Introduction

It is not well-established why cancer screening rates are lower among Mexican-Americans in the US. Our study is the first to assess whether access to care, time in the US, and socioeconomic position explain these lower rates using data from the 2001 California Health Interview Survey (CHIS), the largest sample of Mexican-Americans available in any US survey.

Immigrants in the US have lower cancer screening test rates than other groups [1–3]; however the reasons are unclear. Foreign birth and language are barriers to cancer testing even after adjustment for other factors [4], possibly because lack of fluency in English limits knowledge about services and ability to navigate US health care markets [5–7]. Inadequate access to health services also explains lower rates of cancer testing among immigrants [8–10]. Previous studies have shown that Latinos, the largest immigrant group in the US, are less likely to obtain cancer testing than non-Latino whites [11–13]. But, in a prepaid health plan in San Francisco, California, Perez-Stable et al. [14] found no difference in cancer test use between Hispanic and non-Hispanic women even though 77% were born in Latin America and 80.8% responded to survey questions in Spanish. The foreign-born Latinos had been living in the United States for an average of 12.9 years and 30% were of Mexican background, 53% were Central American, and 14% were from other Latin American countries or Spain. This study suggests that equal access eradicates differences in cancer test use regardless of language or time in the US.

California has more Mexican-Americans than any other state, about 8.5 million in 2000. Latinos comprised about 32% of California’s population in 2000, and three-quarters of Latino immigrants are from Mexico. Hunt et al. [15] have criticized research that presumes certain ethnic groups have particular cultural characteristics, fails to define what might constitute the host or ‘‘mainstream’’ society, or makes erroneous assumptions about the historical origins and movement of the populations under study, because these can result in an undertone of ethnic stereotyping. Ramirez et al. [16] recommend specifying the exact nationalities and regions of Hispanics. To address these concerns, we limit our analysis to the Mexican-American population in California. All respondents in our study have similar national origins even though they vary in their birth place and whether they speak English or Spanish. Thus we compare cancer testing among three Mexican-American groups: born in the US, immigrated at least 10 years ago, and in the US less than 10 years.

Of all major racial-ethnic groups in the US, Latinos are least likely to be insured [17], especially in California. A third of Latinos reported lacking insurance for all or part of the year, the highest rate of all racial/ethnic groups [18], so it is unclear whether findings for the general US population will hold for Mexican-Americans in California. For example, income, which could be used to purchase medical services, has been shown to be less important in the general US population than measures of access to health care (usually measured by health insurance and usual source of health care) for obtaining cancer screening services [19]. Because of lower rates of health insurance coverage, use of health services and cancer screening tests may depend more on income level among Latinos. As a result, income may be more important for obtaining cancer screening tests for Latinos than for population groups with higher rates of health insurance. If neither access to care nor income is associated with lower screening rates among Latinos, then do language or other cultural barriers explain them?

Our study addresses whether differential access to care for immigrants and non-immigrants is associated with different cancer test rates. We draw on the behavioral model of health services utilization as well as the extensive literature on cancer screening nationally and in various subgroups to investigate how health care access and time in the US shape use of cancer testing among Mexican-Americans living in California. The behavioral model of health services utilization is a useful conceptual framework for examining factors associated with use of health services, including cancer screening [20–22]. The model emphasizes a systems perspective that incorporates multiple dynamic influences that interact to impact the use of health services, including both individual-level and health system factors. For example, predisposing characteristics, such as patient age and education may influence the likelihood that an individual will get screened for cancer. Other factors, such as having health insurance coverage can enable access, while having a usual source of health care and a continuous relationship with a provider may reinforce screening recommendations. To the Andersen model, we explicitly add language, a predisposing characteristic, and time in the US, an enabling factor. Other authors have used these variables to understand cancer screening use among Asian immigrant groups [23, 24].

Our enhanced conceptual model combined with high-quality local statewide data can be used to design evidence-based policy and programs to increase cancer screening use among Mexican-Americans in California, especially as the state considers strategies to provide medical coverage to its uninsured population. California has some of the best health care access policies in the US, so problems found there are likely the tip of the iceberg nationally.

Ours is the first cancer screening study to compare immigrants and native-born from a single country of origin. This has the effect of controlling for differences that could arise from differences in nationalities and addresses the concern raised by Ramirez noted above. The large Mexican-American population in California and the large statewide sample provided by the CHIS allowed us to delimit our study sample to Mexican-Americans in California. While the literature on cancer screening in minority and immigrant populations is growing, small sample sizes have limited the ability of researchers to test for associations in carefully delimited populations such as our study does. The type of focused local analysis that we perform in this study has the potential to provide a powerful evidence base for policy and program purposes.

Methods

The CHIS is the largest state health survey in the country, and California has the largest population of Mexican-Americans in the nation. We use the CHIS 2001 data to examine cervical, breast, and two types of colorectal cancer test use among Mexican-Americans. Though social structures pertinent to health care delivery include provider-related and other contextual factors [21] not measured by the CHIS, the self-reported items that are available in the CHIS can be usefully grouped into demographics, socioeconomic position, and access to health care. Below, we describe the study sample, define our outcomes of interest and the covariates considered for analysis, and explain our statistical approach.

It is noteworthy to indicate that we view Latino/Latina and Hispanic as having the same essential meaning; however, we favor Latino (masculine sex or both sexes) or Latina (feminine sex) since they are the preferred terms in California. When referring to publications, we use the terms employed by the authors. The focus of our study is on Mexican-Americans, a subgroup of the Latino/Latina or Hispanic population.

Data

We analyzed the 2001 CHIS which interviewed a representative sample of California’s population in Spanish, English and four Asian languages. The 2001 CHIS randomly selected 55,428 California households for an Random Digit Dial (RDD) telephone survey. One randomly selected adult aged 18+ from each participating household was interviewed. The CHIS data were weighted to the 2001 California Department of Finance projections at the county and state levels [25]. Adult completion rates were 59% for the testing and 64% for the extended interview, resulting in a response rate of 38% (59 × 64%) for the extended interview [26]; this is comparable to the telephone-administered California Behavioral Risk Factor Surveillance System Survey [27]. The CHIS provides representative data for California’s ethnically diverse population [28]. More detailed descriptions about data collection methods [29], sample design [30], and weighting [25, 31] are published elsewhere. The CHIS was approved by the University of California at Los Angeles Institutional Review Board and by the California State Committee for the Protection of Human Subjects, and received OMB clearance.

Our study sample includes 9,079 respondents aged 18 and older who reported Mexican origins and place of birth from the 2001 CHIS. Within immigrants, we distinguished between those in the US 10 years or longer because previous research showed that immigrants who have been in the US 10 years or more are not statistically different from the native born [32]. Of the Mexican-Americans living in California, 18% of women and men were recent immigrants, 47% of women and 49% of men were in the US 10 years or longer, and 36% of women and 34% of men were born in the US.

Spanish bilingual interviewers were trained according to protocol in sessions that included both English-only and bilingual interviewers. All bilingual interviewers practiced with the Spanish language instrument before proceeding to live interviewing in Spanish. The training was monitored by Spanish-speaking team leaders in each interview site [29].

Variables

Outcomes

Recent cancer test use was defined for four types of cancer testing according to recommendations for age and sex [33]. A mammogram was considered recent if conducted within the previous 2 years, and a Pap test was considered recent if conducted within the previous 3 years. Two different colorectal cancer tests were measured. In the regression analysis, respondents were considered to have had a recent colorectal exam if they reported a home stool blood test in the last year or a colorectal endoscopy within 5 years. The questions were as follows:

Mammogram:

  • Have you EVER had a mammogram? [IF NEEDED, SAY: “A mammogram is an X-ray taken of each breast separately by a machine that flattens or squeezes each breast.”

Pap smear:

  • Have you ever had a Pap smear test to check for cervical cancer? [IF NEEDED, SAY: “A pap smear is a routine cancer test for women in which the doctor examines the cervix during a gynecological exam, and takes a cell sample from the cervix with a small stick or brush and sends it to the lab. This is not a test for detecting sexually transmitted diseases.”]

Endoscopy:

  • Have you ever had a sigmoidoscopy, colonoscopy, or a proctoscopy to look for signs of cancer or other problems in your colon? [IF NEEDED SAY: “A SIGMOIDOSCOPY is an exam in which a health care professional inserts a flexible tube into the rectum and the lower part of the colon to look for signs of cancer or other problems. A COLONOSCOPY is a SIMILAR exam but uses a longer tube to examine the entire colon. Before a colonoscopy is done, you are usually given medication through a needle in your arm to make you sleepy. A PROCTOSCOPY is an exam that uses a rigid tube.”]

FOBT:

  • The following questions are about the blood stool or occult blood test, a test to determine whether you have blood in your stool or bowel movement. The blood stool test can be done at home using a kit. You smear a small amount of stool on cards at home and send the cards back to the doctor or lab. Have you ever done a blood stool test, using a HOME test kit?

Those who responded ‘Yes’ to any of the above questions were asked “When was your most recent exam?” The possible choices for the response were:

  • A year ago or less

  • More than 1 year ago up to 2 years ago

  • More than 2 years ago up to 3 years ago

  • More than 3 years ago up to 5 years ago

  • More than 5 years ago

  • “Refused”

  • “Don’t know”

An additional category, more than 10 years ago was asked only for endoscopy (and the 5 years ago category was capped at up to 10 years ago). Respondents who reported “refused” or “don’t know” were deleted from the analysis.

Independent Variables

Time spent in the US was based on place of birth and time in the US. Respondents were grouped by: (1) recent immigrants, (2) immigrants in the US 10 years or longer, and (3) born in the US.

Access to care was based on health insurance coverage and usual source of care. Access is defined as reporting both insurance and a usual source of care. The CHIS contains several questions designed to elicit type of health insurance coverage, including full and part-year coverage. If the respondent reported any coverage over the past year, they were counted as insured. To determine whether a respondent had a usual source of care, we used the questions, “Is there a place you usually go for Health care? (yes/no) and “What kind of place do you go to most often? (a doctor’s office, a clinic or hospital clinic and some other place were counted as “yes; Emergency Room was counted as no).

We adjusted for covariates known to be predictive of cancer testing in the general population [34]. Though social perspectives of demographic factors can be changed, demographics themselves cannot. We therefore included demographic variables as controls and do not hypothesize about them. Age is continuous. Education was grouped into less than high school, high school, more than high school. Income was grouped relative to the federal poverty level as 0–99% FPL, 100–199% FPL, 200% + FPL. Having seen a medical doctor in the past 12 months, residence in a metropolitan statistical area (MSA), and family history of cancer were grouped as yes or no. Language of interview is English or Spanish.

Statistical Analysis

To consider associations in the Mexican-American population, we first examine the characteristics of Mexican-Americans in California stratified by sex and the amount of time they report having lived in the US; we then stratify the data by sex and by both time in the US and access to medical care in order to examine whether these subgroups are obtaining cancer screening tests at similar rates; finally we test for associations with covariates shown to be predictors of cancer exams in previous studies [16, 34]. Proportions are weighted to the population.

We used logistic regression adjusting for other covariates listed above to assess the relationship of the two major predictors to each cancer test—time spent in the US and access to care. From this regression model we computed adjusted percentages called predicted marginals [35]. Using the entire sample, we fit a logistic regression model to predict the odds of use of cancer screening tests controlling for other covariates. We then estimated the predicted marginals from this model by assuming that all subjects belong to one group (e.g., have access to care) and averaging over the entire group to obtain the probability of using a screening test for this group. (For more details refer to Korn and Graubard [35]).

Predicted marginals are interpreted like percentages but are adjusted for other covariates. If all observations belonged to one group (males), then their average predicted value for the outcome is the predicted marginal for that group (males). Predicted marginals (PM) are more intuitive than odds ratios and make it convenient to compare screening use between different sub-groups (for example, percentage of men using colorectal cancer screening versus percentage of women using colorectal cancer screening).

Analyses were restricted to age and sex groups for which cancer tests are recommended: women and men 50 and older for colorectal cancer screening; women 40 and older for mammogram; women 18 and older for Pap test. Analyses were adjusted for the complex survey design using SAS V8.2 [36] and SUDAAN V8.0 [37]. Significance tests are two-sided and P-values < 0.05 are considered significant.

Results

Characteristics of Mexican-Americans stratified by length of time lived in the US are shown in Tables 1 and 2. The youngest median age was among recently immigrated men and women. Men and women born in the US were most likely to report access to care (70% of men, 78% of women), followed by those who lived in the US at least 10 years (54% of men and 58% of women). Men and women in the US less than 10 years were least likely to report access (29%, 31%).

Table 1 Characteristics of 3,778 Mexican-American Men in California by Time in the United States, 2001 CHIS
Full size table
Table 2 Characteristics of 5,301 Mexican-American women in California by time in the United States, 2001 CHIS
Full size table

Fewer than 25% of men and women born in the US had less than high school education. The overwhelming majority of immigrants were less educated. There was a distinct gradient in the percentage of men and women living in poverty across the three groups. Among men, 13% born in the US, 30% who lived in the US 10 years or more, and 52% who lived in the US less than 10 years reported family income below the federal poverty level. Among women, 22% born in the US, 45% who lived in the US 10 or more years, and 60% who lived in the US less than 10 years reported family income below the federal poverty level. Even though women in all three groups were more likely than men to report living in poverty, they were more likely than men to report seeing a medical doctor in the past year. Moreover, rates for seeing a medical doctor declined with how recently respondents immigrated. For women, 89% born in the US, 83% who lived in the US 10 years or more, and 76% who lived in the US less than 10 years reported seeing a medical doctor in the past year. For men, 76% born in the US, 66% who lived in the US 10 years or more, and, 58% who lived in the US less than 10 years reported seeing a medical doctor in the past year.

Few Mexican-Americans lived in rural (non-metropolitan) areas in 2001. The longer in the US, the more likely they were to report a blood relative with cancer. Whereas nearly all men and women born in the US took the interview in English (97% and 96%), less than a quarter of long-term immigrants did (24% and 22%), and few recent immigrants did (7% and 4%).

Table 3 shows recent cancer test use by time in the US for the three groups, stratified by access to health services. Regardless of time in the US, respondents reporting access had similar test rates. Lower rates were reported by respondents without access except for FOBT where we had a small sample. Differences were largest for colorectal endoscopy followed by mammography and Pap testing.

Table 3 Recent cancer test use among Mexican-Americans in California by access to health care and time in the US, 2001 CHIS
Full size table

Respondents reporting more time in the US were more likely to report access to medical care (Tables 1, 2) and to report getting a cancer screening exam (Table 3). To test whether time in the US was significantly associated with access, we regressed access on the same factors that we used to test for associations with cancer screening exams (data not shown). Significant associations would suggest that time spent in the US is a pathway to access to care. Results showed a gradient for years lived in the US with Mexican-American women born in the US most likely to have access to care (63%, 95% CI = 61–65%;) followed by those in the US 10 or more years (58%, 95% CI = 57–59%) and finally those in the US less than 10 years (47%, 95% CI = 45–49%).

Predicted marginals, corresponding odds ratios and their respective confidence intervals are shown in Tables 4 and 5. Access to care, having seen a doctor, and family history predicted a mammogram. Age, access, and having seen a doctor were predictive of a Pap test. Age, access to medical care, having seen a medical doctor in the past 12 months and poverty level were associated with colorectal cancer testing for men. Except for poverty level, the same variables were associated with colorectal cancer test use for women.

Table 4 Adjusted correlates for cancer screening tests, Mexican-Americans in California, 2001 CHIS
Full size table
Table 5 Adjusted correlates for cancer screening tests, Mexican-Americans in California, 2001 CHIS
Full size table

Discussion

The rate of mammography for Mexican-American women aged 40 and older in California was relatively high, 78% in 2001 [38]. However, we found biennial mammography was far higher for women with than without access. For Mexican-American women with access, 78% of those born in the US and 79% of recent immigrants reported mammogram in the previous 2 years. The lowest rates, 38%, were among recent immigrants who reported access problems. In Mexico, in 2000, less than 8% of women reported a mammogram in the previous year [39]. The median age of the Mexican-American women we studied ranged from 29 years for women who lived in the US less than 10–38 years for women who lived in the US for 10 or more years. The median age of women diagnosed with breast cancer is 62 years. Both incidence and mortality rates from breast cancer are lower for Hispanic whites than for non-Hispanic whites [40].

Differences between the groups were smallest for the Pap exam. Rates of Pap testing were relatively high for all groups, ranging from 68% for women born in the US with access problems to 91% for immigrants who lived in the US for 10 years or more and reported access. Several contextual factors may help explain these counterintuitive findings. First, Pap test was the most widely used preventive service in Mexico in 2000: 55% of women 40 and older reported having one in the 12 months prior to their interview [39], though use of this test dropped to 30% in rural areas [41]. Among Mexican-American women respondents aged 40 and older in the CHIS 2001, 63% reported a Pap test in the previous year (data not shown); this was the same rate estimated for women in Mexico City [39]. Second, California’s extensive program to provide breast and cervical cancer screening for low-income women, Every Woman Counts, especially targeted Spanish-speaking women. This may help explain the generally high rates of Pap tests reported among Mexican-American women; and particularly why women without access are getting Pap testing at rates nearly as high as women with access. High rates of pap testing are clinically important because the cervical cancer incidence rate is twice as high among Latina Whites as non-Latina Whites [40].

Our study showed that Mexican-American men were more likely than women to use colorectal endoscopy. Another study among all racial-ethnic groups 50 and older using the CHIS found that men were more likely to report endoscopy [42]. That study also found that women were less likely than men to report a recommendation for the exam from their doctor. Findings of test use by sex in national surveys are inconsistent. Pollack analyzed 2002 BRFSS and found more non-Hispanic men than women reported CRC testing, but these results were not statistically significant. Nor was sex significant in the logistic regression, which included both Hispanics and non-Hispanics [43]. In Seeff’s multivariate analysis of the 2000 NHIS [44], sex (not stratified by ethnicity) was significant. Men were more likely than women to report endoscopy or any CRC test, but women were more likely to report having an FOBT. Meissner et al. [45] examined the 2003 NHIS and stratified by sex. Prevalence of testing for sigmoidoscopy and any recommended test was higher for men than for women, but there were no significant differences by sex for test use among Hispanics. It may be that much smaller sample sizes for Hispanics in these national samples and low prevalence of test use is masking differences by sex. Our data specifically examine Mexican-Americans in California and cannot be generalized to all Hispanic or national populations. None of the national studies cited above examined physician recommendation.

We found that time in the US was significantly associated with access, suggesting that time spent in the US is a pathway to access to care. To our knowledge, our study is the first to show that access to care is directly associated with time living in the US. Crude rates showed that more time in the US also was associated with higher reported rates of family history of cancer. We suspect that higher rates of reporting reflect greater awareness of family history as a risk factor for breast cancer. It is unlikely that rates of cancer family history are actually higher among Mexican-Americans with more time in the US.

A comprehensive intervention to promote mammography in Washington [46] showed just how much test use can increase when cost barriers are eliminated. Eighty-seven percent of Latina migrant workers receiving vouchers obtained a mammogram within 30 days, compared with 17.5% of controls. These findings confirmed self-reports that cost was the major barrier to accessing mammograms and that use would increase when financial barriers were removed.

Understanding barriers specific to subgroups is key to developing appropriate policy and other interventions to increase use of cancer screening exams. Because the experience of immigrants from different countries is so varied, we limited our analysis to Mexican-Americans as suggested by Ramirez. Having access to medical care and having seen a medical doctor recently were predictors for all cancer tests. Neither time lived in the US nor language of interview predicted testing in the multivariate models. Though culture seems intuitively important, Abraido-Lanza et al. [11], using the 1991 NHIS, found that health services quality and access were more strongly associated than acculturation with cancer testing for Latinas in the US. Our study using the 2001 CHIS data confirmed that health services access was more strongly associated than time in the US for Mexican-Americans in California. Whereas Abraido-Lanza compared Latinas with non-Latina whites stratified by low or high acculturation [47, 48], we examined Mexican-Americans stratified by time reported in the US.

Other authors have shown how delivering health care in a more culturally competent way might help reduce health disparities [49] and how language-assistance programs could improve health care delivery [50]. In our study, nearly all Latino immigrants were interviewed in Spanish. We assume that respondents who interviewed in Spanish would also prefer to speak Spanish when obtaining medical services.

Earlier studies by Yoon, Bindman et al. located the Spanish language barrier within the context of financial barriers in California. In California, the large number of Spanish-speaking physicians seemed ample to serve the Latino population, but Spanish speakers were not getting health services at the same rate as other Californians. To investigate why, this research team surveyed physicians in 2001 and found that many Spanish-speaking physicians were unwilling to provide services to uninsured patients or those with MediCal coverage [51] because reimbursement rates did not cover the costs that physicians incurred for delivering their services [52]. These studies show that Spanish in-language services are available in California; however resources are needed to fully cover their costs either through adequate public and private insurance reimbursement or with screening-specific programs. In a study of Pap test use in California, Ponce found both Spanish and English-speaking Latinas were among the highest users [23].

The CHIS provides the largest sample of Mexican-Americans in a single state of any health survey in the United States. Nevertheless our study is limited in several ways. Survey responses are subject to self-reported error. Validation studies have shown that respondents tend to telescope reported screening. To reduce the likelihood of this type of self-reported error occurring we used recent rates as our dependent variables. Cell sizes for recent immigrants are small, especially for recent immigrants aged 50 and older. The CHIS 2001 data is cross sectional which does not allow us to draw cause and effect conclusions. Finally, because CHIS data were collected in a single state, our study cannot be generalized to other states or the nation. Rather our findings await confirmation or refutation by subsequent researchers.

In conclusion, we found access to health care for Mexican-Americans strongly associated with cancer testing and, the longer they lived in the US, the more likely they were to report health care access and cancer tests. If respondents reported access to care, cancer test rates were similar regardless of time in the US. Cancer screening tests that are covered by Every Woman Counts had smaller gaps among groups with different amounts of time in the US and with access to medical care than colorectal cancer screening which is not covered by a similar program.

Two natural experiments would be captured by the CHIS data if proposed policy changes are implemented. One is if California were to extend its early detection program to include colorectal cancer screening; the other is if the governor and legislature extend health insurance universally. As the only state with a survey able to monitor changes in subpopulations, California would provide an opportunity to monitor the extent to which making health care coverage universal will deliver on the promise of equality in cancer prevention, early detection, and other medical services found in the prepaid health plan studied by Perez-Stable et al. [14].