Modeling the Impact of Navigated Care on Population-level Breast Cancer Outcomes among African-American Women in Chicago
Michael Cronin, Ganga Vijayasiri, Garth H. Rauscher, Dana Villines, Karriem S. Watson, Yamilé Molina, Aditya S. KhannaBackground: From 2008 to 2012, African American women were 42% more likely to die from breast cancer than non-Latino white women, in part due to later stage diagnoses. Patient navigation, which provides individualized assistance to patients in order to specifically address barriers to care, is suggested to be a helpful strategy to address this disparity. Additionally, patients who have positive experiences with patient navigation are more likely to engage members of their social network regarding breast cancer screening and treatment, which could lead to behavioral changes in their social networks. We model how information provided during patient navigation might diffuse through social networks and effect population-level health outcomes for African American women.
Methods: In this NCI-funded study (R21 CA 215252), we develop an agent-based model (ABM) of African-American women in Chicago, aged 50-74 years to predict how patient navigation might diffuse into and affect population health. The model simulates social network ties of agents, using separable temporal exponential random graph models (ERGMs), developed in the statnet suite of packages in R. Agents in the model develop and screen for breast cancer at empirically determined rates. Agents diagnosed with breast cancer are then eligible for patient navigation, which may influence the behavior of agents within their simulated social network. Given that the effects of these individual-level changes on the broader network are unknown, we will investigate a range of scenarios and quantify the population-level effects in the presence of navigation compared to no navigation.
Data: The model is parameterized using a number of empirical data sources. Demographic and healthcare access data are drawn from public and local data sources (e.g., 2014 Healthy Chicago Survey, 2016 Behavioral Risk Factor Surveillance System data). The social network data are obtained from the Breast Cancer Care in Chicago (BCCC) study. Data on navigation and its effects across different scenarios are obtained from the Patient Navigation in Medically Underserved Areas (PNMUA) and published data sources (e.g., Marshall et al., 2016). Data on genetic risk factors are obtained from the Survey Epidemiology and End Reports (SEER) Program report for African American women. The effects of genetic and hormonal risk factors are parameterized using published data sources (Braithwite 2018 and Munsell 2014). The cancer screening and diagnostic processes are parameterized using input from breast cancer clinicians and epidemiologists.
Progress to Date: The model has been coded in four modules. They are: (1) social network generation and vital population dynamics; (2) breast cancer risk factors and disease progression; (3) breast cancer screening and navigation; (4) diagnosis and social network adjustment. The code can be found at: https://github.com/khanna7/bc-navigation. Currently, data generation and calibration of the model is underway.
Expected Results and Conclusions: We will dynamically simulate the code modules and present data on the simulated effects of patient navigation on the number of breast cancer diagnoses per year and the proportion of early stage breast cancer diagnoses. Additional areas of investigation may include method of navigation implementation and incorporating more nuanced social network structures.