Background
Between 1990 and 2000, the U.S. Hispanic population increased from 22.3 million to 35.3 million, and by 2050, it is expected that over 97 million Hispanics will live in the United States, accounting for nearly one-fourth of the U.S. population [
1,
2]. In the Western United States, the Hispanic presence is much larger (in 2005, Hispanics, primarily originating from Mexico and Central America, comprised 35% of the adult population in California) [
3]. Hispanics are a diverse group, varying significantly with regard to country of origin, demographic factors, health-related characteristics and acculturation level including generation in the U.S., and dominant language. In general, Hispanics of Mexican and Central American descent (hereafter referred to as Latinos) are significantly more disadvantaged on many socioeconomic and health measures than those of Cuban, Puerto Rican, or South American descent [
4].
Among Latinos, being born in the US tends to be positively associated with higher educational attainment, higher income, English proficiency [
5] and Internet access [
6], but negatively associated with health-promoting behaviors and chronic health problems [
4]. As a result of generational differences, the overall demographic and health characteristics of the Latino population are likely to change substantially as the percentage of the Latino population that is born in the U.S. or immigrate at a very young age increases. The 2002 National Survey of Latinos found that nearly all Latinos born in the U.S. speak English, with 46% of second generation and 22% of third-plus generations being bilingual, compared with only 28% of foreign-born Latinos (24% of whom are bilingual) [
5]. Educational attainment is also significantly higher among second and third generation Latinos as compared with foreign-born Latinos, with 75% vs. 46%, respectively, completing at least high school and 14% vs. 9% having college degrees. While Latino adults are significantly less likely than non-Hispanic Whites and African-Americans to use the Internet, English-dominant (prefers to communicate in English) and bilingual Latinos are approximately three times more likely to use the Internet than Spanish-dominant (very limited or no English language proficiency) Latinos, and the difference between English-dominant Latinos and non-Hispanic Whites Latinos significantly diminishes after adjusting for education [
6].
Health behaviors and chronic diseases have also been shown to be subject to the effects of generational differences due to acculturation. Based on evidence of several studies of the effects of acculturation on Latino health behaviors and health, Lara
et al posit that the effect of acculturation, while not absolute, is generally negative with regard to health behaviors such as dietary practices, smoking, obesity, and physical inactivity, resulting in an increasing rate of diabetes and onset at younger ages [
7]. For example, while studies have found dietary changes that are healthy (e.g., decreased use of lard, cream, and sausage) and unhealthy (e.g., less fresh fruits and vegetables, rice, beans, and more sweets) among more acculturated versus less acculturated Latina women, the researchers judged the overall effect of acculturation on diet to be more negative than positive [
7‐
9]. The effect of acculturation on Latino smoking appears to be gender-dependent; acculturation is associated with increased smoking prevalence among women, but has little or no effect on men. Ultimately, this results in the more acculturated men and women having smoking rates similar to those among non-Hispanic Whites [
7,
10,
11]. Three studies found a higher prevalence of obesity among more acculturated Latino adults and adolescents than among the less acculturated [
12‐
14]. In contrast to these negative effects, Crespo
et al. found that acculturation was positively associated with participation in leisure-time physical activity [
15].
According to the 2005 California Health Interview Survey, there were approximately 900,000 Latinos of Mexican or Central American descent aged 20–64 covered by non-Medicaid health insurance in California in 2005, approximately 23% of all insured adults in this age group [
3]. Of these, approximately 24% had limited ability to speak English, 51% spoke English well or very well, and 25% spoke only English. The objective of our study was to explore the heterogeneity of Spanish-dominant, bilingual, and English-dominant Latinos aged 25–64 who were members of a large Northern California health plan with regard to educational attainment, income, health-related characteristics, access to information technology, and preferred methods for receiving health information and health education. Our intent is to provide information to health plans serving a culturally diverse Latino population with information that might aid in understanding differences in health care needs and outcomes for these populations, as well as for planning health education services.
Methods
Study design and populations
The Kaiser Permanente of Northern California (KPNC) Adult Member Health Survey, conducted every three years starting in 1993, is a project designed to inform policy makers, researchers, administrators, and clinicians both inside and outside the Health Plan about healthcare-related characteristics and preferences of insured adults [
16]. This confidential survey, conducted with an age, gender, and geographically stratified random sample of 42,000 KPNC adult health plan members aged 20 and over, covers sociodemographic characteristics, health status and health conditions, health-related behaviors and lifestyle risk factors, use of selected medications and dietary supplements, preventive services, and previous use of and interest in different methods of obtaining health education. While in 2005 the survey was primarily conducted using a mailed questionnaire, the survey could also be completed on a secure website or by phone. Due to cost constraints, the survey has only been conducted in English, and people known to require a translator or require written information in a language other than English have been excluded from the sample. However, since the numbers of members who speak little or no English have been increasing over recent years, and since learning about health and health care disparities has become a priority for not only the health plan, but for society at large, the decision was made to conduct a pilot survey in Spanish for Latino members with limited or no English proficiency (Spanish-dominant group). Both surveys were approved by KPNC's Institutional Review Board.
The English-dominant Latino group was comprised of 532 women and 388 men aged 25–64 who had self-identified as being of Mexican-American or Central American ethnicity on the 2005 KPNC Member Health Survey (MHS). Of these 920 Latino members, 93 women and 87 men who indicated that they preferred to use Spanish when talking about or learning about their health were classified as bilingual; the remaining 433 women and 301 men were classified as English-dominant. Since the Health Plan did not have computerized information about race-ethnicity, country of birth, or generation in the United States for all members, it was not possible to estimate what percentages of bilingual and English-dominant Latinos responded to the survey.
The Spanish Member Health Survey questionnaire was a slightly modified version of the 2005 Member Health Survey, translated into Spanish using terms and wording chosen to be understandable to a monolingual Mexican-American or Central American adult with limited education. The survey sample included an age- and gender stratified random sample of 309 adults aged 25–64 who were identified from a health plan patient demographics database as requiring a Spanish-language interpreter for clinic visits and Spanish-language preference for written materials (at the time, information about language preference and interpreter needs was available for 95% of KPNC members). A survey packet consisting of a cover letter, information sheet, self-administered questionnaire, and return envelope was sent in May 2006, with the offer of a $10 gift card for participating in the survey via phone interview or self-administered questionnaire. Nonrespondents were called approximately 3 weeks after the initial mailing, at which time the bilingual research assistant encouraged them to complete the survey with her over the phone or offered to send another copy of the self-administered questionnaire if that was their preference. Those who were sent a second copy of the questionnaire but did not mail back a completed form were contacted one final time to offer the phone interview. In all, 78 women and 93 men (66.7% and 57.3%, respectively, of people who were contacted by mail or phone) completed the survey, 126 by self-administered and 45 by interviewer-administered questionnaire over the phone. Of the 171 respondents in the Spanish-dominant group, 69.8% (n = 120) self-identified on the survey as Mexican-American, 16.3% (n = 28) as Central American, and 13.9% (n = 24) as Other Hispanic/Latino.
Study variables
Sociodemographic measures included age, education (highest level of school completed), and total household income from all sources in 2004, before taxes. Country of origin was not ascertained because we had been advised that including this question would reduce the number of recent immigrants and non-documented workers willing to participate. Both the education and income questions had categorical responses. Based on past research, a greater number of response categories for very low levels of educational attainment were included in the Spanish language survey.
Health status was assessed using the standard question of "In general, would you say your health is excellent, very good, good, fair, or poor." Follow-up questions asked for separate ratings of physical health (including pain) and emotional/mental using the same scale. For the Spanish survey, the rating categories were "Muy bueno, bueno, regular, malo." Respondents were also asked to indicate from a checklist whether during the past 12 months, they had or had taken medication for a number of health conditions, including diabetes, high blood pressure, high cholesterol, frequent heartburn or acid reflux, severe back pain or sciatica, depression, sadness, or very low spirits lasting at least 2 weeks, and frequent problems with sleep.
Health behaviors were ascertained using several questions. Current smoking status was derived from two questions that asked whether the individual had ever regularly smoked cigarettes, and if so, whether the individual smoked now, even occasionally. Obesity (BMI ≥ 30 Kg/m
2) was derived from self-reported height and weight. Exercise frequency was based on response to a question about how often the individual usually got physical exercise, such as walking, swimming, gardening, golf, and tennis. People who indicated "3 to 4 times a week" or "5 or more times a week" were considered to be exercising ≥ 3 times a week. People who indicated an exercise frequency of less than 1 to 2 times a week (i.e., "2 to 4 times a month", "once a month or less," or "never") were considered to get exercise less than once a week (sedentary behavior). Dietary questions focused more on behaviors than evaluating content of the diet. Usual number of servings of fruits and vegetables was based on a single item "During an average day, about how many servings of fruits and vegetables do you usually eat (1 serving = a half cup or a medium piece)." This single item fits with the public health message to consume "5 A Day," not differentiating between fruits and vegetables. DiSogra and Hudes previously showed that using two items that asked about average number of servings for total fruit and total vegetables separately was a good way to estimate the fruit and vegetable consumption to obtain population estimates for tracking purposes, although it would likely produce a slightly higher estimate than estimates based on multiple items about servings of specific foods [
17]. Dietary behavior related to fat intake was assessed by asking about how often the individual tried to eat reduced fat (low fat or nonfat) foods. People who indicated doing this "all of the time" or "most of the time" were considered to usually try to eat reduced fat foods. Daily multivitamin use was taken from a checklist of dietary and herbal supplements that the individual had used during the past 12 months. Belief about the relationship of behavioral risks and health was based on response to a question "How much do you think habits/lifestyle, such as exercise, what you eat, and your weight, can affect your health." People who indicated "quite a bit" or "extremely" were considered to believe that these had a large effect, while those who indicated "not at all" or "a little bit" were considered to believe that these had little or no effect.
Access to information technology (personal computer, Internet) was ascertained by the questions "Do you have access to a personal computer?" and "Do you have access to the Internet?" with response options of "Yes, at home," "Yes, at another location," or "No" [access]. Use of selected health education modalities in the prior 12 months and interest in future use of different health education modalities were assessed by two checklist questions, the latter of which read, "In addition to talking with your doctor, how would you prefer to learn about taking care of health problems and improving your health?" The exact wording and format of questions in English and Spanish is available upon request from the first author.
Analysis
All analyses were done using weighted data. The Bilingual and English-dominant respondents had all previously been assigned post-stratification weights so that analyses using weighted data would reflect the age (in 5-year intervals), gender, and geographic composition (KPNC service populations) of the full adult membership at the time of the survey. The Spanish-dominant sample was similarly assigned weights based on age (5-year intervals) and gender of Spanish-dominant members aged 25–64 at the time of the survey. Because previous research has shown significant gender differences in population-based sociodemographic characteristics and behavioral health risks, analyses for men and women were performed separately.
Analyses were performed using weighted data and SAS (Statistical Analysis Software) procedures for analysis of data obtained from complex survey designs [
18]. Proc Surveyfreq was used to estimate the unadjusted percentages and confidence intervals (see Additional file
1). Because the linguistic groups differed by age (Table
1), and age was associated with many of the health-related behaviors and health indicators, the weighted percentage estimates for all three linguistic groups were then age-adjusted to the age distribution of the English-dominant Latinos (women ages 25–39: 0.516, ages 40–64: 0.484; men ages 25–39: 0.523, ages 40–64: 0.477) using Proc Surveyreg as outlined by Gossett et al. [
19]. To assess whether the age-adjusted percentages significantly differed between pairs (e.g., English-dominant women vs. Bilingual women), Proc Surveyreg was also used to subtract one age-adjusted estimate from a second and apply a t-test for difference [
19]. Finally, Proc Surveylogistic models were used to confirm the Proc Surveyreg results and evaluate whether controlling for education in addition to age reduced differences between the linguistic groups found after adjusting for age alone.
Table 1
Age distributions of Spanish-dominant, Bilingual, and English-dominant Latino study groups prior to age-adjusting
25–39 yr | 53.2% | 43.3 | 51.6 | 56.9 | 60.4 | 52.3 |
40–64 yr | 46.8% | 56.7 | 48.4 | 43.0 | 39.6 | 47.7 |
Mean (SE) | 40.2 (1.0) | 42.3 (1.0) | 40.3 (0.5) | 39.4 (1.2) | 39.0 (0.8) | 40.3 (0.6) |
Median | 39 | 42 | 38 | 37 | 38 | 39 |
Discussion
Our comparison (in the context of a health care plan) of Spanish-dominant, bilingual, and English-dominant Latino men and women aged 25–64 revealed several differences that have implications for the provision of health education services. While the linguistic groups did not differ with regard to prevalence of diabetes, high blood pressure, or high cholesterol, the Spanish-dominant Latinos had higher percentages of men and women reporting heartburn and women reporting back pain, suggesting that Spanish language educational materials for prevention and management of these health problems might help improve quality of life for this population segment. Similar to findings of other studies [
7], Spanish-dominant Latinos were also significantly less likely than the bilingual and English-dominant Latinos to consider their health to be good, even after adjusting for education. Our relatively small Spanish-dominant Latino sample did not allow us to assess whether the lower rating of health reflects actual differences in health and well-being or, alternatively, cultural differences in how people think about health status. Spanish-dominant and bilingual Latinos were also significantly less likely than English-dominant counterparts to believe that health risk factors such as diet, exercise, and weight had a large impact on health, partly due to low level of formal education and less prior exposure to the concept of risk reduction in their home countries. Thus, they may be less ready to act on the general messages being given by health care providers and the public health community about eating more fruits and vegetables, trying to reduce the amount of fat in their diet, and increasing frequency of exercise to improve health.
With regard to behavioral health and lifestyle factors, our results confirm the findings of Crespo
et al that Spanish-dominant Latinos are more likely to report being physically inactive than English-dominant Latinos [
15]. Moreover, their hypothesis that Spanish-dominant Latinos may get substantial amounts of physical activity during work or activities of daily living (which would not have been captured by our exercise question) is indirectly supported by our finding that Spanish-dominant Latinos were less likely than English-dominant Latinos to be obese.
In our sample, a very large proportion of Spanish-dominant Latinos expressed an interest in receiving health education. Because Spanish-dominant Latinos had much lower levels of formal education, household income, and access to personal computers and the Internet than English-dominant Latinos, it was not surprising to find greater preference among those speaking only Spanish for lower technology modalities of health education, including videos, television programs, and taped health messages accessible by phone, and lower preference for Internet-based sources. Despite the lower levels of education, more than one-fourth of the Spanish-dominant Latino group indicated that they had gotten information from the health handbook they received from the health plan and more than half were interested in receiving health newsletters and short articles or brochures about health. This suggests that while providing Spanish-language materials and programs via the Internet may have great logistical and economic appeal, these modalities will likely not reach a majority of Spanish speaking Latinos who require and are eager to receive health education. Moreover, because of the low household income, it is likely that even those with Internet access from home are not using very high speed DSL or broadband connections nor computers with very fast processing speed and large RAM, making it very difficult for them to interact with websites that have extensive graphics, to interact with online programs, and to download materials.
This study has both strengths and limitations. While the fact that the data all came from a health plan population in Northern California might affect the generalizability of the results to the broader insured or uninsured population, this can also be viewed as a strength in that it provides a more controlled environment for observing differences associated with dominant language. All of the survey respondents were not only insured for primary care, but being seen by primary care clinicians whose practice guidelines calls for screening for and counseling their patients about behavioral/lifestyle health risks and making referrals to the health plan's health education resources. Limitations of the study include its use of non-validated self-report data; the relatively small numbers of Spanish-dominant and bilingual Latinos; likely under-participation in the English language survey by bilingual and English-dominant Latinos of very low educational attainment and relatively low overall response to that survey; lack of information about country of birth and generation in the United States; and inability to weight the bilingual and English-dominant Latinos data to the age-gender distribution of Latino health plan members with some English proficiency because census-type race/ethnicity statistics for the membership were not available.
Conclusion
Our findings highlight important differences among Latinos of different English language proficiency with regard to educational attainment, income, health status, health behavior, technology access, and health education modality preferences that ought to be considered when planning and implementing health programs for this growing segment of the U.S. population. Specifically, there is a great need to provide Spanish-dominant and bilingual Latinos with education about the relationship between current health habits/lifestyle and future health and functional status in order to sustain the favorable health status indicators (lower morbidity/mortality rates, lower prevalence of many chronic diseases, lower rates of disability) that have been well-documented among Spanish-dominant or lesser acculturated Latino populations in the United States [
4,
7]. For the Spanish-dominant population, education about depression and back pain targeted to women and heartburn/acid reflux targeted to both men and women might provide a teachable moment to explore this relationship between lifestyle choices and health outcomes. Finally, Latinos with limited English proficiency appear to be very interested in obtaining health education, but it will likely be more accessible if it is provided through non-Internet-based modalities such as videos, taped telephone messages, Spanish language print materials, and small group visits with a doctor or patient educator. Further research should be done to determine the generalizability of these results to other U.S. Latino populations and to monitor whether the identified differences in education, income, health behaviors, health issues, and preferred health education modalities diminish or widen over time.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
NG conceived, designed, and conducted the surveys, performed the statistical analyses, interpreted the data, and drafted the manuscript. CI consulted on the data interpretation and helped to draft the manuscript. Both authors read and approved the final manuscript.