Machine learning reveals systemic barriers blocking breast cancer screening for India's poorest women

Breast cancer is now the leading cause of cancer death among Indian women, yet fewer than one in a hundred women in the screening age group have ever been tested for it. The gap between what national health policy promises and what actually reaches women on the ground is not a mystery—it is measurable, predictable, and deeply rooted in who has money, where they live, and how much say they have in their own medical care.

A new study published in Frontiers in Artificial Intelligence applied machine learning to nationally representative health data to map exactly where screening happens in India and, more importantly, where it does not. The researchers used decision trees, random forests, and gradient boosting algorithms to sift through the patterns, asking: which women get screened, and which ones are systematically left out? The answer is stark. Screening is concentrated almost entirely among urban women with higher education and household wealth. Rural women, poor women, less educated women, and those with little say in their own healthcare decisions face overlapping obstacles that make screening nearly impossible, regardless of whether clinics exist nearby.

Geography is one barrier. Rural women live farther from screening facilities, lack affordable transport, and often cannot travel alone without permission or accompaniment. But distance is only part of the story. The machine learning analysis revealed that education ranks among the strongest predictors of who gets screened. Women with more schooling are significantly more likely to seek screening—partly because they understand health risks better, but also because education correlates with greater independence and access to information. Household wealth follows close behind. A woman without money cannot pay for screening, cannot afford to lose a day's wages to travel for it, and cannot navigate a healthcare system that assumes a baseline of resources she does not have.

Two other factors emerged as critical. Contact with community health workers—the frontline staff who work in villages and neighborhoods—substantially increased the odds of screening. These workers serve as a bridge between the formal healthcare system and communities that might otherwise never reach it. And women's autonomy in healthcare decision-making mattered enormously. A woman who could seek medical care on her own, without needing permission or a male relative to accompany her, was far more likely to be screened than one constrained by gender norms or family control. The machine learning models showed that these barriers do not operate independently. A poor rural woman with limited education and low autonomy does not face three separate obstacles—she faces a compounded wall of exclusion that makes screening statistically unlikely.

The researchers used inequality decomposition methods to show how these factors reinforce each other. Economic disadvantage, educational gaps, geographic isolation, and gender inequality do not simply add up; they multiply. A woman missing one of these advantages might still find a way to screening. A woman missing all of them almost certainly will not.

The study's implications are clear: expanding clinics and issuing guidelines will not close this gap. Policymakers need targeted interventions designed for the women most likely to be left behind. Strengthening community health workers—giving them resources, training, and authority to educate and refer women in their own neighborhoods—could reach populations that formal healthcare systems have failed to touch. Health communication must be redesigned for low-literacy and rural audiences, using trusted local channels and culturally appropriate messaging rather than top-down campaigns. Policies that support women's independent access to healthcare—safe transport, women-friendly clinics, the ability to seek care without permission—could measurably shift screening rates. And policymakers should use data-driven approaches to identify high-risk groups and allocate resources where they will have the most impact, rather than relying on broad national averages that mask deep inequality.

The machine learning analysis does something important: it moves beyond asking whether women choose screening and instead asks what systems prevent them from choosing it. The answer is not individual behavior or awareness. It is the architecture of inequality itself—economic, social, geographic, and gendered—that determines who lives long enough to catch cancer early and who does not.