I Introduction
Anthropology and epidemiology are dedicated directly or indirectly to the study of human cultural practices and how those practices affect human health and disease. They must be fundamentally concerned with the theories that help guide and explain their discoveries as well as with the methods used to make those discoveries. Chapter 2 explained that these disciplines began with a fundamental concern for fieldwork, with the researchers always refining and adjusting how and why they collect information. Chapter 3 reviewed a few of the variables used to describe disease patterns, showing that new questions and concerns are raised when social scientists, particularly anthropologists, unpack the assumptions underlying such variables. This chapter pays more specific attention to the collection of data. I argue that data collection is built on a series of cultural conventions, not all of which facilitate valid measurement. Data collection is improved when those conventions are acknowledged and confronted.
The chapter-opening quote and Figure 4.1 show some of the cultural conventions around data collection. The quote portrays a standard critique of anthropology, namely that it studies too few people and mistakes mere anecdote for data. Figure 4.1 shows a standard kind of cartoon genre, the "public opinion poll." This one pokes fun at the pollsters and all the invisible interpretive errors that can take place between a household interview and a final monolithic summary statistic. Both jokes manifest some underlying discomforts about how we learn what we know. Why can't stories be understood and used as data? How can we know when snores are miscounted as presidential support?
i/OO
qu&rTiOM.
i/OO
qu&rTiOM.
AMCINTCRflaATCTHE
HOW THE SMTlRE MATlcW Fg£L£A0CUTW£ VITAL
issued THE pw:
- Figure 4.1. Public opinion poll in "Shoe" comic strip. C. Cassat and G. Brookins, "Shoe." 3/11/90. Tribune Media Services, Inc. All rights reserved. Reprinted with permission.
Three concepts are usually helpful in guiding discussions of the quality of data collection: validity, reliability, and generalizability. Although scientists define validity in many ways, the concept refers to correspondence between what one thinks one is measuring and what one is really measuring. Reliability, on the other hand, is the likelihood that a measure will repeatedly yield the same results. Reliability can be defined as similar results over time from multiple uses of the same test, or it can be defined as similar assessments reached by multiple observers. Generalizability refers to the possibility that a study's outcomes based on a sample also will apply to the broader group from which the sample is drawn.
Anthropologists do many things to avoid anecdote. They often seek to increase the accuracy of their methods by relying on participant observation, an approach to doing research that relies on long-term contact with a specific place and community. (See the description in the next section of how participant observation methods helped an anthropologist suggest culturally valid measures of infant mortality in Brazil.) Anthropologists commonly write fieldnotes to describe their daily experiences, but they also rely on guided observations, surveys, maps, and reviews of written records such as newspapers, parish records, or other archival data. Because participant observation emphasizes duration, breadth, and depth of contact, it increases validity at the expense of reliability and gener-alizability. That is, its strength is that it provides detailed information from many different kinds of sources about a particular place. Whether other observers would conclude the same things about that place, and whether that place is really like many other places, remains untested in most anthropological studies.
A. Developing Ethnographically Sensitive Vital Statistics: Measuring Infant and Child Mortality in Brazil
Marilyn Nations has undertaken a series of innovative projects in Brazil combining anthropology and epidemiology. One of these studies sought to compare official infant mortality rates with popular (unofficial) meanings and experiences of infant death in the northeast region of Brazil, which has among the highest infant mortality rates in the world (Nations and Amaral 1991). The goals of this study were to measure death accurately and to correct any apparent errors in official population-level measures of infant and child mortality.
The investigators undertook 118 interviews with parents of dead or dying children in three poor rural communities. They observed healing ceremonies, child wakes, funerals, and burials. They learned that deaths in Brazil only get counted in official statistics when registered at government-authorized sites. But this process often necessitates multiple trips to a capital city and requires the services of a doctor to ascertain the cause of death. Parents are offered incentives to register a child's birth but disincentives to register a death: although registering a child's birth can lead to benefits such as milk and food supplements, school enrollment, voting, and health care, registering a child's death can incur penalties such as losing food supplements. Not surprisingly, births tend to be registered, but death registry is comparatively incomplete.
The anthropologists also studied the types of rituals involved in the dying process and the community members involved in those rituals. They used their data to develop a network of people who had inside knowledge of infant and child deaths: grave diggers, coffin makers, indigenous midwives, undertakers, priests, and the like. They then trained these people to detect and record infant and child deaths in one town, and they compared over 12 months the number of deaths detected by these "popular death reporters" to those officially registered and to those found in a census of all households with children under age five. Nine deaths of children under five were reported from all these sources put together: the popular reporters identified eight of them; the survey six, and the official registry only four. The sources assembled through anthropological field-work provided more accurate counts of infant and child deaths than the official registry or a household survey. The authors conclude that they have painted a picture of infant and child death that is more nuanced and valid than the cold and incomplete numbers provided by the state or national government.
And what do epidemiologists do to increase the accuracy of their measurements? They increase accuracy by relying on large random samples, standard (pretested) measurement scales, and tested survey questions. These methods improve reliability and generalizability, sometimes at the expense of validity. That is, epidemiologists focus on the likelihood that another researcher would find the same results that they did, and that the site under study is a good example of other sites like it. But because of unexamined assumptions and the popularity of particular types of variables at particular times, the validity and accuracy of epidemiologic measures may not always be just what epidemiologists intend them to be.
How do epidemiologists and anthropologists design research to obtain valid and reliable data? Epidemiologists divide their studies into experimental and observational designs. Experimental designs involve some type of purposive and measured intervention on the part of the researcher, whereas observational designs involve data collection without researcher intervention. Observational studies are called descriptive if they present the overall distribution of disease in a population, and analytic if they explain observed patterns in terms of proposed causal or etiologic factors. Observation and description also are essential parts of any anthropological study, but few anthropologists would describe intervention and experiment as appropriate objectives. By participating in the daily activities of a given group over time, anthropologists believe they come to understand what it means to be a member of that group. Anthropologists are thus likely to be able to compile more intimate and comprehensive pictures of daily life than could be obtained by a complete outsider.
One way to try to resolve some of the differences between epidemiological and anthropological forms of observation has been to create interview tools that combine descriptive narrative accounts of local concepts of illness with systematic coding and analysis. One such tool called the EMIC (Explanatory Model Interview Catalogue) was developed primarily to facilitate the study of local concepts of mental health (Weiss 2001). The EMIC uses a combination of open-ended and closed-ended questions, so that respondent categories unknown to a researcher can be uncovered and measured, but so also can measures of the frequency and causes of categories presented by researchers.
Anthropology and epidemiology are typically thought to involve dramatically different analytic processes, but there is actually a great deal of similarity between them. One traditional approach to anthropological analysis has been described as follows: the anthropologist commonly finds a behavior pattern in the multiple singular acts of particular individuals, then creates abstract roles and relationships from those patterns, and then seeks or proposes principles that can account for those roles and relationships (Nadel 1957). Epidemiologists also generalize about the behavior of individuals when they construct measures of exposure. They find disease patterns in data aggregated across multiple individuals, and they seek or propose principles that can account for those patterns. Anthropologists sometimes seem to forget that expanding from a unique individual utterance or action to a more general description of group context requires an act of descriptive abstraction. On the other hand, epidemiologists sometimes seem to forget that categorizing risk factors or choosing measures requires qualitative judgment. Thus descriptive abstraction occurs in both epidemiology and anthropology, although epidemiologists commonly seek and describe their abstracted patterns of disease using quantitative descriptions built from a statistical vocabulary, whereas anthropologists more often present their abstracted patterns of culture using qualitative descriptions built from an everyday vocabulary.
Post a comment