnotics (32.3%) and, among women, oral contraceptives (45.4%). The proportion of concurrent use of sedatives (25.5% vs. 36.4%, χ2 258.7, df 1, P < 0.001), hypnotics (27.4% vs. 34.1%, χ2 99.4, df 1, P < 0.001) and oral contraceptives (41.7% vs. 46.6% of women, χ2 27.7, df 1, P < 0.001) was lower among early discontinuers compared to those with refills. There were no major differences for the other drug groups studied.

All selected variables met the inclusion criteria for adjusted logistic regression analyses (table 3). We observed only minor changes in the estimates in the adjusted model compared to the univariate logistic regression models, although we noted some changes for marital status and socio-economic status. In the adjusted model, early discontinuation was more common among those born outside Sweden with two non-Swedish parents [odds ratio (OR) = 1.76] compared to those born in Sweden with Swedish parents, and it was slightly elevated for those born in Sweden with non-Swedish parents (OR = 1.20). Furthermore, early discontinuation was more common among those receiving social assistance (OR = 1.26), but less common among those with ≥2 years of higher education (OR = 0.71). Early discontinuation was more common among those who started treatment with a TCA (OR = 2.58) or anti-depressant other than SSRI, TCA and SNRI/NRI (OR = 2.90) but less common when the first filled prescription was issued by a general psychiatric health care facility (OR = 0.81).

View this table:
Table 3

Results from logistic regression models analysing predictors for early discontinuation in the study population (n = 25 003), presented as crude and adjusted ORs with 95% CIs

OR (95% CI)OR (95% CI)
Women0.86 (0.81–0.91)0.82 (0.75–0.89)
Age (years)
    25–291.03 (0.96–1.11)1.04 (0.93–1.17)
    30–341.06 (0.98–1.13)1.00 (0.88–1.13)
Marital status
    Married1.18 (1.10–1.27)1.03 (0.92–1.15)
    Divorced or widow/widower1.35 (1.19–1.54)1.10 (0.90–1.35)
    Born in Sweden, two parents born in Sweden1.001.00
    Born in Sweden, one parent born in Sweden1.04 (0.94–1.15)1.01 (0.88–1.15)
    Born in Sweden, two parents born outside Sweden1.34 (1.18–1.52)1.20 (1.01–1.43)
    Born outside Sweden, one parent born in Sweden1.27 (0.80–2.01)1.23 (0.66–2.28)
    Born outside Sweden, two parents born outside Sweden2.10 (1.85–2.38)1.76 (1.48–2.10)
Highest attained education
    Mandatory education (0–10 years)1.001.00
    Upper secondary school0.73 (0.68–0.79)0.85 (0.76–0.95)
    Higher education <2 years0.52 (0.46–0.59)0.62 (0.51–0.75)
    Higher education ≥2 years0.57 (0.52–0.62)0.71 (0.61–0.83)
Socio-economic status
    Upper white-collar worker1.001.00
    Lower white-collar worker1.16 (1.06–1.27)1.07 (0.95–1.21)
    Blue- collar worker1.29 (1.18–1.42)1.09 (0.97–1.23)
    Other occupation1.21 (0.86–1.70)1.19 (0.81–1.75)
    Student0.81 (0.75–0.86)0.97 (0.87–1.09)
    Unemployed0.94 (0.87–1.01)1.03 (0.94–1.13)
    Social assistance1.52 (1.42–1.63)1.26 (1.11–1.44)
    Income <60% of median income1.07 (1.00–1.14)1.01 (0.89–1.15)
Type of anti-depressant purchased at index date
    TCA2.60 (2.35–2.86)2.58 (2.24–2.98)
    SNRI/NRI1.11 (1.03–1.20)1.08 (0.96–1.22)
    Other anti-depressant2.37 (1.85–3.05)2.90 (2.05–4.10)
Sphere of activity of health care facility issuing the index prescription
    Primary care1.001.00
    Psychiatry—general0.80 (0.75–0.86)0.81 (0.73–0.89)
    Other psychiatry1.00 (0.81–1.24)0.79 (0.58–1.07)
    Occupational health service0.80 (0.66–0.98)0.90 (0.72–1.13)
    Internal medicine1.59 (1.33–1.88)1.18 (0.94–1.48)
    Neurology1.41 (1.06–1.88)0.99 (0.68–1.44)
    Pain care0.96 (0.51–1.80)0.42 (0.19–0.93)
    Oncology1.80 (0.78–4.16)1.54 (0.53–4.48)
    Other2.64 (2.22–3.13)2.88 (2.33–3.56)
  • a: Including all selected variables in the regression model.

There were no major differences in the results from the multiple regression analyses stratified by age (5-year classes) compared to those for the whole study population. While female sex was associated with early discontinuation in age groups 25–29 and 30–34 years in the unadjusted analyses, these associations were not significant in the multivariate models. In age groups 20–24 and 25–29 years, we observed no difference between Swedes and those born in Sweden to two non-Swedish parents and for recipients of social assistance. Furthermore, we detected no differences between those with upper secondary school education and for those with mandatory education only in the youngest and the oldest age groups and for those with a higher education <2 years in the oldest group. Early discontinuation was more common among blue-collar workers in the oldest age group.

Women, individuals with only mandatory education, recipients of social assistance, blue-collar and lower level white-collar workers were overrepresented in the study population when compared to the reference population (table 4). Further, the proportion of the youngest age group, individuals with a higher education two years or longer and employed individuals was smaller in the study population.

View this table:
Table 4

Characteristics of new anti-depressant users, i.e. study population (n = 25 003), and reference population (n = 500 000)

CharacteristicStudy populationReference populationP-value
nRate per 1000nRate per 1000
    Male9298372259 264519<0.0001
    Female15 701628240 736481
Age (years)
    20–246519261162 153324<0.0001
    25–298310332160 972322
    30–3410 174407176 875354
Marital status
    Unmarried19 262777392 554794<0.0001
    Married430817491 631185
    Divorced or widow/widower12104910 06220
Number of people in the household
    112 356499238 374477<0.0001
    2294311948 28497
    3–59065366203 466408
    6 or more40316943619
    Born in Sweden, two parents born in Sweden17285782355 793812<0.0001
    Born in Sweden, one parent born in Sweden240110937 70186
    Born outside Sweden, at least one parent born in Sweden884438410
    Born in Sweden, two parents born outside Sweden12255520 07746
    Born outside Sweden, two parents born outside Sweden11125020 37947
Country of birth
    Sweden21 085845420 489841<0.0001
    Western industrial countries8943623 12646
    Eastern Europe9663916 83634
    Other20138132 05064
Highest attained education
    Mandatory education (0–10 years)506420844 58393<0.0001
    Upper secondary school11 910489247 430517
    Higher education <2 years19698143 51791
    Higher education ≥2 years5415222143 440300
Activity on labour market
    Employed11 851479276 986554<0.0001
    Student6510263142 768286<0.0001
    Received social assistance463518737 65975<0.0001
    Unemployed445018075 191150<0.0001
Socio-economic status
    Upper white-collar4271239109 045291<0.0001
    Lower white-collar643236198 146262
    Blue-collar6951390162 221433
    Disposable income <60% of the median value5725229125 226251<0.0001
Annual disposable income [hundreds of EUR (SEK 1= EUR 0.1)]
    Median (IR)121.878.4128.199.5
Annual disposable household income (hundreds of EUR)
    Median (IR)177.4197.6213.7229.3
  • Percentages per category were calculated excluding missing values. Pearson's chi-square test was used to analyse differences between the study population and the reference population.


Over one-fourth of new anti-depressant users discontinued treatment early, suggesting suboptimal use. Early discontinuation rates were lower among individuals who started treatment with an SSRI and those who used sedatives and hypnotics concurrently. We detected higher discontinuation rates among individuals born outside Sweden with non-Swedish parents, individuals born in Sweden with non-Swedish parents and among individuals who received social assistance. Early discontinuation was less common among individuals with higher education and among women.

Although some previous studies reported similar early discontinuation rates,2,5,6,8 others reported lower11 or higher7,10,12–15,18,20 rates. However, most of these studies included a broader age span than our study. Disparate results might be attributable to several factors, such as different definitions of early discontinuation, whether prescribing or dispensing data were used, age groups studied, setting (primary care or specialized health care) and whether inclusion in the study required a diagnosis of depression. The variation in early discontinuation rates by anti-depressant type in this study was in accordance with that in previous studies.12,27

Being born outside Sweden with non-Swedish parents was the most strongly associated factor with early discontinuation. Other studies have also reported differences in early discontinuation rates by background. In the Netherlands, being a non-western migrant was the strongest predictor for early discontinuation in adults aged ≥18 years.4 A Danish study also reported that early discontinuation was more common among those with a foreign citizenship.12 Olfson et al.14 reported that early discontinuation differed by ethnic origin, i.e. it was more common among Hispanics than whites in the USA. Thus, although studies have used different definitions to indicate migration background, which might influence results to some degree, all results point in the same direction.

Our finding that early discontinuation was less common among individuals with higher education concurs with previous studies,5,12,14 possibly because more highly educated individuals easier understand information about the onset of effect and the side effects of anti-depressant treatment. A study investigating beliefs about anti-depressants among Australian adults determined that one- fourth of all respondents believed anti-depressants would harm a depressed and suicidal person.28 This group was less educated, had less experience with depression and underestimated both the seriousness of depression and the need for intervention. Taken together, these findings pinpoint the necessity of discussing patients’ beliefs about treatment and disease when initiating anti-depressant treatment.

The present study determined that early discontinuation was more common among individuals who received social assistance. We detected no excess risk for unemployed individuals, for those with an income <60% of the median income or for those who received student allowances in the preceding year. While some previous studies reported no differences by income level, others did.12,14 Milea et al.29 reported higher levels of early discontinuation among individuals enrolled in Medicaid, which can indicate low income.29 In general, income levels are lower and income differences are less pronounced in young adulthood compared with older age groups, possibly explaining the absence of differences by income in our study.

On average, our study population used two co-medications. Earlier studies reported slightly higher numbers but they also included older age groups.4,9 Excluding women's use of oral contraceptives, the largest groups of concurrently used drugs were antibiotics, sedatives and hypnotics. Concurrent use of sedatives and hypnotics in our study agreed with data from studies of anti-depressant users in mixed adult age groups, which are generally older.4,5,9 Concurrent use of sedatives and hypnotics was less common in early discontinuers, possibly indicating fewer contacts with health services or less severity. To avoid sleep disturbances and anxiety, healthcare providers commonly add a sedative or hypnotic when initiating SSRI treatment, possibly explaining why early discontinuation occurred less commonly among patients with concurrent use of these drugs. However, concurrent use as a proxy for severity is very unspecific.

We observed disparities between our study population and the reference population. Women and socially vulnerable groups, i.e., individuals with lower education and those needing financial support were overrepresented among anti-depressant users, indicating the importance of including socio-economic factors when studying anti-depressant use.

Our findings regarding early discontinuation and factors associated with early discontinuation were similar to previous studies.4,5,12,13 Our findings regarding early discontinuation likely can be generalized to other settings with similar safety nets for health care and medication costs. These safety nets are an important aspect that substantially impact access to health care and medicines.

Strengths and limitations

Our study presents data from a national database that encompasses all filled prescriptions in Sweden, independent of reimbursement status. Our data does not include drugs used for in-patient hospital care, and we did not gather any data on hospitalizations. Since depression and anxiety are largely treated in ambulatory care, the majority of patients likely are included in the database. Since some individuals classified in our study as early discontinuers may have been hospitalized after they filled their first prescription, they possibly were misclassified. However, since such individuals probably would need new prescription medicines following discharge and thus would need to fill a second prescription, it is unlikely that hospitalization within the first 6 months altered our results. Importantly, only 4% of all psychiatric inpatient hospitalizations in Sweden had a length of ≥85 days.30 Since earlier studies reported higher persistence among individuals who received their first anti-depressant prescription from a psychiatrist rather than a GP, we tried to describe the treatment setting that issued the index prescription.9 However, register data on treatment setting is scarce, the sphere of activity was missing on some dispensed prescriptions and the register does not include information on prescribers’ specialty.

Some individuals in our study group lacked socio-economic information. Indeed, the most common variable recorded as unknown or missing was parental country of birth. The procedure for recording such data has varied over time and uncertainties can exist for migrating adolescents or adults who arrive in Sweden independently. While this lack reduced the numbers included in the multiple regression models, it likely did not result in overestimated effects.

Information about indication was not available in a format that allowed statistical processing, a significant but common limitation in register-based research. Consequently, we did not know whether an anti-depressant was intended for treatment of depression or another indication. However, prescriptions for anti-depressant treatment of both depression and anxiety disorders should last at least six months. We also lacked information about diagnosis or severity, an important factor in assessing whether individuals with a more severe illness also discontinued treatment earlier.


We show here that more than one fourth of new anti-depressant users discontinue anti-depressant treatment after filling a prescription only once, suggesting suboptimal use in relation to therapeutic guidelines. Early discontinuation occurred more commonly among individuals born outside Sweden with non-Swedish parents and among individuals who received financial support from social services. Taken together with previous studies reporting that early discontinuers tend not to inform their physician, our data indicate that these groups might need greater monitoring during the initiation of anti-depressant treatment. Combined with systematic follow up regarding adherence and treatment experience, prescriber awareness of characteristics associated with increased risk for early discontinuation of anti-depressant treatment, can improve treatment outcomes.


Data collection for this study as supported in part by a post-doctoral scholarship from the Swedish Lundbeck foundation. The study was also financed with internal funds.

Conflicts of interest: None declared.

Key points

  • Among new anti-depressant users aged 20–34 years, more than one fourth discontinued anti-depressant use early.

  • Early discontinuation occurred more commonly among individuals born to non-Swedish parents and among those who received social assistance, but less commonly among women and among individuals with at least two years of higher education.

  • Individuals with foreign background, both first and second generation migrants, and recipients of social assistance might require greater support during initiation of anti-depressant therapy to facilitate initiation and maintenance of a sufficiently long treatment period.


The authors are grateful to Ms. Karen Williams, KWills Editing Services, for help with language editing the manuscript.


View Abstract