Data sources
This study is a dynamic national register-based prospective cohort study. We used data from the total resident population of women in Norway. A unique de-identifiable version of a personal identification number (PIN), was used to combine the information from four national registers. PIN is assigned to all Norwegian citizens at birth, as well as to all individuals registered as residents in Norway for at least six months. The Central Population Registry provided demographic information such as birth year, country of origin, migrant background and marital status. The KUHR database contains information on compensation claims from health professionals such as those working in OPMH services. From this we were able to identify individuals with OPMH consultations during the years 2006-2013. Information on the highest obtained education level was extracted from the National Education Database, while the FD-Trygd database provided information on work-related income and child benefits for the years 2006-2013.
Study design and population
We included women aged 23-40 years, born between 1968 and 1988, who resided in Norway for at least three consecutive years during the study period 2006-2013. This age group was chosen because by age 23, the majority of women have completed their education and are entering the labour force. A disruption in this process, such as the onset of a mental disorder, can have both short-term and long-term effects.
All women were followed from the start of the study in 2006, when turning 23 or from the year of migration to Norway. Censoring occurred at the end of the study in 2013, when turning 40 years or the year of emigration or death. Furthermore, we excluded women who had no work-related income throughout the study period (N= 33,755), as we were interested in the change in work-related income. The eligible sample consisted of 640,527 women.
Measures
The outcome variable was work-related income, defined as yearly pre-tax wages and income from self-employment for the years 2006-2013. Income for all years was inflated to 2013 levels by using the Norwegian consumer price index. Negative values were recoded to 0 and values higher than 2,000,000 Norwegian kroners (NOK) were set to 2,000,000 NOK. Furthermore, to ease the interpretation of the results, income was recalculated into percentiles and presented as percentile change in income following the uptake of OPMH treatment.
The main exposure variable was OPMH treatment, used as a proxy for mental disorder. Since the first consultation is mainly used to map whether there is a need for further OPMH follow-up, to be exposed, a woman needed to have at least two consultations within a six-month period. In cases where the two contacts occurred in two consecutive calendar years, the year of exposure was set as the year of first contact with OPMH services. Furthermore, to detect changes in income following the uptake of OPMH treatment, we introduced a washout period where each woman defined as exposed had to have at least two years free of OPMH consultations prior to exposure. This was to increase the probability that the contact detected between 2008 and 2013 was a new, and not an ongoing, case.
Additionally, we considered differences in loss of work-related income following the uptake of OPMH treatment by migrant background. Migrant background was divided into three major groups: majority (Norwegian-born with at least one Norwegian-born parent), descendant (Norwegian-born with two foreign-born parents) and migrant (foreign-born with two foreign-born parents). We divided migrants into eight regions of origin: 1) Nordics, 2) Western Europe, 3) EU Eastern Europe, 4) non-EU Eastern Europe, 5) Middle East and North Africa (MENA), 6) Sub-Saharan Africa, 7) South Asia, and 8) East/ South East Asia. Women from countries not fitting into the categories presented above (N=9220) were excluded from the study sample. The Norwegian majority group was used as a reference group in all analyses.
Covariates included age, education level, marital status and motherhood, all time variant. Age was measured as a continuous variable with values between 23 and 40. Education level was a categorical variable with categories 1) compulsory or lower, 2) upper-secondary, 3) tertiary, and 4) unknown. Marital status was also a categorical variable with values 1) unmarried, 2) married/partner, and 3) previously married. Motherhood (yes/no) was based on whether the woman was receiving child benefit or not, commonly entitled to mothers with a dependent child below the age of 18 [31]. We also controlled for the year of observation to adjust for time fixed effects, with a dummy variable for each year.
Statistical Analysis
To investigate the effect of OPMH treatment on work-related income, we applied a linear regression model with individual fixed effects. Use of fixed effects models provides a method of assessing the association between exposure (OPMH treatment) and outcome (income), adjusting for all measured time-varying and both measured and unmeasured time-invariant factors within an individual, eliminating confounding from such factors [32]. Each woman is treated as her own control [33]. A disadvantage of using a fixed effects model is that coefficients for time-invariant variables, such as for migrant background, are not estimated [32]. However, by including an interaction term between OPMH treatment (time-variant) and migrant background (time-invariant), the fixed effects models provide coefficients for the interaction between these two variables. Due to a large difference in income between the investigated migrant groups, the analysis was stratified into low- (the bottom third of the income distribution) and high- (the top two thirds of the income distribution) income levels, based on mean work-related income during the years in the study. This distribution divide was chosen to ensure there was a sufficient number of individuals in each group. The regression analyses had a hierarchical set-up with adjustment for age and an interaction term between OPMH treatment and migrant background in Model 1 and educational level (coefficient for the individuals with unknown education were omitted from the analysis), marital status, motherhood and time fixed effects in Model 2. Results for time fixed effects are not shown. Stata 15.0 was used to perform all analyses.
Ethics
Approval for the use and linkage of data, and the conduction of the study was granted by the Regional Committee for Medical and Health Research Ethics, South East Norway (REK 2014/1970) and the owners of the different registries. Since this study uses already existing administrative data consent to participate was not required.