Inequities in breast cancer screening utilisation in Spain - Using decision trees to identify intersections
More details
Hide details
University of Bremen, Institute for Public Health and Nursing Research, University of Bremen; 2 Leibniz ScienceCampus Digital Public Health Bremen, Grazer Str. 4, 28359 Bremen, Germany
University of Bremen, Germany
Publication date: 2023-04-26
Popul. Med. 2023;5(Supplement):A1476
Background and Objective:
Organised breast cancer screening programmes can be effective in reducing incidence, mortality and illness burden. Currently in Spain, women between 50-69 are invited to attend screenings bi-annually. However, roughly 25% do not utilise this free service. Here, we take an intersectional perspective using machine learning techniques to identify social groups at risk of not utilising breast cancer screening.

Women were drawn from the 2020 European Health interview Survey in Spain, which targets the (young) adult population > 15 years old living in private households (N = 22,072; 59% response rate). Using available indicators of socioeconomic status based on the PROGRESS-Plus framework, we applied machine learning (Classification and Regression Trees (CART), Chi-square Automatic Interaction Detector (CHAID), and Conditional Inference Trees (CIT)) to data from the target population (women 50-69) to estimate models that disentangle existing social intersections. We then used accuracy, sensitivity and specificity indicators to identify the best-fitting tree.

A non-parametric CHAID model suggests (overall accuracy of 75.07%) primary education or below to be the strongest discriminating factor for not attending screening (n=1,060 Pr=0.3448 vs n=2,790 Pr=0.2462). The second strongest factor was country of birth (lower education group, born in Spain: n=983 Pr=0.3313 vs born outside Spain n=77 Pr=0.5195; higher education group, born in Spain: n=2,596 Pr=0.2369 vs born outside Spain: n=194 Pr=0.3711). Finally, for those born in Spain with lower education, being married, widowed or divorced predicted attendance compared to being single or separated (n=869 Pr=0.3026 vs n=124 Pr=0.5081).

In order to reduce inequities in screening attendance, programs in Spain should particularly support women with lower education and migration background, and potentially focus on social support measures. CHAID is a useful tool to identify groups at higher risk of not utilising an organised public health program and inform tailored prevention programs.