Culture and epidemiology special issue: Towards an integrated study of culture and population health

Introduction

The field of population health, which tackles social, environmental, and biological determinants of health in an integrated framework, is growing rapidly as both a research philosophy and intervention model in global health (McGarvey 2007). In 2007, the Global Fund provided USD$2.6 billion to fight AIDS, tuberculosis, and malaria (Global Fund 2008), the Gates Foundation provided grants totaling USD$1.2 billion for global health projects (Gates Foundation 2008), and the President's Emergency Plan for AIDS Relief (PEPFAR) provided USD$1.8 million to focus countries (PEPFAR 2008, fiscal year 2006). With this tremendous investment, multidisciplinary population health research can play an essential role in achieving global health and wellbeing.

One of the areas that has challenged the field of population health is the incorporation of culture into epidemiological studies (Hahn 1995; Patel 2001; de Jong and Van Ommeren 2002; Trostle 2005). An ability to understand, assess, and interpret the role of culture is important for studying race and ethnicity, which are seen increasingly as socially constructed rather than biological categories. It is also important for situating demographic factors in local context and understanding how and why the associations between demographic factors and health vary across time and space. Despite its presumed importance as a determinant of population health, culture has received substantially less attention in terms of both theory and method compared with the tremendous headway made by social epidemiologists in identifying, measuring, and assessing the impact of social factors on health (Berkman and Kawachi 2000; Oakes and Kaufmann 2006). The papers in this special edition of the Annals of Human Biology aim to highlight the ways that culture can be meaningfully, systematically, and rigorously incorporated into studies of population health.

Towards this aim, Hruschka offers a systematic review of the varied ways in which the concept of culture has been utilized and deployed in the epidemiological and population health literature (Hruschka 2009). Brown and colleagues the describe their use of ethnography with Cherokee and White youth to identify locally meaningful life course barriers and the steps required to translate these elements into an instrument that can be utilized in an epidemiological study design (Brown et al. 2009). Importantly, he and his colleagues attempt to assess the value added to their understanding of the correlates of mood disorders by this ‘ground-up’ instrument design. Kohrt and colleagues build upon the theme of the importance of local context in a study using ethnography, historical political economy, and epidemiology to uncover pathways driving mental health disparities among Hindu caste groups in Nepal (Kohrt et al. 2009). Worthman and colleagues conclude this special issue of Annals by challenging traditional biological approaches to population health (Worthman et al. 2009). Worthman's use of biomarkers opens a broader window onto studying wellbeing across cultural settings. The papers collectively examine the issue of how anthropologists and human biologists can examine with the emerging field of population health. They also raise interesting questions about core concepts, methods of data collection and analysis, and interpretation. We briefly explore these ideas here.

Core concepts

Several factors have limited the integration of culture into models of the social determinants of health: These include the lack of a consensus definition, challenges in identifying and measuring locally important cultural factors, and issues in assessing the causal importance of these cultural factors. These are clearly interlinked. Without a clear definition, theorizing and measurement become problematic, and it is difficult to assess causal pathways without good measurement and theory.

Here we define culture as a system of beliefs, values, norms, and behaviors that are transmitted through social learning. Constellations of beliefs, values, and norms that pertain to a particular domain of activity (e.g. the signs, causes, and treatments for malaria) are called cultural models, and can provide a framework for interpreting and reacting to the world. Cultural models can lead to differing health outcomes by influencing risk behaviors, care-seeking, and care-giving (Hruschka and Hadley 2008). For example, cultural models that elevate the status of males over females may lead to poorer health among females (Mahalingam and Jackson 2007) or lead certain groups to receive systematically sub-optimal treatment (Green et al. 2007).

Methods and sampling

Measuring culture in a way that is amenable to integration into population health studies has been a formidable challenge. One of the central findings of 20th century anthropology was that learned systems can differ dramatically (and in non-intuitive ways) between groups, and thus that each new context requires in-depth study. Ethnography is the traditional method for uncovering such locally-specific beliefs, norms, values, and behaviors. Ethnography is also well-suited to address three common critiques of epidemiology and population health studies that employ culturally constructed categories such as race and ethnicity (Comstock et al. 2004; Gravlee and Sweet 2008). First, it can clarify the local meaning of racial and ethnic categories, which are often used without explicit definition. Second, it can identify what counts as membership in a particular group. And, third it can illustrate the causal pathways that would justify the use of such variables. In other words, it clarifies how cultural groups are defined, how people belong to groups, and why these groups are important predictors of health in the first place.

Ethnographic data and qualitative data can be transformed into survey items, although care must be taken to ensure that survey items are understood in similar ways across all study groups. Part of this challenge can be addressed through translation that also takes into account the content, criterion, technical, conceptual, and semantic equivalence of the instruments across each group (Van Ommeren et al. 1999). Validation methods have been developed that are culturally sensitive and produce generalizable findings (Van Ommeren 2003). Quantitative models now exist to statistically assess the degree of sharing within a cultural group and there are now several methods available to statistically assess if respondents are drawing on the same cultural models (Romney et al. 1986) or multiple different models (Hruschka et al. 2008). This allows a formal test of the extent to which different individuals share similar cultural models. Once a consensus is established epidemiologic studies can assess the extent to which individuals can (or do) approximate their community's cultural model. Some evidence suggests that individuals’ failure to match community models is associated with hypertension, depression, and anxiety (Dressler et al. 2005). Or, alternatively, McDade et al. have recently show that among a large sample in Bolivia, mothers who score higher (i.e. greater agreement) on a culturally agreed upon measure of traditional knowledge had children who scored higher on several measures of health, including nutritional status and C-reactive protein, the latter an indicator of infection. This result maintained even after controlling for a host of other potentially confounding factors (McDade et al. 2007).

Assessing the differential impact of cultural norms on the health outcomes of sub-populations often requires broad sampling of different sub-populations, much broader than is characteristic of ethnographic studies. For instance, if one is interested in assessing how cultural norms around issues of obesity and body size contribute to differences in obesity between ethnic groups, then one's sample would need to include data on the distribution of cultural norms both within and between groups. When comparing across multiple groups, important issues arise that will challenge the population health researcher interested in culture. For instance, based on what anthropologists learn from ethnography, many epidemiological predictions about the impacts of culture on health may involve effect modification by cultural background. This will require statistical tests of interactions, which in turn require sufficiently large sample sizes to ensure that investigators have the statistical power to detect such interactions and assess effect modification. This will be especially true when outcomes are rare, as is the case with severe and persistent mental disorders. Finally, in many cases different cultural groups will occupy asymmetric positions of power in which resources are unequally distributed across groups. This can create problems for statistical modeling when there is little or no overlap in the covariate distributions across groups. When this inequality is deeply entrenched in the structure of a specific society, it is referred to as structural confounding (Oakes and Johnson 2006) and no amount of data can undo the problem.

Analysis

One of the pitfalls of population health research and cross-cultural epidemiology is assuming that the role of ethnography and cultural context is no longer required once data collection is completed. Quite the contrary, ethnographic understandings are just as important during analysis as for earlier stages. First, the ethnographic findings can guide the questions asked in statistical analyses to assure that the hypotheses tested are relevant to the participant community. One possible way of assuring this is conducting community consultations with participants at the conclusion of data collection to identify and discuss analytic questions. Second, when conducting analyses, epidemiologists can refer back to the ethnographic codebook to assure that the variables reflect the categories and meanings of the participant community (Trostle 2005). Brown's work in this special issue (2009) is an excellent example of this as he devotes extensive effort to developing a construct of ‘life barriers’ that fits the experiences and perspectives of Cherokee and White residents in Appalachia.

After ethnographically grounding the constructs, variables, and analytic questions, the next step is to determine the appropriate analytic procedures. A major critique of population health studies that attempt to explain health disparities in terms of culture is that they typically used a dummy variable, where it may be a ‘residual explanation for population differences unexplained by existing variables’ (Hruschka and Hadley 2008). One approach to improve upon this is to specify particular pathways by which culture might account for health differences between populations, measure variables relevant to these pathways, and then assess to what degree these account for the population differences (Holmbeck 1997; Kraemer et al. 2001; Frazier et al. 2004; Wong et al. 2007). Kohrt and colleagues’ analysis in this special issue (2009) demonstrates the process of disentangling risk factors to identify mediators of the relationship of caste with mental health.

Ultimately, regardless of the analytic strategy, when dealing with culture it is especially important to consider alternative hypotheses such as ecological/structural explanations and the influence of inequality shaped by historical political economy.

Future directions

We have outlined what we see as some of the key steps in a cultural epidemiology research program. The suggestions provided above hopefully decrease the chances of drawing sloppy conclusions about the role of culture versus biology in understanding population health (McGarvey 2007). Moreover, through rigorous ethnographic and epidemiologic approaches, the black box of culture is opened to shed light on the underlying mechanisms by which culture influences health outcomes. Even with rigorous methods and analysis, three common traps of interpretation related to culture and epidemiology should be kept in mind and addressed in future research.

First, researchers should understand the limitations of certain kinds of studies for making claims about the causal relationship between culture and health. Health beliefs and health behaviors may be strongly associated in a cross-sectional study, but without longitudinal observation it is unclear if the behaviors lead to current belief patterns or if beliefs motivated behavior. Similarly, an ethnographic study in isolation may not give a representative picture of the kinds of mechanisms at play, and may take as face valid the justifications of informants for particular actions. Longitudinal studies, and ideally intervention studies, are one step toward addressing this issue.

Second, and especially relevant to population health research, interpretation should keep in mind the degree to which the outcome variable of interest is associated with disability and daily functioning. Even if an individual or population endorses high subjective distress or displays a biomarker outside normal values, the meaningfulness of this should be connected to a level of disability. This helps to isolate the areas of wellbeing most requiring intervention. Assessment of disability or daily functioning should become a routine part of population health research (Bolton and Tang 2002).

Third, the locus of where culture is acting should be considered with reference to individual, microsocial, or macrosocial processes. Whereas some psychologists have placed culture at a level independent and above individuals, family, social institutions, and infrastructure, it may be worthwhile to consider how each level is influenced by culture. Multilevel modeling can be employed in future research to explore these levels and cross-level interactions (Blakely and Woodward 2000).

The articles in this special issue provide some initial steps towards addressing these issues. There is still much ground to be gained in determining the most effective approaches to incorporating culture into population health research to promote wellbeing across the globe. Declaration of interest: The authors report no conflicts of interest. The authors alone are responsible for the content and writing of the paper.