Rugiatu Jalloh

Assessing predictors of ongoing participation in a breast cancer screening program for low income women in several Michigan counties

BACKGROUND: Research has shown that many deaths due to breast cancer can be prevented by early detection, which provides women with wider treatment options and better chances of recovery. For this reason the BCCCP (Breast and Cervical Cancer Control Program) has been put into operation via a subcontract with the Michigan Department of Community health, funded by the CDC as one of the initial cancer control programs. The BCCCP in Genesee County has engaged in a variety of outreach methods to bring and maintain low and uninsured/underinsured women into the program, yet are sti1l faced with lower screening rates among certain groups of women (specifically African Americans). 

RATIONAL: Working exclusively with the populations that reside in Genesee County and utilize its surrounding screening facilities is valuable because the majority of related published studies focus on utilization data from screening locations in the city of Detroit. Genesee County is an urban community with its own unique characteristics and chal1enges. Data collected on individuals in Detroit and the findings published from them may not be representative of the individuals who use the services offered in Genesee and the surrounding screening facilities. Moreover, government policies may be derived from these studies, resulting in an inadvertent bias against the needs of Genesee area residents

OBJECTIVE: (1) Identify a list of variables from an aggregated dataset collected and compiled from enrollment forms and medical records of women who qualify based on age, income and insurance coverage (or lack there of). It consists of women who had a t least one mammogram from FY1996 through FY2006 (10/1/1995 through 9/30/2006). (2) Find out if there are factors which are more likely predict whether a participant of this program is more likely to be an ongoing participant by receiving subsequent mammograms and therefore is compliant to mammography guidelines based on age and risk factors, or not. (3) Determine if these factors explain the differences between women who routinely receive mammograms vs. those who do not.

METHODS: The effect of each variable of interest will be estimated, both independently and jointly, on likelihood (via Logistic Regression models) for ongoing participation according to the dataset in which several personal identifiers have been removed. Individuals classified as on-going participants vs. non participants (depending on the number of return visits) will be compared and contrasted in respect to the variables which have been deemed relevant in the prediction of the outcome of interest.