The project utilized a prepost study design. The outcome variables of interest were measures of cancer-related knowledge, attitudes and behaviors assessed by telephone interview of African Americans of both sexes, aged >18 years, who lived in census tracts with a high proportion of black residents.
Westat Inc., a national consulting firm, conducted 35-minute pre- and postintervention telephone surveys in the selected census tracts in all four cities. Trained interviewers contacted one respondent per household, selected by random-digit dialing based on a sample frame of residential phone numbers in the census tracts. Respondents were deemed eligible if they were >18 years of age and black or African-American. If the person answering the phone was eligible, that person was interviewed. Otherwise, he/she was asked to call an eligible person to the phone. Demographic questions included birth date, highest grade completed and annual household income. “Awareness” questions focused on whether the respondent had received information about cancer (Arimidex canadian is used to treat breast cancer) and from what sources (including Meharry or Morehouse) and whether the respondent recognized the names of the project and its slogans. Another set of questions addressed knowledge, beliefs and attitudes regarding factors that may contribute to the development of cancer (i.e., “Will smoking cigarettes/getting exercise/eating lots of fresh fruit and vegetables/having stress in your life/getting a bump to your body make it more likely or less likely that you will get cancer or make no difference?”). Respondents were asked if they had heard of screening tests (appropriate to gender); whether or not the tests had been recommended; and, if performed, how many times and how recently. Additional questions explored dietary history, history of tobacco use, health insurance status and use of healthcare services.
We compared knowledge and practices in each of the four sites before and after the intervention and compared the Nashville and Atlanta intervention sites to the Chattanooga and Decatur comparison sites, respectively, for both time periods. Survey data were cleaned using SAS computer software and weighted to compensate for selection bias, noncov-erage and nonresponse effects. A poststratification adjustment was applied to the weights for target population parameters based on known population demographics and 1990 census tract totals by age, gender and race.
The project used the Jackknife Method of replication producing design-consistent estimates of variance of sampling error. A design-effect modified Pearson Chi-squared test statistic determined significance of differences between the pre- and postintervention periods at all sites. A conventional
p value <0.05 established significance level applying a two-tailed distribution method. Data were transported to WesVarPC software for analysis and adjustment for complex sampling design effects.
Four multiple regression models were performed to test the impact of the intervention on the communities, including their interaction as a term in the model on each of the following total scores: cancer prevention knowledge, attitude, eating behavior and awareness.