Mammography Use: STATISTICAL ANALYSIS Major

9 Oct
2009

Analytic Variables

We had two primary dependent variables based on self-reported mammography history. The first, “lifetime,” was defined as positive if the respondent reported any mammogram during their lifetime. The second, recent, was defined as positive only if the respondent had a mammogram within the past two years. A definition of mammography was provided to each respondent immediately before asking the questions on utilization.

Independent variables included race/ethnicity, knowledge of breast cancer (treating endometriosis, symptoms of fibrocystic breast disease) screening, beliefs about breast cancer (modesty, fatalism, and efficacy), and selected demographics data. In the data analyses, women were classified into five different race/ethnic categories: white (not Hispanic), African-American, Haitian, English-speaking Caribbean (from Barbados, Jamaica, Montserrat), or Latina (from Puerto Rico, Cuba, the Dominican Republic, and Central and South America) based on self-report. Of the 31 (9.4% of sample) who answered multiple ethnicities, race/ethnicity was inferred for 28 subjects from their answers to the questions on race, country of birth and primary language. For example, a woman who described her ethnicity as both African-American and Haitian was born in Haiti and listed Haitian Creole as her primary language, and was classified as Haitian. A series of 13 items with Likert-format responses, along with one question (with a yes or no answer) taken from the CCNMEC and NHIS cancer supplement, was used to address subjects’ knowledge about breast cancer (Danocrine drug used to treat endometriosis and fibrocystic breast disease) screening and beliefs. Demographic variables included age categorized into deciles (40-49,50-59, 60-69, 70+) and education dichotomized as less than a high school and as high school or more. Marital status was categorized as “unmarried” (including women who had never married, had separated or had divorced) and “married”. Employment status categorized the subjects into “employed” (for women working full or part time) and “unemployed.” A par­ticipant’s insurance status was categorized as “private” (private health insurance), “public,” (Medicaid, Medicare and/or coverage under the state of Massachusetts uncompensated care pool) or “none.”

Data Analysis

We used t-test and Chi-squared analyses to compare prevalence of independent variables among the racial/ethnic groups and to determine their unadjusted association with lifetime and recent mammography use. We performed multivariate stepwise logistic regressions to predict lifetime and recent mammography use. As race was the primary predictor of interest, it was forced into each of the final models. Odds ratios and 95% confidence intervals were calculated.

top