The research team, led by Sarah Derveeuw and Eva Gezels of Ghent University’s Department of Public Health and Primary Care, together with colleagues at the Centre for Cancer Detection in Bruges, took advantage of a rare scientific resource: individually linked population registry, cancer registry, health insurance and administrative socioeconomic records. Rather than surveying a sample and hoping it represents the whole, the investigators assembled a retrospective cohort of essentially every eligible woman aged 31 to 64 residing in Flanders between 2018 and 2023. This population-wide design eliminates the selection biases that plague smaller studies and allows researchers to detect patterns in groups that are often too small or too hard to reach for conventional surveys, including women who never interact with the health system at all.

The study’s central outcome was deliberately strict. A woman was classified as a long-term non-attender if she had no registered cervical screening test of any kind during the entire six-year observation window, a period spanning more than two full screening rounds under Flemish guidelines. Among the 1,220,227 women with complete data, 21.5 percent fell into this category. That figure is striking on its own, but the true analytical power of the study lies in how the researchers dissected it. They organised their analysis around Andersen’s Behavioural Model of Health Services Use, a long-standing framework that sorts the drivers of healthcare utilisation into predisposing factors such as age, migration background, household type, education and occupation; enabling factors such as household income, subsidised insurance status, recent childbirth and contact with a general practitioner; and illness-level factors such as chronic disease and major invalidity.

Using hierarchical multivariable logistic regression, the team estimated the odds of long-term non-attendance associated with each factor while adjusting for all the others, a statistical technique that isolates the independent contribution of every variable. The results delivered an unambiguous headline: nothing predicts screening behaviour better than whether a woman is already plugged into the healthcare system. Women who had recently given birth had 63 percent lower odds of long-term non-attendance compared with other women, with an odds ratio of 0.37, almost certainly because pregnancy and childbirth funnel them through routine medical contacts where screening can be offered or performed. Frequent contact with a general practitioner produced a similar protective pattern, with an odds ratio of 0.54. In other words, every additional touchpoint with primary care appears to act as an implicit gateway to cancer prevention, and women who lack those touchpoints are systematically left behind.

Education emerged as the most powerful socioeconomic determinant, and it did so in a classic gradient. Women with the lowest educational level had 79 percent higher odds of never attending screening across six years compared with the highest-educated group, an odds ratio of 1.79 that remained essentially unchanged after full statistical adjustment. The persistence of this gradient is important because it suggests that education captures something more fundamental than income alone, perhaps health literacy, familiarity with preventive medicine, the ability to navigate administrative systems, or the cultural capital to feel entitled to ask for preventive care. Intriguingly, household income told a much weaker story, showing only a modest and non-graded association with non-attendance. This dissociation between income and education is a technical finding with real policy consequences: simply making screening free or cheap, which Belgium already does, does not neutralise the social stratification of participation.

The study also produced one of its most nuanced results in the domain of migration background. Among first-generation migrant women, those born outside Belgium, elevated odds of non-attendance were largely attenuated once socioeconomic factors were taken into account, implying that their lower participation is substantially explained by the socioeconomic positions they occupy rather than by their origin per se. But the same was not true for the second generation, women born in Belgium to migrant parents. Elevated odds persisted among second-generation women of Maghrebi, Turkish, Asian, Sub-Saharan African and Eastern European origin even after adjustment for education, occupation, income and healthcare contact. That residual excess risk, surviving a battery of statistical controls, points toward mechanisms that administrative data cannot fully capture, including experiences of discrimination, cultural norms around intimate examinations, language and trust, and the specific ways healthcare services are perceived and navigated by ethno-racially minoritised communities.

Several other groups carried independently elevated odds of long-term non-attendance. Older women within the eligible age range were less likely to attend than younger ones, a pattern with direct biological relevance because the risk of cervical cancer and the persistence of high-risk human papillomavirus infections both increase with age. Women living alone had higher odds than those in larger households, consistent with the idea that partners and family members often prompt and accompany preventive care. Receiving an increased reimbursement, the Belgian marker of low income and financial vulnerability, was associated with non-attendance, as were chronic illness and major invalidity status. The latter finding may seem counterintuitive, since chronically ill patients see doctors frequently, but it likely reflects a clinical reality in which care becomes dominated by the presenting condition and preventive services fall by the wayside, a phenomenon researchers describe as preventive care crowding-out.

What makes this study methodologically significant is not only its scale but its data architecture. Linking the national population registry to the Belgian cancer registry, health insurance claims and administrative socioeconomic records allowed the researchers to construct a near-complete picture of each woman’s social circumstances and healthcare behaviour without asking anyone a single question. The authors handled missing data transparently, reporting characteristics for the full population of 1,283,569 women before restricting to complete cases, and they formally assessed multicollinearity among predictors using generalised variance inflation factors, ensuring that the large odds ratios they report are not statistical artefacts of correlated variables. The study was conducted under ethical approval from the Belgian Information Security Committee using pseudonymised data in compliance with the European General Data Protection Regulation.

The implications reach well beyond Flanders. Organised screening programmes, in which women are invited by name at regular intervals, were designed precisely to flatten social gradients in participation, and the Flemish data show that such programmes reduce but do not erase them. The authors argue that equity-oriented strategies should prioritise the women who remain systematically unreached: those with few socioeconomic resources, limited engagement with primary care, and second-generation migrant backgrounds. Concretely, that could mean embedding screening invitations in maternity and postnatal care, training general practitioners to opportunistically offer cytology or human papillomavirus testing during unrelated consultations, deploying self-sampling kits that women can use at home, and designing outreach that is culturally attuned rather than merely translated.

The study also carries a warning for the HPV vaccination era. As vaccinated cohorts age into screening programmes, maintaining high participation among the socially disadvantaged becomes even more critical, because unvaccinated women are not evenly distributed across society and tend to cluster in the very groups least likely to attend. The researchers are careful to note that their work identifies who is left behind, not why, and they call for further research into the specific mechanisms and barriers underlying these inequities to guide tailored programme adaptations. But the population-level message is already clear: a screening programme is only as equitable as its reach, and in Flanders, as in much of the world, that reach still stops short of the women who need it most.

Subject of Research: Sociodemographic and healthcare-access determinants of long-term non-attendance in cervical cancer screening in Flanders, Belgium

Article Title: Determinants of long-term non-attendance in cervical cancer screening in Flanders, Belgium: a retrospective population-wide cohort study using administrative data

Article References: Derveeuw, S., Gezels, E., Hendrickx, M., Kellen, E., Toma, S., Vanthomme, K., & Willems, S. (2026). Determinants of long-term non-attendance in cervical cancer screening in Flanders, Belgium: a retrospective population-wide cohort study using administrative data. International Journal for Equity in Health. https://doi.org/10.1186/s12939-026-03056-x

Image Credits: AI Generated

DOI: 10.1186/s12939-026-03056-x

Keywords: cervical cancer screening, health inequities, social determinants of health, Flanders, Belgium, population-based cohort, Andersen behavioural model, migration background, primary care, HPV, cancer prevention, public health