The first screening round took place from 28 February 2008 to 31 December 2009 where women were offered a pre-booked mammogram appointment at one of the region’s six screening sites. Appointment and site could be changed by the women who could also decline to participate. Non-participants received no reminders. A population-based study was performed including all women (aged 50–69) invited to the first screening round in the Central Denmark Region (n = 149,234).
According to Danish Legislation and the Central Denmark Region Committees on Biomedical Research Ethics (j.no.: 181/2011) the study did not need formal ethical approval, as it was based on registry data. The project was approved by the Danish Data Protection Agency (j.no.: 2009-41-3471 and j.no.: 1-16-02-31-11).
Registries and variables
Data on participation was collected from a regional database containing administrative information. Participation was categorised as ‘participants’ or ‘non-participants’ based on whether the women participated in the first screening round or not. Non-participants were further divided into ’active non-participants’ (ANPs), defined as women who actively called and declined participation, and ‘passive non-participants’ (PNPs), defined as women who stayed away without cancelling or rescheduling the appointment (‘no-shows’).
Data on socio-demographic variables was obtained from ‘The Danish Integrated Database for Labour Market Research’ (IDA) run by Statistics Denmark
, which has been updated annually since 1980. The following variables from IDA were included: Educational level classified according to UNESCO classification as low (≤10 years), middle (11–15 years), and higher education (>15 years). Labour positions were classified as: 1. employed, 2. self-employed and chief executive, 3. unemployed or receiving supplementary benefits other than social welfare, 4. retired women, 5. social welfare recipients, and 6. others. Marital status was classified as married, living in a registered partnership, cohabitating, or being single. Ethnicity was categorised as Danish, immigrants from western countries, and immigrants from non-western countries, according to Statistics Denmark’s definition of developed countries
. Residential ownership was divided into residence-owners and tenants. ‘OECD-adjusted household income’ in the year prior to mammography, adjusted for number of persons in the household
, was used as an income measure. Based on tertiles and rounded off to the nearest 100 Euros, OECD-adjusted household income was categorised as: low (1st tertile, ≤34,600 Euros ), middle (2nd tertile, >34,600 to ≤53,200 Euros ), and high (3rd tertile, >53,200 Euros). From the National Vehicle Registry, information on the household’s access to a vehicle was obtained and divided into access or no access. From The Danish Cancer Registry
, all registered cancer diagnoses prior to the screening date were obtained, and all women with a prior breast cancer diagnosis were identified and excluded. The remaining women were divided into either ‘no registrations of a cancer diagnosis’ or ‘registered with one or more cancer diagnoses (excluding breast cancer) prior to the screening date’. Women diagnosed with cancer prior to the age of 14 were categorized as ‘no registered cancers’ (n = 62).
Travel distance to screening site was calculated according to the Danish road network using ArcGIS Network Analyst (version 10.0)
. Geographical coordinates were used to locate each woman’s residence on the date of her screening appointment and the coordinates were obtained from the Centralised Civil Register. It was possible to calculate the shortest driving route in km for 137,417 of the women (95.3%). Travel distance was categorised into: 1) 0–20 km, 2) >20-40 km, 3) >40-60 km, and 4) >60 km.
In Denmark 98% of citizens are listed with a general practice, which they must contact for medical advice
. From the ‘Patient Registry’, data on which general practice the woman was listed with at the screening date was obtained.
All registries could be linked by the women’s unique civil registration number (CRN). A woman’s socio-demographic position was described the year prior to the booking date. This means that for a women being assigned to screening in e.g. 2009, socio-demographic registries updated at the end of 2008 were used. Missing information on the registry-based variables ranged from 0% for the variables ‘age’ and ‘participation’ to 1.8% for data on ‘residential ownership’.
Generalised linear models (GLM) with log link and the Bernoulli family regression models
[26, 27] were used to quantify the association between socio-demographic factors and screening participation. Restricting the analyses to the group of non-participants (n=30,453), the same method was applied to investigate whether ANPs and PNPs differed in terms of socio-demographic factors.
Initially, to check for multicollinearity between independent variables the mean Variance Inflation Factor (VIF) was calculated. Values above 10 indicate multicollinarity
. Following, unadjusted analyses were done with each independent variable. A multiadjusted model was also done, adjusting for statistically significant variables (p <0.01) from the unadjusted analysis. Prevalence ratios (PR) with 95% confidence intervals (CI) were used as association measures using robust variance estimates to allow for clustering of patients by general practice in both unadjusted and adjusted models
. Due to a high prevalence of the outcome measure (more than 20% non-participants), odds ratios tend to overestimate the association
[26, 27]. Age was calculated on the date of the screening appointment using the women’s CRN containing date and year of birth. Statistical analyses were conducted using STATA version 12.