Study design

A national register-based cohort study was performed based on data from national registers in Denmark (5.8 million inhabitants). The cohort consisted of inpatients admitted in the period 2011–2017 with major depressive disorder (MDD) as the primary diagnosis. The Danish public healthcare system is mainly tax-funded with universal health coverage to ensure, in principle, free and equal access to hospital care for all Danish residents. Since no private mental care units exist, all patients with MDD are admitted and treated in public hospitals. A unique personal identifier, assigned to all residents, was used to register the use of services as well as the quality of care in national registers [Reference Mainz, Hess and Johnsen11]. This identifier was used to retrieve and merge individual data from the different registers.

Data sources

The Danish Depression Database is a nationwide clinical quality registry [Reference Videbech and Deleuran12]. It was established in 2011 and contains information on the quality of care, admission and discharge dates for all adult patients (≥18 years) with permanent residence in Denmark, admitted at psychiatric hospitals with a primary diagnosis of MDD (International Classification of Diseases [ICD]-10: F32, F33, F34.1, and F06.32 including all subcodes) [Reference Videbech and Deleuran12]. It is mandatory by law for all Danish mental care units to report data to the registry [Reference Videbech and Deleuran12].

Data from the Danish National Patient Registry was used to identify readmissions with a primary diagnosis of MDD, all-cause readmission, and suicide attempts. Danish National Patient Registry includes patients treated at all non-psychiatric hospitals from 1977 and patients from psychiatric hospitals from 1995 and onwards [Reference Schmidt, Schmidt, Sandegaard, Ehrenstein, Pedersen and Sørensen13].

The patients’ vital status was obtained through the Danish Civil Registration System. Established in 1968, it is a national register containing basic personal data of anyone with a social security number in Denmark [Reference Schmidt, Pedersen and Sørensen14]. The causes of death, including suicide, were obtained through the Register of Causes of Death, which has existed in its current form since 1970 [Reference Helweg-Larsen15].

Study population

The study population was identified through the Danish Depression Database and included the first documented admission of adult (≥18 years) inpatients in the period 2011–2017 with MDD as the primary diagnosis (see Figure 1 for study population selection). Stays shorter than 24 h were excluded, since data collection in the Danish Depression Database only includes those who stay longer than 24 h. If multiple hospital contacts occurred within periods of up to 4 days, they were considered to reflect the same admission. Some of these combined admissions contained contacts shorter than 24 h. However, the combined admission was excluded if all contacts were <24 h. In total, 20,750 individual first-time documented admissions between 2011 and 2017 were identified. Among these, subjects were excluded if they had migrated to Denmark within 1 year before (n = 50) or had left the country before the first registered admission in the Danish Depression Database (n = 30). This was done to ensure complete data on the covariates and clinical outcomes. Four subjects lacked all data in the Statistic Denmark registry and were therefore also excluded. Finally, descendants of migrants were excluded since they represent a subgroup with characteristics of both migrants and the native population (n = 154). In Statistics Denmark, descendants are defined as people born in Denmark to parents, neither of whom was born in Denmark and has Danish citizenship. The final study cohort included 20,508 patients. In the analyses of the clinical outcomes, follow-up time was set to 1 year. Thus, patients admitted in 2017 were omitted from these analyses (n = 2,085).

Figure 1.

Flow chart of study population selection.

Exposure

According to Statistics Denmark, migrants are individuals born abroad to parents born abroad who are not Danish citizens [16]. Data from the Danish Depression Database were linked to the Population Register to obtain individual data on country of birth as well as immigration and emigration dates. The population was classified as migrants and non-migrants (Danish-born). Migrants were further subclassified into “Western” or “non-Western” according to their country of birth using the categorization of countries made by Statistics Denmark [16]. Due to the limited number of migrants in the Danish Depression Database, subclassification of the migrants into separate countries was not possible. This meant that heterogeneous groups in terms of migration experience, trauma exposure, socioeconomic status, and culture were grouped together. Asylum seekers who have not yet received a residence permit in Denmark are not provided a unique personal identifier and were therefore not included in the study.

Outcomes

Quality of care

Quality of care was assessed by nine evidence-based performance measures obtained from the Danish Depression Database reflecting recommendations from national clinical guidelines [Reference Videbech and Deleuran12] (see Table 1).

Table 1.

Performance measures of quality of care in the Danish Depression Database for inpatients.

Abbreviation: HAM-D17, Hamilton depression scale (17-item version).

These measures were selected by a national multidisciplinary expert group appointed by scientific societies and professional organizations [Reference Mainz and Bartels17]. All measures reflected clinical services, which should be provided to all patients regardless of their individual characteristics, severity of disease, and so forth. However, some of the measures did not need to be performed if it could be demonstrated that they had recently been performed and documented elsewhere in the healthcare system. Others required a certain minimum length of stay to be relevant. Based on these criteria, the patients were individually classified as eligible or ineligible for the individual performance measures.

The quality of care was assessed both using the individual performance measure as well as by using two opportunity-based composite measures. Using the patient average method [Reference Reeves, Campbell, Adams, Shekelle, Kontopantelis and Roland18], one composite measure was constructed as a continuous variable of the total percentage of fulfilled eligible measures per individual. In addition, a dichotomous composite indicator was created with a cut-off of 70% or more of the eligible performance measures fulfilled as a measure of high quality of care. This cut-off is often used in analyses of composite indicators [Reference Reeves, Campbell, Adams, Shekelle, Kontopantelis and Roland18] and was a pragmatic estimate in which most of the care processes were fulfilled while still providing enough patients with the outcome to conduct meaningful analyses [Reference Rasmussen, Videbech, Mainz and Johnsen19]. Additional cut-offs of 60 and 80% were applied in sensitivity analyses.

Covariates

Relevant covariates were identified a priori using directed acyclic graphs and included age as a continuous variable and sex [Reference Shrier and Platt20]. Models adjusting for these covariates constituted the primary analyses. However, socio-economic factors (SEF) could be considered potential confounding factors [Reference Knudsen, Valentin, Videbech, Mainz and Johnsen21], although it is debated whether it is more correct to consider them as intermediate factors [Reference Ingleby22]. To explore the role of SEF, educational level, income, occupational status, and residency were included as covariates in a sensitivity analysis. Educational level was categorized in accordance with The International Standard Classification of Education (ISCED). Income was calculated as the average yearly total family income in the 5 years before admission. Employment status was classified into four categories: employed, public benefits, pensions, and students. Residency was registered based on home address and categorized in accordance with residency in one of the five administrative regions in Denmark.

Statistical analyses

Analysis was first performed with all migrants compared with the Danish-born population and then performed with the two migrant subgroups (non-Western and Western) compared with the Danish-born population. To account for missing data, multiple imputation using chained equations was applied with missing data imputed 10 times using available patient characteristics including exposure and outcomes. Patients with missing exposure were excluded from the respective analysis.

The dichotomous composite score as well as the individual performance measures were examined using Poisson regression models with robust error variances, reporting relative risk (RR) with the corresponding 95% CI. The continuous composite score was examined using multivariable linear regression models, reporting percentage point difference (PPD) with corresponding 95% confidence intervals (CI).

The associations between migration status and clinical endpoints were determined using Cox proportional hazards regression analysis to compute hazard rate ratios (HR) for all-cause mortality and cause-specific HR (csHR) for suicidal behavior and readmissions presented with 95% CI. Administrative censoring at 365 days follow-up was applied. Unadjusted event proportions and adjusted risk differences (RDs) for the clinical endpoints were calculated from Aalen–Johansen cumulative incidence using bootstrap and death as a competing risk. Inverse probability of treatment weights was applied for adjusted analyses.

Analyses were performed using STATA v.16 (StataCorp LLC, College Station, TX) [23].