Phenotype Discovery and Geographic Disparities of Late-Stage Breast Cancer Diagnosis across U.S. Counties: A Machine Learning Approach

Abstract

Background: Disparities in the stage at diagnosis for breast cancer have been independently associated with various contextual characteristics. Understanding which combinations of these characteristics indicate highest risk, and where they are located, is critical to targeting interventions and improving outcomes for patients with breast cancer. Methods: The study included women diagnosed with invasive breast cancer between 2009 and 2018 from 680 U.S. counties participating in the Surveillance, Epidemiology, and End Results program. We used a machine learning approach called Classification and Regression Tree (CART) to identify county “phenotypes,” combinations of characteristics that predict the percentage of patients with breast cancer presenting with late-stage disease. We then mapped the phenotypes and compared their geographic distributions. These findings were further validated using an alternate machine learning approach called random forest. Results: We discovered seven phenotypes of late-stage breast cancer. Common to most phenotypes associated with high risk of late-stage diagnosis were high uninsured rate, low mammography use, high area deprivation, rurality, and high poverty. Geographically, these phenotypes were most prevalent in southern and western states, while phenotypes associated with lower percentages of late-stage diagnosis were most prevalent in the northeastern states and select metropolitan areas. Conclusions: The use of machine learning methods of CART and random forest together with geographic methods offers a promising avenue for future disparities research. Impact: Local interventions to reduce late-stage breast cancer diagnosis, such as community education and outreach programs, can use machine learning and geographic modeling approaches to tailor strategies for early detection and resource allocation.

Risk Factors and Geographic Disparities in Premature Cardiovascular Mortality in US Counties: A Machine Learning Approach

“…The above advantages of using CART and random forest methods, together with geographic information systems, have been demonstrated in a prior study investigating the phenotypes of late-stage breast cancer diagnosis and their geographic distributions. 21 This study further demonstrates the validity of this approach in uncovering geographic disparities in PCVM. Future studies should also consider spatially varying relationships using a geographic random forest model to evaluate risk factor importance of PCVM across space.…”

Variation in and Factors Associated With US County-Level Cancer Mortality, 2008-2019

“…Indeed, the lack of correspondence in some locations between prevalence and VI suggests that cancer mortality risk factors may be modified by place-specific factors. [38][39][40][41] It is conceivable that the interplay of a given risk factor with comorbidities or other exposures, observed or unobserved, may potentiate or ameliorate the association of that risk factor. Thus, the difference between prevalence and VI might be associated with modulating factors, either positive or negative, where further investigations are needed.…”

“…For example, although obesity prevalence was generally higher in the South compared with that in the West, obesity VI in several western states around Colorado was higher than in most parts of the South except for Florida. Indeed, the lack of correspondence in some locations between prevalence and VI suggests that cancer mortality risk factors may be modified by place-specific factors . It is conceivable that the interplay of a given risk factor with comorbidities or other exposures, observed or unobserved, may potentiate or ameliorate the association of that risk factor.…”

Risk factors and geographic disparities in premature cardiovascular mortality in US counties: a machine learning approach

“…The above advantages of using CART and random forest methods, together with geographic information systems, have been demonstrated in a prior study investigating the phenotypes of late-stage breast cancer diagnosis 21 and cancer mortality 22 . This study further demonstrates the validity of this approach in uncovering the combination of risk factors and their relative importance in predicting county-level PCVM.…”