Factors associated with smoking in immigrants from non-western to western countries – what role does acculturation play? A systematic review

Introduction We aimed to identify factors associated with smoking among immigrants. In particular, we investigated the relationship between acculturation and smoking, taking into consideration the stage of the ‘smoking epidemic’ in the countries of origin and host countries of the immigrants. Methods We searched PubMed for peer-reviewed quantitative studies. Studies were included if they focused on smoking among adult immigrants (foreign-born) from non-western countries now residing in the USA, Canada, Ireland, Germany, the Netherlands, Norway, the UK, and Australia. Studies were excluded if, among others, a distinction between immigrants and their (native-born) offspring was not made. Results We retrieved 27 studies published between 1998 and 2013. 21 of the 27 studies focused on acculturation (using bidimensional multi-item scales particularly designed for the immigrant group under study and/or proxy measures such as language proficiency or length of stay in host country) and 16 of those found clear differences between men and women: whereas more acculturated women were more likely to smoke than less acculturated women, the contrary was observed among men. Conclusion Immigrants’ countries of origin and host countries have reached different stages of the ‘smoking epidemic’ where, in addition, smoking among women lags behind that in men. Immigrants might ‘move’ between the stages as (I) the (non-western) countries of origin tend to be in the early phase, (II) the (western) host countries more in the advanced phase of the epidemic and (III) the arrival in the host countries initiates the acculturation process. This could explain the ‘imported’ high (men)/low (women) prevalence among less acculturated immigrants. The low (men)/high (women) prevalence among more acculturated immigrants indicates an adaptation towards the social norms of the host countries with ongoing acculturation.


Introduction
High-income or economically developed western countries, such as North America, North and West Europe, Australia, and New Zealand are characterised by considerable ethnic diversity. The immigrant populations residing in western societies are heterogeneous, comprising people from different countries of origin, with different motivations to migrate, cultural identities etc. Health outcomes among immigrants in terms of mortality and morbidity have been largely covered in international research. Fewer studies focused on immigrants' health behaviour and especially on the determinants of health behaviour such as smoking. Smoking is one of the leading causes for premature death and thus a particularly risky health behaviour. It causes a variety of cancers and cardiovascular diseases [1][2][3][4]. On the individual level, smoking patterns vary with socioeconomic statusand here especially with educationwith stress, age, and gender. Smoking is also influenced by peer group social norms on the interpersonal level and country-specific tobacco control interventions (public and work place bans, cigarette prices etc.) or the society's social support system of smoking on the socio-political level [5][6][7][8][9][10][11][12][13][14][15].
Smoking behaviour as well as norms and attitudes towards smoking also differ between countries. Lopez et al. [16] speak of the 'smoking epidemic' , a four-stage model describing the progress of smoking among men and women: in the 1 st stage smoking predominantly involves men. In the 2 nd stage, smoking prevalence increases steeply among men and slightly among women. The 3 rd stage is characterised by a further increase in smoking among women. Smoking among men starts declining and prevalences among men and women converge. The 4 th stage shows a further decline in prevalence among both men and women (see Figure 1). Rather than looking at the stages as isolated parts, the model should be understood as a process or continuum over decades [17]. Whereas economically developed western countries are in the advanced phases of the epidemic, countries such as China, India, Syria or some Southeast Asian or African countries are more located towards the beginning or early phases of the epidemic [18][19][20][21][22].
Thus, immigrants from non-western to western countries move from an earlier to a more advanced stage of the epidemic. As health risks and resources they acquired in their countries of origin are not static and may be subject to change in the host countries, immigrants might adapt to, for example, the host country's smoking behaviour [23]. This change in smoking behaviour might be the result of an acculturation process which starts immediately after arrival in the host country. Acculturation is a complex phenomenon: it refers to a dynamic process through which behaviours of immigrants change as a result of interactions with individuals in their (new) social and cultural environment [24]. The concept of acculturation can be either understood as unidimensional -where immigrants move along a continuum ranging from a weak adaptation to the host culture to a strong oneor as bidimensionalwhere immigrants may independently maintain the culture of origin and adapt to the host culture. The best known bidimensional acculturation model was developed by Berry [25], although recent research activities propose a more extensive approach to acculturation. Schwartz et al. [26], for example, also incorporate cultural practices, such as customs and traditions, cultural values, such as individualism or independence and a cultural identification, such as the attachment to a certain cultural group. It becomes apparent that acculturation is influenced by three main contextual areas: the context prior to immigration, the immigration context itself and the settlement context [27].
Acculturation models are widely used in health behaviour research with various measures, ranging from indirect proxy measures (e.g. time spent in host country, or proficiency in the language of the host country), to multi-item scales [27][28][29][30]. The relationship between acculturation and, for example, substance abuse, dietary practices, leisure-time activity or health services use has mainly been studied in the United States [31][32][33][34][35][36][37][38]. There is also evidence from the US linking smoking behaviour to level of acculturation [39][40][41][42][43][44][45]. In Europe, research on this relationship is scarcesmoking among immigrants has been investigated mainly with regard to the association between smoking and socioeconomic status.
A combination of the 'smoking epidemic' model and the acculturation model should allow to predict the smoking behaviour of immigrants: populations that migrate from non-western to western countries will 'bring along' the smoking behaviour from their countries of origin and maintain it for some time after migration. With the commencing acculturation process in the host Figure 1 The stages of the 'smoking epidemic' proposed by Lopez et al. in 1994. countries, these 'imported' risks will change and immigrants will adapt to the smoking behaviour of the majority population in the host country (see Figure 2). This assumes that acculturation is an important (albeit not necessarily the only) determinant of smoking behaviour among immigrants.
The aim of this paper is to identify common patterns for smoking behaviour among immigrants (or foreignborn persons) from non-western to western countries with a special focus on the role of acculturation through a systematic review over the international literature.

Search strategy
The database PubMed was searched in May 2012 and again in March 2014 using the MeSH terms displayed in Figure 3. All studies were filtered first by title and then by a narrower filter procedure following the inclusion and exclusion criteria listed below. Next, all full-texts of the remaining articles were read and then entered for the review or dropped. Additionally, the reference lists of included articles were scanned for studies that were not detected by the database search.

Eligibility criteria
All peer-reviewed primary studies in English language published after 1990 were included. Studies were included if they matched the following criteria: quantitative study; analysis of factors associated with cigarette smoking (irrespective of frequency and quantity); focus on immigrant adults (18 years and older); specification of country of origin; personal migration experience from an economically developing non-western country of origin to an economically developed western host country (North America, North and West Europe, Australia, and New Zealand). Studies that focused on foreign-and native-born persons were included if the proportion of immigrants or foreign-born persons was more than 90%. Studies with less than 90% were only included if a clear distinction between foreign-and native-born persons was made and factors associated with smoking were examined and presented separately for each group.
Studies were excluded if no distinction between immigrants (foreign-born) and their offspring (born in the host country) was made, if the immigrant group under study was defined by a specific characteristic (e.g. pregnancy, occupation, addiction etc.); if the study focused on risk factors for (or causes of ) certain diseases, was not empirical but theoretical, merely investigated time trends in smoking prevalence, focused on any form of drug intake or any form of tobacco use other than smoking cigarettes, was conducted for validating a questionnaire, evaluated a prevention programme or focused on media/advertisement influence on smoking behaviour.

Synthesis of results
The results reported in the articles were included as findings in this review if they were derived from any form of statistical analysis, regardless of whether descriptive analyses or regression models. For example, variables which were not significantly associated with smoking in simple regression analyses might have not been included in the subsequent multiple regression analyses. As non-significance does not equate with non-relevance, all factors known to be relevant were covered in this review. In the case of conflicting results between descriptive and analytical procedures, those derived from the highest level of statistical analysis were counted as finding. A metaanalysis was not performed as definitions and measurements of the variables of interest were not consistent between the studies.
The retrieval procedure was performed independently by two investigators. By the end of the first retrieval round, there were five articles with conflicting decisions regarding inclusion. After discussion, four articles were excluded and one was included in the review. After the second search there were no discrepancies. The final number of studies included was 27. The retrieval procedure and the criteria for exclusion and inclusion are outlined in Figure 3.
The two investigators also independently assessed and documented methodological quality of the studies to trace possible bias within as well as across the studies. The following information was documented for each study: aim, definition of the immigrant group, data source (if secondary data was used) or sampling procedure (if primary data was generated), method of data collection, study design, number of immigrant participants, statistical methods applied, operational definition of dependent and independent variables, and limitations reported.

Study characteristics
The 27 studies included were published between 1998 and 2013 and all of them used a cross-sectional design; the majority (19 of 27) were conducted in the USA; two were conducted in Germany, and one each in Ireland, Canada, the Netherlands, the UK, Norway, and Australia. Most of the studies focused on Asian immigrants, originating from China (5 of 27), Korea (3 of 27), Vietnam (2 of 27), and the Philippines (1 of 27). Three studies investigated more than one Asian immigrant group. The second-largest immigrant group under study were Latinos/Hispanics from Mexico (2 of 27) and El Salvador (1 of 27). One study focused on African immigrants. One study compared the smoking behaviour between Asians and Latinos/Hispanics among others. Two studies focused on Arabs. One study each investigated smoking among Ethiopians, Polish immigrants, Turkish immigrants, and immigrants from the former Soviet Union; and two studies focused on more than one immigrant group. The sample sizes ranged from 96 to 16,738 persons. Eleven studies had a sample size below 1,000, in ten studies the sample size ranged from 1,000-5,000 participants and six studies surveyed more than 5,000 persons. Fifteen of 27 studies focused on foreign-born persons only, in seven studies more than 90% were foreignborn, in five studies between 7% and 77% of the participants were foreign-born. Four studies restricted their study sample to men and five studies performed their highest statistical analysis only for men as the number of female smokers was too small. Eleven of 27 studies stratified their analysis by gender, seven adjusted for gender. Table 1 presents contextual information of the different countries, such as the GDP per capita as prosperity indicator and the smoking prevalence in the immigrants' host countries, their countries of origin and among immigrants themselves as reported by the studies included: smoking prevalence among men in the countries of origin is higher than that among men in immigrants' host countries, whereas the contrary applies to women. Additionally, the gap in smoking prevalence between men and women is larger in the countries of origin than in the host countries. Among immigrants, this genderdifference is not as large as in their countries of origin but still larger than in the host countries. Table 2 presents the factors associated with smoking among immigrants, for men and women combined if analyses were only adjusted for gender or no genderspecific differences were observed; or for men and women separately if gender-specific differences were observed or analyses were restricted to one gender only.

Study findings
Acculturation was focused on in 21 of all 27 studies. Table 3 illustrates the different acculturation measures used. Only five studies applied multi-item acculturation scales that were developed particularly for the immigrant group under study and were based on a bidimensional concept of acculturation. Multiple questions were used to identify the preference for and the fluency in the culture of the country of origin or the host country. Higher scores on the scales reflected an orientation more towards the host country culture. All other studies used proxy measures for acculturation, mainly proficiency in the language of the host country and length of stay. The majority of the studies (16 of 21) that included acculturation emphasised differences by gender, irrespective of the acculturation measure applied (acculturation scale or proxy measure for acculturation): less acculturated men (or men with lower language proficiency or short length of stay) had higher smoking prevalences than more acculturated men (or men with higher language proficiency or long length of stay). However, the contrary applied to women. Women with a lower level of acculturation (or lower language proficiency or short length of stay) had lower smoking prevalences than women with a higher acculturation level (or higher language proficiency or long length of stay). These gender-specific patterns were observed both in studies that explicitly stated that they used proxy variables to measure acculturation (11 of 21) and in studies that did not link variables such as length of stay or language proficiency to acculturation (5 of 21).
Gender was one of the factors most frequently reported to be associated with smoking behaviour in immigrants (23 of 27). The majority of the studies reported marked differences by gender, with men being more likely to be current smokers than women.  Additionally, educational level (19 of 27), age (9 of 27), employment (6 of 27), and marital status (7 of 27) were frequently reported. Younger and non-married persons were more likely to smoke than older and married persons. Smoking was positively associated with a low educational level among men but with a high educational level among women (11 of 27). In 4 out of 27 studies, employed persons had a higher smoking prevalence than unemployed persons. Other factors, such as income, religion, alcohol consumption or knowledge of tobacco health risks, were reported by less than five studies (see Table 2), and often only for men.

Quality assessment of the studies included for the review
All studies presented a clear study aim, defined their study population accordingly and clearly specified dependent and independent variables. Besides descriptive analyses, logistic regression analyses were applied in 20 of 27 studies (see Table 4).
Concerning the sampling procedures of the studies generating primary data (19 of 27), the vast majority used a list-based (mainly a telephone list-based) technique with focus on names specific to the immigrant group or a community-orientated (via migrant organisations or networks) sampling strategy. It has to be noted that (I) a list-based approach is restricted to only those appearing on, for example, the telephone list, and (II) a random sample for a quantitative study can hardly be realised by using a community-orientated approach as it is highly likely to include predominantly socially integrated people.
A major part of the studies (18 of 27) applied both translated instruments and bilingual interviewers. Six studies reported either the use of translated instruments or the employment of bilingual interviewers. Two studies did not provide any information in this regard and one study was conducted in the host country language (English) only. Applying translated, culturally adapted, and validated instrumentsespecially on acculturation and smokingas well as bilingual interviewers may lower the risk of selection and information bias [46,47]. If, however, the risk of selection bias is high, the generalisability of the study results to a group other than the selected one is questionable (see Table 4).
Moreover, using self-reported data to assess smoking behaviour may lead to an under-or overestimation of the smoking status, for example, due to socially desirable responses. As all studies used self-reported data, information bias cannot be excluded. However, one study validated the self-reported data by using expired carbon monoxide measures and showed a significantly high correlation between self-reports and measurements. Furthermore, all studies applied a cross-sectional design. Here, factors associated with smoking and smoking status are measured simultaneously. Thus, it is neither possible to assess causal pathways between factors associated with smoking and smoking status nor to draw conclusions on changes over time. Additionally, as the sample in cross-sectional studies comprises different age or birth cohorts, the study findings may also be the result of a cohort effect.

Discussion
Our analysis shows that smoking among immigrants from economically developing non-western to economically developed western countries is positively associated with (I) a low acculturation level among men, (II) a high acculturation level among womenas measured by acculturation scales or proxy measures such as language proficiency or duration of stayand in general (III) with being male, of younger age, non-married, and  Pre-migration life events Exposed vs. non-exposed Non-exposed vs. exposed 20 Post-migration life events No association Exposed vs. non-exposed 20 Socio-economic and socio-demographic factors such as marital status, education, employment status, age, and gender are already known to determine smoking in any population group, irrespective of their migration status. Among immigrants, acculturation is an additional factor linked to smoking. One possible explanation can be found in the transition brought about by the immigration from non-western to western countries, which occupy different positions in the 'smoking epidemic'. Western countries tend to be located towards the advanced phases of the epidemic, whereas non-western countries tend to be located more towards the early phases, and smoking among women tends to lag behind that in men [48,17]. This is supported by higher prevalences among men from non-western countries compared to men from western countries, lower prevalences among women from non-western countries compared to women from western countries and by the large gap in smoking between men and women from non-western countries as seen in Table 1. It is compatible with the hypothesis that with the immigration from non-western to western countries immigrant men and women may 'import' their smoking behaviour from the country of origin to the host country. If acculturation is regarded as process which starts with the immigration to the host country, this might explain the high smoking prevalence among less acculturated men and the low smoking prevalence among less acculturated women. Furthermore, with increasing duration of stay in the host country, immigrants might 'move' towards the advanced phases of the smoking epidemic and adapt to the smoking behaviour of the host country. This is likely to be the result of an ongoing acculturation process. In many nonwestern countries smoking is still uncommon and socially unacceptable among women but highly acceptable among men [42][43][44][45]. This pattern of social support of smoking might account for the smoking behaviour among recent immigrants. In western societies smoking is equally accepted among men and women [42][43][44][45]. Immigrants might identify with these values and attitudes and adopt them in the course of the acculturation process towards a higher smoking prevalence among women and a lower one among men. These findings correspond well to the 'operant model of acculturation' as proposed by Landrine & Klonoff [49]. It predicts that health behaviours with a high prevalence among initially low-acculturated immigrants will decrease in prevalence with increasing acculturation, whereas health behaviours with a low prevalence among low-acculturated immigrants will become more common with advancing acculturation. This model has already been successfully applied to health behaviours such as smoking, diet, alcohol use, and exercise [49,50]. The phenomenon that higher levels of acculturation are associated with increases in unfavourable health behaviours is also known as the 'immigrant paradox'.
Only few studies so far have explicitly focused on the relationship between acculturation and the stages of the 'smoking epidemic'; the studies investigating the course of the epidemic among immigrants have mainly focused on its association with educational level [51,52]. In general, research on smoking behaviour among immigrants with focus on acculturation is in its infancy in Europe. Only one of the European studies included in this review explicitly stated that it also measured acculturation.
However, this should not divert from the fact that, besides acculturation, there are several other determinants of smoking on the individual, cultural, economic and political level. There might also be interactions between certain determinants (e.g. between education and employment status). Differences between countries of origin and host countries regarding, for example, cigarette prices or bans on smoking in public/work places might additionally affect smoking in immigrants.         Table 4 Methodological quality assessment of the articles included in the review (Continued) 19 Wiecha, Lee, Hodgkins [74] 1998 in USA

Strengths and limitations of the review
This review combines the concept of acculturation and the model of the 'smoking epidemic' for the research on smoking in immigrants. So far, studies focused on either of both aspects. Linking the health transition in terms of the 'smoking epidemic' to the acculturation process is the major strength of this review. Besides this strength, the review has some limitations: first, factors associated with smoking were dichotomised only (e.g. high vs. low, long vs. shortsee Table 2). This was done in order to better illustrate the main findings. Such a simplification leads to a loss of information and to a possible misclassification as specific definitions and measurements of the variables (e.g. long vs. short length of stay) are likely to vary between the studies. Second, 19 of 27 studies were conducted in the USA and most focused on Asian immigrants. This might bias the findings of the review towards this immigrant group or towards US immigrants in general. At the same time it reveals the need to further investigate factors associated with smoking among immigrants in countries other than the USA. Third, the retrieval procedure was performed by using the database PubMed only. Thus, it cannot be excluded that additional articles on the topic under study appeared in journals not listed in the database. However, as the database covers a large number of journals from the life sciences and biomedicine, it is not likely that a substantial number of papers on smoking behaviour among immigrants have been missed.

Conclusion
Implications for future research While 11 of 27 studies stratified their analysis by gender, seven studies only adjusted for gender. Future studies should not only appreciate gender differences in smoking behaviour by adjusting for gender, but explicitly assess gender as a potential modifying variable. This might also help to further explore the 'immigrant paradox' where higher acculturated women are more likely to participate in unhealthy behaviours. Moreover, measures of acculturation should not just be unidimensional, with either a weak or a strong orientation towards the culture of the host country. As immigrants do not have to give up their traditional culture in the process of acculturation, unidimensional measures do not satisfactorily reflect the inherent multidimensionality of acculturation. Such multi-item scales should furthermore be based on theoretical or conceptual work. This could promote a more direct approach to studying the process of acculturation (rather than relying on proxy measures) and thus help to better understand the association between acculturation and health behaviour. Additionally, information bias might be minimised.
All studies we could include in this review were crosssectional. As acculturation is a process, only longitudinal studies will properly track how changes in acculturation affect health-related behaviour over time. This also applies to other variables which may have an effect on smoking such as education or economic factors; they are time-variant as well and therefore prone to change with increasing duration of stay in the host country. Besides, not only the situation in the host country influences the smoking behaviour among immigrants, the process of immigration itself and the situation in the countries of originand thus the context of migrationmay have an impact on health behaviour. Consequently, future studies should focus on a life-course perspective to reveal mechanisms and determinants responsible for uptake, maintenance and cessation of smoking [53,54].

Implications for prevention and health promotion
Health professionals need to be aware of the patterns of smoking among immigrants identified in this review and adapt existing health promotion programs or plan and initiate new programs accordingly. In particular, they need to employ different strategies for immigrant men and women. Programs for male immigrants should aim to further decrease smoking prevalence, whereas programs addressing immigrant women should aim at preventing smoking initiation. Immigrants of both sexes need to be made aware of the social and cultural forces that operate during the process of acculturation and may affect their smoking behaviour. As smoking and acculturation are not only individual but also social and group phenomena, key persons from immigrant communities should be involved in implementing strategies in their communities, settings, and networks [55]. Thus, knowing the factors associated with smoking among immigrants is only the first step towards its prevention.