We conducted an observational population-based cohort study using Danish nationwide register data. All IBD (CD or UC) incidences in the Danish adult population with an initial hospital contact between January 1, 2000, and December 31, 2017 were included. The diagnosis (index) date was the day of the first hospital contact after 2000 with a diagnostic code for IBD. According to our inclusion criteria, incident cases were patients with at least two hospital contacts with IBD diagnosis (ICD-10 K50 & K51) within five years at a Danish hospital. To ensure the correct identification of new incidences, we checked the register for IBD prevalence in the period 1995–1999. Fig. 1 depicts the inclusion process of the studied patients.

Flow chart of the patient population inclusion process

Cases were followed up until the onset of biologic treatment. Patients without any registration of biologic treatment were censored at the date of IBD-related surgery (procedure codes starting with “KJFK”/“KJFB”/“KJFA58”/“KJFA38”/“KJFA6”/“KJFH”/“KJFF”), emigration, death, or at the end of the observation period (31/12/2018).

The study was carried out in accordance with the Strengthening the Reporting of Observational Studies in Epidemiology (STROBE) guidelines [9].

Each official resident in Denmark is assigned a unique personal identification number, which in its encrypted form is used to link person data from different registers. The Danish national population registers are of high quality and comprehensiveness and offer the opportunity to study complete population samples using systematically collected long-term data [10]. Demographic patient information was retrieved from the Civil Registration System [11]. Data on income, education and occupation were retrieved from the Income Statistics Register [12], the Student Register [13] and the Employment Classification Module [14], respectively. The National Patient Register (NPR) provided medical data on comorbidity, disease management activities and hospital contacts [15].

The main outcome is indicated as the number of days between initial IBD diagnosis and the date of the first received dose of a Tumor Necrosis Factor‑α inhibitor (TNF-i) (treatment code: BOHJ18A1/BOHJ18A3/BOHJ18A4), an anti-integrin (BOHJ19H4) or anti-interleukin (BOHJ18B3) as recorded in the NPR.

The socioeconomic indicators included measures of education, income and occupation. They were handled as separate characteristics to explore each factor’s individual association with the outcome.

Educational level: The highest completed education before the index date was classified into four categories (lower secondary education or less, upper secondary education, vocational education and academic education) based on the 2011 International Standard Classification of Education [16].

Equivalent disposable household income: Baseline income was specified by the annual disposable amount of money available to the subjects based on the total household income relative to the number of persons living in the respective household. Data from the year preceding the index date was used toallocate the patients. Thereby, we made sure that the IBD diagnosis and disease course would not interact with the income subsequently. Baseline income was divided into quartiles.

Occupational status: Occupational status in the year preceding the IBD diagnosis as according to the Employment Classification Module described the main source of income in six categories: Employed/self-employed, unemployed including receivers of governmental financial support (social aid), student, retired, sick leave and other in case the main income source was registered as not identifiable.

The analyses were adjusted for age, sex, presence of comorbidities and index year, as the availability of biologic drugs increased substantially after 2004. Prevalence of comorbidity was identified via ICD-10 codes as registered during hospital contact. Diseases considered most relevant in the uptake of biologic treatment were included. These comprised hypertension, cerebrovascular disease, psychiatric disorders, rheumatoid arthritis, other inflammatory diseases (psoriasis, multiple sclerosis, ankylosing spondylitis), tuberculosis and hepatitis B/C [17,18,19]. Comorbidities leading to hospital contact within one year before the IBD diagnosis were considered clinically relevant for the analysis.

Cox proportional hazards regression models were applied to analyse the association between SES and first biologic treatment. We estimated hazard ratios (HRs) with corresponding 95% confidence intervals (CIs) and a significance level of α = 5%. The models were grouped by IBD type to map disease type-specific associations in the analyses.

The proportional hazards assumption was assessed graphically using Schoenfeld residuals. Consequently, income was included with a time-varying coefficient. Martingale residuals were used to evaluate the model fit regarding the linearity of covariates. We did not adjust the p-values for multiple testing, as each analysis was conducted individually. The E‑value was computed for estimates as a sensitivity analysis of the robustness of the models where the 95%-CI excluded the null [20]. Missing data due to unavailable information from registers were few (< 3%) and mean imputation was conducted. All analyses were conducted in Stata version 17.