The setting explains why geography looms so large. BMC serves as the only tertiary referral hospital for a vast region of the Lake Zone, providing cancer services to more than twenty million residents. Its catchment stretches roughly 200 kilometers to the south, north, and west, and the nearest alternative tertiary hospital lies more than 1,000 kilometers away. Tanzania’s health system is built as a referral cascade: patients are meant to start at local district facilities, move up to regional referral hospitals, and finally reach zonal or national institutions where cancer diagnostics and comprehensive treatment are concentrated. Although about 90 percent of Tanzania’s population lives within five kilometers of some health facility, specialized oncology services remain concentrated in urban centers, and residents of sparsely populated areas can face travel times of three to six hours for higher levels of care.

To quantify these patterns, the team conducted a cross-sectional ecological geospatial analysis across 45 administrative subdistricts in the BMC catchment area. Each patient’s residence was geocoded to a subdistrict defined by publicly available boundary data from the United States Census Bureau. One island subdistrict, Ukerewe, was excluded because it shares no borders with neighboring areas, and subdistricts containing municipalities were merged to match how medical records recorded addresses, yielding the final 45-unit analytic dataset. The researchers calculated age-standardized observed breast cancer rates per 100,000 women aged 20 and older for each subdistrict, standardizing to Tanzania’s overall population structure. They then applied two classic spatial statistics: Global Moran’s I, which tests whether values across a map are more clustered than chance would predict, and Anselin Local Moran’s I, known as LISA, which pinpoints where specific clusters and outliers sit.

The cohort itself revealed the steep hurdles patients already overcome. The 379 women ranged in age from 21 to 90, with a median age of 50 years. They had traveled a median distance of roughly 150 kilometers to reach BMC, with interquartile distances spanning 18 to 274 kilometers. Just under half, 45.9 percent, held private insurance. Only 52.2 percent completed pathological staging, and among those who did, about 65 percent were diagnosed with late-stage disease, stage III or IV. Nearly 48 percent of the women were lost to follow-up after diagnosis, a striking signal of how difficult continuity of care remains in this setting. At the community level, 52.2 percent of patients lived in a subdistrict with a regional hospital and 67.3 percent lived in one with a district hospital.

The spatial statistics delivered an unambiguous verdict: diagnosis rates were geographically dependent, not random. Global Moran’s I reached 0.289 with a p-value of 0.001, confirming significant clustering across the entire catchment. Cases came from 32 of the 45 subdistricts, with a median of three cases per subdistrict and an overall observed rate of 4.2 per 100,000 women. LISA analysis then located the clusters precisely. High-high clusters, areas of elevated rates surrounded by similarly elevated neighbors, appeared closest to BMC in Ilemela, Magu, and Nyamagana, with rates of 67.8, 18.0, and 47.7 per 100,000 respectively. Low-low clusters, persistently low rates surrounded by low-rate neighbors, emerged in Biharamulo at 2.7 per 100,000 and Bukombe at 4.7 per 100,000. A single high-low outlier appeared in Kibondo, which recorded 13.6 per 100,000 despite its low-rate surroundings.

Comparing community characteristics across cluster types sharpened the interpretation. High-high clusters sat in areas with a mean population density of 840 persons per square kilometer and a mean distance to BMC of just 27.7 kilometers. Low-low clusters, by contrast, lay a mean of 260 kilometers from the hospital, had a mean population density of only 26.8 persons per square kilometer, and lacked regional hospitals entirely. Travel distance and population density both differed significantly across cluster types, with p-values below 0.01 and 0.03 respectively. When the researchers examined patient-level characteristics such as age, insurance status, pathology completion, and loss to follow-up by cluster type, they found no significant differences, suggesting that the geographic patterning of diagnosis operates at the community level rather than reflecting differences in the patients themselves.

Multivariable ordinary least squares regression translated these patterns into effect sizes. For every additional kilometer separating a subdistrict’s centroid from BMC, the observed breast cancer rate fell by 0.04 per 100,000, an association that held in adjusted models with a p-value below 0.001. Higher population density, conversely, was associated with higher observed rates, with a coefficient of 0.008 and a p-value of 0.003, indicating that densely settled areas detect more cases. Quadratic terms tested for nonlinear relationships proved nonsignificant. Notably, the mere presence of a regional or district hospital was not significantly associated with observed rates in the adjusted analysis, implying that facility counts alone fail to capture whether communities actually have functional diagnostic capacity. After adjusting for distance and density, residual spatial autocorrelation disappeared, with a Moran’s I of 0.116 and a p-value of 0.104, meaning these two covariates accounted for the clustering the maps revealed.

The magnitude of the gap between observed diagnoses and expected disease burden is the study’s most sobering implication. The overall observed rate of 4.2 per 100,000 women sits far below Tanzania’s estimated national breast cancer incidence of 35 to 40 per 100,000 reported by the Global Cancer Observatory for 2022. The authors attribute this difference not to genuinely less disease but to women who never reach diagnostic services and remain undiagnosed. Elevated rates near BMC likely reflect better diagnostic availability, while low-rate clusters in remote areas may signal limited detection rather than lower disease burden. This interpretation aligns with broader evidence from sub-Saharan Africa, where greater travel distance to breast cancer facilities has been linked with delayed diagnosis and advanced-stage presentation, and where rural residence is associated with poorer outcomes such as more advanced tumors at diagnosis.

The findings also fit into a wider structural picture. Many hospitals across sub-Saharan Africa provide only fragmented components of breast cancer care without full diagnostic capacity, forcing patients into prolonged pathways of repeated visits and delayed referrals before receiving a formal diagnosis. Across Tanzania, Nigeria, Ghana, Uganda, and Kenya, transportation costs and financial burdens have been shown to delay biopsies and undermine women’s ability to start or continue treatment. Pathology staffing shortages, equipment failures, and limited immunohistochemistry constrain health systems, while drug stockouts and centralized oncology services keep essential therapies far from where patients live. Against this backdrop, breast cancer incidence in many sub-Saharan African countries is projected to rise by 85.7 percent by 2040, while five-year survival in the region remains near 40 percent, making equitable early detection an increasingly urgent goal.

The authors caution that the analysis has real limitations rooted in routinely collected data. Observed rates reflect diagnoses actually made, not population-based incidence, and are shaped by healthcare utilization and diagnostic availability. Thirteen subdistricts recorded no cases at all, an absence that should not be read as an absence of breast cancer but may reflect differences in referral, access, or case ascertainment. Medical records did not capture patients’ initial points of care or referral pathways before arriving at BMC, and subdistrict-level data on household income and education were unavailable, limiting the ability to adjust for socioeconomic factors. Because exact patient addresses were missing, distances were estimated from subdistrict centroids, which may have introduced measurement error and attenuated observed associations. Even so, the study points toward a concrete strategy: decentralizing diagnostic capacity, particularly into the low-rate spatial clusters, could shorten the road to diagnosis for women facing the steepest geographic barriers and improve equity in early detection and outcomes across northern Tanzania.

Subject of Research: Geospatial patterns of breast cancer diagnosis and access to comprehensive cancer services in northern Tanzania

Article Title: Characterizing geospatial access to comprehensive breast cancer services in Northern Tanzania

Article References: Friebel-Klingner, T. M., Rositch, A. F., Iyer, H. S., Wagner, R. B., Joo, E., Kirahi, M., Masalu, N., & Kohler, R. E. (2026). Characterizing geospatial access to comprehensive breast cancer services in Northern Tanzania. Frontiers in Oncology, 16, Article 1952705. https://doi.org/10.3389/fonc.2026.1952705

Image Credits: AI Generated

DOI: 10.3389/fonc.2026.1952705

Keywords: breast cancer, Tanzania, geospatial analysis, healthcare access, cancer diagnostics, sub-Saharan Africa, Bugando Medical Center, spatial clustering, health equity, women's health, cancer epidemiology, referral systems