JMIR Publications

ABSTRACT

Background:

Background:

Geographic variation in late-stage breast cancer remains an important public health concern, particularly in large and socioeconomically heterogeneous states such as Texas. Although mammography availability is central to early detection, modern cancer prevention and care increasingly depend on digital connectivity for appointment scheduling, telehealth, patient communication, health information access, and care navigation. However, the relationship between digital access and geographic variation in late-stage breast cancer burden has received limited spatial epidemiologic attention.

Objective:

Objective:

This study examined the association between county-level digital access and late-stage breast cancer burden across Texas while accounting for mammography availability, socioeconomic characteristics, and spatial dependence.

Methods:

Methods:

We conducted an ecological spatial analysis of all 254 Texas counties using late-stage female breast cancer case counts from the National Cancer Institute State Cancer Profiles, mammography facility data from the Texas Cancer Information Program, and county-level digital access and socioeconomic indicators from the Centers for Disease Control and Prevention Social Vulnerability Index and related American Community Survey data. Late-stage breast cancer counts were modeled using Bayesian Poisson conditional autoregressive models with county female population as an offset. Spatial dependence was modeled using a Leroux conditional autoregressive specification. Model fit was assessed using the Deviance Information Criterion, Widely Applicable Information Criterion, and Log Marginal Predictive Likelihood. Sensitivity analyses alternatively coded suppressed county cancer counts as 1, 2, or 3 cases.

Results:

Results:

Late-stage breast cancer burden demonstrated significant positive spatial autocorrelation across Texas counties (Moran I=0.190, P<.001). In the Bayesian spatial model, a higher percentage of households without internet access was associated with greater late-stage breast cancer burden (β=0.027, 95% CrI 0.019-0.035). This association remained positive and statistically credible across sensitivity analyses using alternative coding of suppressed case counts. Mammography facility density was not independently associated with late-stage burden after adjustment for digital access, socioeconomic characteristics, and spatial dependence. The model including household internet access showed improved fit compared with the model excluding this variable, with lower DIC (1400.75 vs 1438.28) and WAIC (1399.70 vs 1436.59) and higher LMPL (-703.52 vs -728.98). Residual spatial dependence remained substantial (ρ=0.687, 95% CrI 0.181-0.983).

Conclusions:

Conclusions:

County-level digital access was independently associated with geographic variation in late-stage breast cancer burden across Texas, even after accounting for mammography availability, socioeconomic characteristics, and spatial structure. These ecological findings suggest that digital connectivity may be an important contextual factor in cancer prevention and care navigation. Strategies to improve breast cancer outcomes may benefit from considering digital infrastructure alongside traditional investments in screening access, while future longitudinal and multilevel studies are needed to clarify the mechanisms underlying these area-level associations.