Understanding Knowledge and Attitudes About Breast Cancer A Cultural Analysis

OBJECTIVE
To evaluate knowledge and attitudes about breast cancer risk factors among Latinas, Anglo-American women, and physicians.


DESIGN
Ethnographic interviews employing systematic data collection methods.


PARTICIPANTS
Twenty-eight Salvadoran immigrants, 39 Mexican immigrants, 27 Chicanas, and 27 Anglo-American women selected through an organization-based network sampling and a convenience sample of 30 primary care physicians in Orange County, Calif.


MAIN OUTCOME MEASURES AND RESULTS
Data analysis using qualitative content analysis and quantitative cultural consensus analysis, a mathematical technique that determines the degree of shared knowledge within groups and estimates "culturally correct" answers (cultural models), was employed. The content analysis revealed different beliefs about breast cancer risk factors, particularly between the Latinas and the physicians. The cultural consensus analysis found two broad cultural models (defined as groups with ratios between the first and second eigenvalues of > or = 3 and no negative competency scores). A Latina model (ratio = 3.4), formed by the Salvadorans, Mexicans, and Chicanas, emphasized breast trauma and "bad" behaviors, including drinking alcohol and using illegal drugs as risk factors. A biomedical model (ratio = 3.0), embraced by physicians and Anglo-American women, emphasized risk factors described in the medical literature, such as family history and age. Within these broad models, each group of respondents also differed enough in their beliefs to form their own, often stronger, cultural models.


CONCLUSIONS
Ethnography can provide important insights about culturally based knowledge and attitudes about disease. An understanding of the distinctive cultural models regarding breast cancer risk factors will aid future cancer control interventions.

larly between the Latinas and the physicians. The cultural consensus analysis found two broad cultural models (defined as groups with ratios between the first and second eigenvalues of \m=ge\3and no negative competency scores). A Latina model (ratio=3.4), formed by the Salvadorans, Mexicans, and Chicanas, emphasized breast trauma and "bad" behaviors, including drinking alcohol and using illegal drugs as risk factors. A biomedical model (ra-tio=3.0), embraced by physicians and Anglo-American women, emphasized risk factors described in the medical literature, such as family history and age. Within these broad models, each group of respondents also differed enough in their beliefs to form their own, often stronger, cultural models.
Conclusions: Ethnography can provide important insights about culturally based knowledge and attitudes about disease. An understanding of the distinctive cultural models regarding breast cancer risk factors will aid future cancer control interventions.
(Arch Fam Med. 1995;4:145-152) BREAST CANCER is the most commonly diagnosed can¬ cer and is second only to lung cancer as the leading cause of cancer deaths among women in the United States.1 In 1991, health care professionals identified 175 900 new cases, and 44 500 women died of this disease. Fortunately, research indi¬ cates that the mortality due to breast can¬ cer can be reduced by 30% to 40% through the use of mammography and clinical breast examinations.2 However, certain popula¬ tion groups are less likely than others to ob¬ tain these screening procedures. For in¬ stance, the National Health Interview Survey found that Latinas were less likely than An¬ glo-American women ever to have had a mammogram or to have had one in the past year.3 There are many reasons for the underutilization of such cancer control ser¬ vices. Latinas frequendy are poor, lack health insurance, and have inadequate finances to pay for medical care out of pocket.4 '5 Fur¬ thermore, they may have limited knowl¬ edge about cancer-related risk factors and cancer screening procedures6"11 and often delay seeking care for cancer-related symp¬ toms. 12"14 Current efforts at health care reform may be successful in removing some of the eco¬ nomic barriers to medical care; however, other obstacles will continue to exist. A par¬ ticular set of problems arises from differences between Latinos and Anglo-Americans in cul¬ turally based beliefs about illness and disease.
For instance, Perez-Stable et al" found that Latino and Anglo-American members of a health maintenance organization, populations that should have similar access to medical i METHODS The findings reported herein are part of a comprehensive study of knowledge, attitudes, and behaviors associated with breast and cervical cancer. Described below is a summary of the entire study, with emphasis on the methodology re¬ lated to this report on breast cancer risk factors. The Uni¬ versity of California, Irvine, Human Subjects Review Com¬ mittee approved the research protocol. ETHNOGRAPHY We conducted ethnographic interviews to obtain our data. Ethnography is a research method that explores cultural beliefs and behaviors, usually through qualitative analysis of in-depth interviews. However, this method may also in¬ clude systematic data collection techniques, such as rank ordering of interview data, that allow for a quantitative data analytic method called cultural consensus analysis1617 (dis¬ cussed below). Ethnography focuses on shared cultural knowledge and does not assume that researchers are aware of all the relevant questions and issues. Thus, this ap¬ proach is useful for exploratory studies, such as this one, that are designed to understand better culturally based be¬ liefs and to generate hypotheses for future research.

INTERVIEW INSTRUMENT AND DATA COLLECTION
We developed a semistructured questionnaire that contained more than 300 closed-and open-ended inquiries regarding cancer in general, breast cancer, cervical cancer, general ac¬ cess to medical care, access to cancer screening and treatment services, and demographic characteristics. The closed-ended questions came from the National Health interview Survey Supplement Booklet Cancer Control.1* The open-ended ques¬ tions came from the National Health Interview Survey, focus group encounters, and advice from the study's Advisory Com¬ mittee on Cancer Among Latinas that included professional and lay Latino community members. We pilot tested the ques¬ tionnaire using Latinas who did not participate in the study.
A group of health services researchers not involved with the project and the study's advisory committee reviewed the ques¬ tionnaire for content validity. Bilingual investigators trans¬ lated the questionnaire from English to Spanish and backtranslated it using well-established methods. 19 Investigators trained in ethnographic methods con¬ ducted the interviews between August 1991 and August 1992. Because of the sensitive nature of some questions, women investigators conducted the interviews with the Latinas and Anglo-American women. A male investiga¬ tor conducted the interviews with the physicians. The interviewers met with the women in their homes and with the physicians in their offices. They conducted and audiotaped the interviews in either Spanish or English, depending on the respondents' preferences. The inter¬ views lasted between 2  This article reports the results of a study that used eth¬ nographic methods to explore knowledge and attitudes about risk factors for breast cancer. In contrast to the deficit knowl¬ edge method, we made no a priori assumptions regarding knowledge and attitudes. Instead, we allowed the women part of the county; therefore, we targeted that area for sam¬ pling. Most Latinos are of Mexican heritage; however, an estimated 25 000 immigrants from Central America, par¬ ticularly El Salvador, also live in the county.

SAMPLING STRATEGY
Organization-based network sampling19 served as the method to select the nonphysician respondents (see the quantitative analysis section below for the sample size justification). Using this approach, one of the investiga¬ tors (L.R.C.) made presentations to social, educational, and religious organizations and asked for women volun¬ teers. He assigned a code number to each volunteer and randomly selected subjects from each study site. To improve the comparability of the groups, the study design restricted interviews to women without college degrees.
The physician respondents consisted of a conve¬ nience sample of primary care practitioners from the community and from the University of California, Irvine. The physicians received a letter, followed by telephone calls, asking for their participation in the study.

QUALITATIVE ANALYSIS
Trained research assistants transcribed verbatim the open-ended responses regarding the risk factors. Three investigators conducted qualitative content analysis by examining the frequency of citations using a text orga¬ nizing program (AskSam, Seaside Soft-ware Ine, Perry, Fla), evaluating the ranking of the risk factors and estab¬ lishing themes from each group. They divided the 29 risk factors into three groups according to their rank¬ ings: most important (rankings 1 to 10), moderately important (rankings 11 to 20), and least important (rankings 21 to 29). Then they independently evaluated the free-listed risk factors, the rankings, and the openended responses and developed themes from them. They later met as a group and discussed the themes until they reached agreement about them. To test the "trustworthi¬ ness" of the data, the investigators presented the find¬ ings to other groups of Latina immigrants, Anglo-American women, and physicians and asked for their comments. These groups agreed that the identified themes were accurate.

QUANTITATIVE ANALYSIS
We used cultural consensus analysis to test for the exist¬ ence of a shared cultural model and to determine the re¬ spondents rank ordering of the risk factors.16·17 Cultural con¬ sensus analysis is a mathematical model that determines the degree of shared knowledge within groups and esti¬ mates the "culturally correct" answers where an answer was previously unknown. The analysis contains a measure known as competence that assesses the individual's exper¬ tise in relation to a set of culturally correct answers (the model) derived from a group of respondents' answers to questions concerning a specific domain of knowledge. Cul¬ tural consensus analysis provides estimates of each indi¬ vidual's competency and the average competency level of the group. The analysis initially solves for individual esti¬ mates of competency by factoring an agreement (correla¬ tion) matrix among raters. The ratio between the first and second eigenvalues determines whether a single factor so¬ lution exists, indicating a single, shared cultural belief sys¬ tem. Researchers in this field generally accept a ratio of 3:1 and all competency scores falling between 0 and 1 (no nega¬ tive competency scores) as a minimum threshold for as¬ serting that there is a single factor (cultural) solution. The higher the ratio, the stronger the amount of agreement among the group. We also provided the correlation ma¬ trix of aggregate rankings for breast cancer risk factors and used metric scaling, employing principal components analy¬ sis on the agreement matrix, to display the results graphi¬ cally.21 Sample size determination for cultural consensus analy¬ sis follows the same principles as those in other types of analyses. For ordinal data, two parameters are necessary: the degree of concordance among respondents (the aver¬ age Pearson correlation coefficient) and the desired level of validity (estimated by the correlation between the an¬ swers obtained from the sample and the "true" answers).
If there is a great deal of agreement about a topic, the num¬ ber of subjects necessary to obtain a high level of validity is small. The lower the average agreement, the larger the number of respondents must be to maintain a specified va¬ lidity level. Because we had no prior knowledge regarding the amount of agreement about risk factors for breast can¬ cer in our subjects, we chose a low competency score of 0.36 and stringent criteria for proportion of items ordered correctly (95% validity). Using these criteria, a minimum of 17 respondents in each group were necessary.17 to inform us about these issues in their own terms. More¬ over, we studied three subgroups of Latinas-Salvadoran immigrants (the term immigrant refers to women born else¬ where who now live in the United States), Mexican immi¬ grants, and Chicanas (women born in the United States who are of Mexican heritage) as well as Anglo-American wom¬ en and physicians.

CHARACTERISTICS OF THE RESPONDENTS
We interviewed 28 Salvadoran immigrants, 39 Mexican immigrants, 27 Chicanas, 27 Anglo-American women, and 30 physicians. Of these, 28 Salvadoran immigrants, 31 Mexican immigrants, 26 Chicanas, 26 Anglo-American women, and 30 physicians completed the rank ordering task. The women respondents were similar in age, with the mean age of the Anglo-American women, Chicanas, Mexican immigrants, and Salvadoran immi¬ grants being 38, 39, 40, and 35 years, respectively. The Anglo-American women had the most education, with a mean of 14 years (range, 12 to 19 years), followed by the Chicanas with 12 years (range, 3 to 17 years), the Sal¬ vadoran immigrants with 8 years (range, 1 to 16 years), and the Mexican immigrants with 6 years (range, 0 to 13 years). All of the Anglo-American women and Chi¬ canas were born in the United States. The Salvadoran im-migrants had lived here for an average of 4.5 years, and the Mexican immigrants had lived here for 10.5 years.
Fourteen of the 30 physician respondents worked for the University of California, Irvine, and the other 16 had community-based practices. The physicians prac¬ ticed internal medicine, family medicine, or obstetrics and gynecology. Twelve were women and 18 were men. Eth¬ nic groups included Anglo-Americans (18), Latinos (3), Asian American (6), African American (1), East Indian (1), and Iranian (1).

QUALITATIVE FINDINGS
The respondents cited a large number of risk factors for breast cancer; the 29 most frequently cited by all groups and their relative rankings appear in Table 1. The risk factors ranged from those generally accepted by the medi¬ cal community, such as family history and age, to those generally not accepted, such as blows to the breast and chemicals in food. The discussion below provides an analysis of the themes illustrated by the open-ended re¬ sponses regarding the risk factors and their rankings, in¬ cluding quotations that epitomize these themes.

PHYSICIANS
The physicians expressed only one major theme, that biomedically recognized risk factors increased the chance of getting breast cancer. Indeed, nine of the 10 most im-portant risk factors (smoking was the exception) were well established in the medical literature. 22 The physi¬ cians gave little credence to the risk factors ranked highly by the immigrants, such as blows to the breast. The phy¬ sicians indicated that they obtained their knowledge from medical training, textbooks, journals, and professional experience. This comment was typical of the physi¬ cians' responses: Family history, nulliparity, children after age 30; a woman who carries her fat or heaviness above the belt. Family history is prob¬ ably the overriding factor, ANGLO-AMERICAN WOMEN Anglo-American women were closest to the physicians in their rankings; however, they did not replicate them. They expressed two major themes. They believed thatbiomedically recognized risk factors and pollution of food and the environment increased the risk of breast cancer.
They also ranked highly two other risk factors that did not fit these themes, highly stressful lives and breast im¬ plants.

Biomedically Recognized Risk Factors
Anglo-American women accepted some of the biomedi¬ cally recognized risk factors such as family history, hor¬ mone supplements, and exposure to radiation. Indeed, like the physicians, they ranked family history of breast cancer as the most important risk factor. A typical quote was I understand that it kind of runs in families. I really can't tell you if this is true, but I would suspect that sometimes a drug that we take may affect breast cancer, but I can't really tell you which ones. I'm kind of hesitant about x-rays, even on my teeth.
Interestingly, some Anglo-American women listed biomedical risk factors but then questioned their veracity: I think it just happens to some people. They say birth control pills and smoking, but that doesn't necessarily mean that I be¬ lieve that.

Pollution
Anglo-American women also emphasized pollution as a risk factor for breast cancer. They ranked chemicals in foods and environmental pollution Nos. 3 and 6 in im¬ portance, respectively. Environmental pollution in¬ cluded a wide array of risks as indicated by this quote: Maybe depending on where she works, you know, there might be like, might be exposed to nuclear radiation. Not radiation but, a, just like waves, and like if she works on televisions or something. You know, it depends. Or maybe sitting at, like, at a computer. Maybe the computer gives off something. CHICANAS Chicanas ranked highly breast cancer risk factors that were similar in some respects to those submitted by the Anglo-American women and in other respects similar to those submitted by immigrants (described below). Thus, the Chicanas were bicultural in their perceptions of breast cancer risks. The themes included biomedically recog¬ nized risk factors, physical trauma, and pollution of food and the environment. In addition, they often mentioned the lack of medical care as a risk factor and ranked it rela¬ tively highly. While the medical community views medi¬ cal care in this context as a secondary preventive mea¬ sure rather than a risk factor, the Chicanas and immigrants did not make this distinction when asked to list risk factors.

Biomedically Recognized Risk Factors
Like the physicians and Anglo-American women, Chi¬ canas ranked as most important risk factors such as fam¬ ily history, exposure to radiation, and hormonal supple¬ ments. One woman said: If it's hereditary, then I think you are more likely to get it. I've heard that if someone within your family, your mother, your grandmother or sister that has it, there is an increased chance that you may get it, not that you will get it. So they recom¬ mend that you get checked.

Physical Trauma
However, they ranked blows to the breast as the third most important risk factor, much like the rankings of the immigrants.

Pollution
Chicanas shared with the Anglo-American women and the immigrants the concern that chemicals in food and a polluted environment could increase the risk for breast cancer.

Lack of Medical Care
Chicanas shared with Mexican women a belief that a lack of medical care posed a cancer risk, as suggested by this quote: If you're not getting any medical attention or yearly checkups, then you're never gonna really know if you start [cancer]. I think it is going to be bad if your symptoms are at a late stage.

MEXICAN AND SALVADORAN IMMIGRANTS
Mexican and Salvadoran immigrants ranked highly a dif¬ ferent list of breast cancer risk factors that we grouped under the following four themes: physical trauma, be¬ havior/lifestyle, lack of medical care, and chemicals in food. We combined the discussion of their rankings be¬ cause of the similarity in themes.
Physical Trauma The first theme pertained to the risk posed by excessive physical use and abuse of the breasts, including blows to the breasts (

Chemicals in Foods
Like the Anglo-American women, the immigrants wor¬ ried that the chemicals in processed food in the United States posed a cancer risk (ranked No. 9 by both Mexi¬ can and Salvadoran immigrants). They contrasted this with life in Mexico and El Salvador, where they ate mostly fresh food. They also spoke of the greater purity of the water and land in their countries compared with what they perceived as the too many chemicals in the United

States. A Salvadoran woman said:
Contamination is a cause of cancer. Here in this environment we live in there is a lot of contamination from the factories, car exhaust, and cigarettes. All this can cause cancer, I say, includ¬ ing the food. This food is bad. I think that canned food is es¬ pecially bad because it is canned so long. When you buy it, it doesn't have any nutrition left for the body. They are not healthy foods. I think that in our environment fewer people die of can¬ cer than here. Perhaps it's because life is different there. The food is more healthy, more natural. Maybe here they use more dangerous fertilizers.

QUANTITATIVE FINDINGS
The qualitative content analysis above indicated that the study groups had varying beliefs about the risk factors for breast cancer. Through cultural consensus analysis, we evaluated the level of consensus within the groups about the importance of the risk factors. The analysis searched for a single factor solution, or cultural model, that would explain the risk factor rankings. We evalu¬ ated all groups in every possible combination (ie, phy¬ sicians and Anglo-American women; physicians, Anglo-American women, and Chicanas; and so forth) as well as separately. Table 2 displays the results for all groups of re¬ spondents combined and for groups in which the analy¬ sis demonstrated cultural models (ie, the ratios of eig¬ envalues were >3 and there were no negative competency scores) for beliefs about breast cancer risk factors. There was no single cultural model of all groups combined. In¬ deed, the eigenvalues ratio was only 1.2, indicating little consensus about the risk factors. However, other com¬ binations revealed definite cultural models. The Anglo-American women were close enough to physicians in their beliefs to share a cultural model with them (ratio=3.0).
Chicanas shared a cultural model with Anglo-American women and with the Mexican (ratio=3.4) and Salva¬ doran (ratio=3.0) immigrants (but not with physi-cians), indicating their bicultural heritage. Moreover, all Latinas (Chicanas and Mexican and Salvadoran immi¬ grants) shared a cultural model that did not include the Anglo-American women or physicians (ratio=3.4). Other group combinations revealed eigenvalue ratios of less than 3 and often contained multiple negative competency scores, demonstrating that no cultural models existed for them.
When evaluated separately, all the groups demon¬ strated their own cultural models that often were stron¬ ger than the general models described above. By far, the physicians expressed the highest level of consensus. Their mean competency score was 0.73, and the ratio be¬ tween first and second eigenvalues was 8.8. Likewise, the Anglo-American women and the Chicanas demon¬ strated separate cultural models with ratios of 4.1 and 4.2, respectively. The Salvadoran and Mexican immi¬ grant women also formed cultural models, although the eigenvalue ratios of 3.0 in each case indicated less con¬ sensus about breast cancer risk factors than the other groups. Table 3 displays the correlations of the aggregate rankings of breast cancer risk factors among the groups. There were relative high correlations between the Anglo- The spatial configuration of the risk factor rankings obtained by plotting the first against the second principal components. The figure displays the relationship of the points to each other, not to the x-and y-axes. Each letter represents the rankings of one respondent. The closer together the letters appear, the more the respondents agreed about the risk factor rankings. Respondents are represented by letters: a, Anglo-Americans; m, Mexican immigrants; s, Salvadoran immigrants; c, Chicanas; and p, physicians.
resented by the clustering of letters representing the groups. The Figure  Indeed, if the physicians believed the risk factors were important, the immigrants believed that they were unimportant and vice versa. However, between these ex¬ tremes, there is much overlap among the groups of re¬ spondents. As we move from the left to the right of the Figure, we find overlap between the Salvadoran and Mexi¬ can immigrants, between the immigrants and the Chi¬ canas, between the Chicanas and the Anglo-American women, and between the Anglo-American women and the physicians. The bicultural nature of the Chicanas is particularly evident in this visual representation. Like¬ wise, the relatively close, but not completely overlap¬ ping, rankings of the Anglo-American women and the physicians becomes clear through the Figure. COMMENT This is the first study, to our knowledge, that has em¬ ployed both qualitative and quantitative ethnographic methods to explore knowledge and attitudes about any type of cancer. The qualitative portion allowed the Latina immigrants, Chicanas, Anglo-American women, and phy¬ sicians to express their beliefs about breast cancer risk factors from their own frames of reference and in their own languages. The quantitative portion allowed the in-vestigators to determine systematically the major be¬ liefs about risk factors and the level of consensus about them within the groups. Anglo-American women, indicating the bicultural na¬ ture of this group. The physicians reached the highest levels of consensus, embracing risk factors found in the medical literature and giving little credence to others. An¬ glo-American women shared enough of the physicians beliefs to be included in a general biomedicai model, but they also indicated that other issues were important, such as environmental pollution and chemicals in food-risk factors not proven by scientific studies.
These findings augment previous research on knowl¬ edge, attitudes, and behaviors regarding cancer. Others have found that Latinos had less information about can¬ cer, including its causes, than Anglo-American populations.6"11·13 However, we found that knowledge and attitudes about breast cancer risk factors were parts of comprehensive and coherent cultural models rather than nonintegrated bits and pieces of beliefs. The models elu¬ cidated the risk factors that the women perceived as mean¬ ingful, the relative importance of the factors, and the logi¬ cal explanations for their beliefs. Moreover, the findings pointed out the schism between the beliefs of the medi¬ cal care providers and those of potential medical care con¬ sumers. It is important for physicians to recognize these differences to provide more culturally sensitive medical care.23 Two limitations of the study should be mentioned. First, the generalizability of the findings may be ques¬ tioned because the number of subjects was small and the sample was not random. We chose the sample sizes us¬ ing the methods of cultural consensus analysis. While small, the samples were large enough to determine the level of consensus among the respondents, a major pur¬ pose of the study, with 95% validity. Indeed, the small sample sizes required for this type of research should en¬ courage its use in future exploratory studies. We inter¬ viewed volunteers recruited through schools, churches, and other organizations, rather than obtaining a true ran¬ dom sample, because of the exploratory nature of the study and because of the extensive questioning involved in the ethnographic interviews. We attempted to improve the generalizability by selecting the subjects randomly from the volunteers. Nevertheless, additional larger studies by telephone or face-to-face interviews will be necessary to determine if these findings apply to other groups of Lati¬ nas, Anglo-American women, and physicians.
Second, the reliability and validity of the qualita¬ tive data maybe questioned. These concepts, derived from quantitative research, cannot be applied directly to quali¬ tative inquiries. More often, qualitative investigators re¬ fer to the trustworthiness of the research.24·25 We dealt with this issue by having researchers not involved with the study and a Latino advisory committee evaluate the questionnaire content prior to its use. We also required three investigators to review independently the openended responses about risk factors, to formulate themes about them and to discuss the themes as a group until they reached consensus. Finally, we presented the re¬ sults to other groups of Latinas, Anglo-American women, and physicians who indicated that the themes were ac¬ curate reflections of their beliefs.
The methods and findings of this study have im¬ portant implications for future cancer control research and interventions. First of all, ethnographic studies, us¬ ing the qualitative and quantitative analytical methods that we described, can lead to a better understanding of cancer-related knowledge, attitudes, and behaviors. In¬ vestigators should consider the use of ethnography in fu¬ ture studies. Second, the ethnographic findings suggest that cancer control educational materials could be im¬ proved by addressing culturally based beliefs that con¬ flict with those of the medical profession, not only among Latinas but also among Anglo-American women. Third, the findings imply that cancer control interven¬ tions should address varying knowledge and attitudes among ethnic subgroups such as Latina immigrants and Chicanas. Finally, we believe that physicians should be educated about the diversity of beliefs among ethnic groups for whom they provide care. This knowledge may help them to understand why patients with dissimilar cultural backgrounds respond differently to their medical recommendations and may lead to better patient-provider communication.