Abstract
In rural India, with hundreds of thousands of villages, a priority from a programmatic perspective is to efficiently determine which villages are at highest risk of HIV/AIDS transmission. The authors first report on the use of a rapid ethnographic approach in 10 rural villages of Karnataka, India, to develop a domains and indicators framework of village-level HIV/AIDS risk and a subsequent rapid assessment tool. They then analyze the rapid ethnographic approach and the rapid assessment tool to discuss differences and commonalities among rapid designs. They also discuss if these studies can be properly categorized as ethnographies, are mainly qualitative in nature, and are in essence participatory, and how appropriate they are to the public health field in general and the HIV/AIDS field in particular.
Introduction
The HIV/AIDS pandemic is one of the most serious threats to global health. Although the previous estimate of 5.7 million individuals infected with HIV in India has recently been revised downward (UNAIDS/WHO, 2006) to about one half of that level (UNAIDS, 2007), this still represents a high burden of infection and disease. The prevention and control of HIV/AIDS has proven to be an immensely complex public health and social challenge due to the interrelationships between a host of societal and individual factors. Added to this, the scale of the population in India taxes the resources of HIV prevention programs. In rural India, with hundreds of thousands of villages, a priority from a programmatic perspective is to determine efficiently which villages are at highest risk of HIV/AIDS transmission. An assessment conducted in India by our group (India-Canada Collaborative HIV/AIDS Project, 2004) observed differentials in levels of high-risk sexual behavior across 239 villages enumerated in the district of Bagalkot, in the northern part of Karnataka state in southern India. HIV prevalence was measured in 10 villages, and the large variations in prevalence observed among these villages (ranging from 0% to 8.2%) (Becker, Ramesh, Moses, & Blanchard, 2006) raised the question of possible community level characteristics that might explain this variability. From a programmatic and research perspective it is highly relevant to determine the characteristics of villages that are at greater risk than others.
A standardized tool that could rapidly assess risk environments should be of enormous utility for the identification of villages that might require more intense intervention activities (the state of Karnataka, with a total population of about 55 million, has 27,481 villages across 29 districts). We collected evidence using a rapid ethnographic approach from the 10 villages where our program had detailed HIV prevalence data (India-Canada Collaborative HIV/AIDS Project, 2003) to develop a domains and indicators framework of village-level HIV/AIDS risk and a subsequent rapid assessment tool. We sought to achieve a more in-depth understanding of the dynamics of the villages to judge the plausibility of a number of village characteristics as causal factors or markers of higher risk for HIV transmission. The use of rapid ethnographic methodology was considered the best approach for several reasons, which we discuss later. The results of the study have formed the basis for a second stage in our research program, to pilot-test the rapid assessment tool in approximately 250 villages from a sampling frame of 7,500 villages in 15 districts, to assess its reliability and predictive validity.
The use of rapid ethnography in research is not new. Known under several guises, such as rapid assessment (Beebe, 1995, 2001), quick ethnography (Handwerker, 2001), rapid appraisal (sondeo) (Hildebrand, 1979), and rapid rural appraisal (Chambers, 1994a, 1994b, 1994c), to mention several, the commonalities and distinctions among these approaches may be of consequence but it is not always clear how. Among the common features is the formation of multidisciplinary research teams, the use of several data collection methods to verify information through triangulation, and the completion of the project quickly, usually in 3 to 6 weeks (Harris, Jerome, & Fawcett, 1997). Rapid ethnography is also considered a compressed form of study design that uses a combination of elicitation techniques, focus groups, and key informant interviews (LeCompte & Schensul, 1999). The distinctions appear to be based on differing emphases. One broad distinction is that between rapid ethnography and ethnography itself. According to Beebe (2001),
rapid ethnography shares many of the characteristics of ethnography, but differs in two main ways: 1) more than one researcher is always involved in data collection and the teamwork is essential for data triangulation; 2) more than one researcher is involved in an iterative approach to data analysis and additional data collection. (p. 1)
Nonetheless, there are several issues to address to locate these study designs properly within the spectrum of community and public health research.
Taking our study as a case, we address several related questions. First, we make a basic assumption, that these “rapid” or “quick” techniques or methods are a particular form of study design. Then, after providing a thorough description of the use of a rapid ethnographic design to investigate village-level dynamic of HIV transmission, we examine whether this particular form of study design is shared across its different guises. In addition, we look at how this particular form (or particular forms) differs from more traditional study designs. Specifically, we examine if rapid designs should be categorized as ethnographies, if they are mainly qualitative in nature, if they are in essence participatory, and how appropriate these rapid designs are to the public health field in general and the HIV/AIDS field in particular.
Rapid ethnographic assessment study and village-level rapid assessment tool
A key premise of our study, as argued by others (Rose, 1992), is that the determinants of variations between individuals within a population might be different from the determinants of variations between populations. In both industrial and developing countries, researchers have suggested that community-level factors have an important impact on sexually transmitted infections. A study in rural Zimbabwe has provided evidence that social capital might be associated with the adoption of safer behaviors and HIV avoidance (Mushati, Mlilio, Campbell, Nyamukapa, & Gregson, 2002). The effect of the social environment on the health of sex workers has been examined in a case study of a sex workers' collective in Sonagachi, Kolkata (Chander & Masaki, 2003), and a dramatic increase in the rates of condom use was observed. A study of U.S. counties concluded that most variations in syphilis rates were accounted for by sociodemographic characteristics (Kilmarx et al., 1997). A study in rural Tanzania suggested that community attributes have an important impact on the risk of HIV infection (Bloom, Urassa, Isingo, Ng'weshemi, & Boerma, 2002). The idea is that community-level characteristics (village-level in the case of our project) might influence proximate factors, and ultimately HIV infection rates, in this population.
Understanding HIV transmission dynamics in India is both a methodological challenge and an important programmatic need. Currently the dominant approach for prevention strategies is based on the UNAIDS/WHO formula, whereby much of the country is considered to have either a low-level epidemic (HIV prevalence less than 1% in the general population and less than 5% in high-risk groups) or a concentrated epidemic (HIV prevalence less than 1% in the general population and greater than 5% in high-risk groups). There is a general acceptance that in the four southern states of Maharashtra, Andhra Pradesh, Tamil Nadu, and Karnataka, the HIV epidemic has become “generalized,” with HIV prevalence consistently exceeding 1% among antenatal women. There has thus been an impetus to redirect prevention programs accordingly, from a focus on targeted interventions in high-risk populations (recommended in low-level or concentrated epidemics) to a more general population-oriented approach. However, such a reorientation of prevention programs to “general populations” is based on unsupported assumptions about the nature of sexual networks and transmission dynamics in India. Furthermore, there is tremendous regional diversity in the social and cultural characteristics of populations in India, which is likely reflected in variations in sexual behaviors and networks. It would also be virtually impossible to provide resources for prevention programs in all rural communities, given the size of the population and the number of communities to be targeted. For these reasons, there is a strong need to characterize the variations in HIV risk across villages.
The challenge of our study was to create a tool that would translate community-level parameters into variables that could be measured in resource-poor settings and that would provide valid indicators for phase-specific intervention programming (Weir Morroni, Coetzee, Spencer, & Boerma, 2002) and empirical inquiries. There are few examples of this type of methodological development in either the developed or developing world. One such experience was the development of a household livelihood security index in Bastar, India, that was used for community and family assessments and social program design (Lindenberg, 2002). Specific to HIV/AIDS, a rapid assessment method to identify geographic areas for focusing HIV prevention efforts in Cape Town, South Africa, was pilot-tested with promising results for focusing interventions in urban areas (Weir, Morroni, Coetzee, Spencer, & Boerma, 2001). Of particular relevance is an HIV impact assessment tool created to account for the potential impact that a given development project might have on the spread of HIV in Asian countries (Regional Bureau for Asia, 2000). Nonetheless, there is a pressing need for a standardized tool that responds to a number of assumptions underlying the implementation of focused HIV intervention strategies. These assumptions are that village-level indicators can define high risk or vulnerability to HIV/AIDS, communities are not all equally “at risk,” and it should be possible to define these risk environment factors and develop a method for measuring them. The ultimate goal is to be able to identify high-risk rural villages to focus interventions in the early stages of an HIV epidemic (Gelmon, 2004). The objective of our study was to develop a rapid assessment tool to detect villages at higher risk of HIV/AIDS so that these communities could then be prioritized for prevention programming. The tool could also provide reliable data for evaluation and research purposes.
Rapid ethnographic assessment study
To develop the village-level rapid assessment tool (VRAT), we used a rapid ethnographic method in two interrelated steps. The data from the rapid ethnography provided evidence for the development of a village-level HIV/AIDS risk environment framework. The data were also used to construct the VRAT itself. Although not reported in detail here, toward the end of our study, the VRAT was field-tested and refined.
The research team initially developed a tentative framework identifying risk environment domains and variables using existing data and experience from intervention programs in hundreds of villages under the Karnataka Health Promotion Trust (KHPT). Prior to the commencement of the study, the research team studied KHPT documentation that had identified potential variables. Furthermore, a workshop with KHPT staff (several of which were also part of the research team) was held to discuss this information and develop an initial framework. The researchers then used a rapid ethnographic method to assess these domains and variables with three objectives: to reformulate these variables and domains and identify further ones, to determine their meaningfulness in characterizing villages as risk environments and assess the plausible connection of village-level variables with HIV/AIDS transmission, and to identify data sources and measures for the variables if used in a rapid assessment tool. This stage of the study was conducted in 10 villages ranging in population size from 522 to 12,135 (average population size: 4,800), from three Karnataka subdistricts (talukas) in Bagalkot district, in the north of the state. The choice of villages was based on findings from a community-based HIV prevalence study conducted in these villages by our group in 2003 (Becker et al., 2006). The HIV prevalence rates of the 10 villages were 8.2%, 6.5%, 6.3%, 3.4%, 3.1%, 2.8%,1.5%, 1.3%, 1%, and 0%.
Before we commenced the fieldwork, the research team visited all of the villages so that we could explain the study to the village governing bodies (Panchayats) and ask for their permission and support. All Panchayats agreed, although in some cases several visits were required to ensure that they were comfortable with the study and that they agreed to its process. Furthermore, most Panchayats provided logistical support to the study (e.g., finding a place for the fieldworkers to rent while in the village). Ethics approval for the study was granted by the research ethics committees of St. John's Medical College (Bangalore, India) and of the University of Manitoba.
The main investigators and the project coordinating team (PCT) selected and hired 20 ethnographic research assistants (ERAs). These individuals were all from the region of study and had university-level education. The ERAs participated in a 3-day training workshop at the study headquarters in Dharwad. The workshop was followed by demonstration and practical training in several nonstudy villages several days prior to commencing the fieldwork. The ERAs stayed in pairs (a female and male) at one village for 15 straight days (they lived in the villages during that time). The fieldwork in villages with populations of 3,000 or less was done by one pair of ERAs, the fieldwork in villages with a population of between 3,000 and 6,000 was done by two pairs, and in villages with more than 6,000 it was done by three pairs of ERAs. The ERAs were provided with field guides that included instructions for using a diary, an observation and mapping information items guide, a key informant interview guide, and a focus group guide. Data from secondary sources were gathered by the PCT.
One source of information was the diary of each ERA, where they kept track of their observations and reflections about the village. The purpose was to keep a record of what each ERA observed during the day, who they talked with, and what they did as well as their impressions about the life of people in the village. They described the village day by day. The main guideline provided to the ERAs for the diary was to gather information about what characteristics of villages could protect villagers from HIV/AIDS transmission or put them at higher risk. Nonetheless, the diary technique was to be purposely kept open ended to allow the ERAs to identify characteristics of the villages not yet considered that could be related to risk of HIV transmission. The diaries were read by the PCT every several days, providing information that assisted them in redirecting if necessary some of the aspects of the village research and to address concerns or difficulties that could arise.
A second source of data was a semistructured observation and mapping guide, used by the ERAs. This guide had two components. The first was an “information items listing” containing 45 structured items. The ERAs gathered information for each item and answered the questions indicated as best they could. This information was to be gathered through direct observation or by what the ERAs heard from the villagers. Information contradicting a previous entry was included without deleting any of the entries. The guide provided sufficient space for various entries, and space was also provided for the ERAs to add items or information not identified in the guide. The guide also included explicit mapping instructions showing aspects such as village geographical distribution by caste, places of gathering, temples, and so on. The second component of the observation and mapping guide was a “sources” section. The ERAs kept track of how they had obtained that particular piece of information for each of the 45 items. They recorded the type of source (one or several young male villagers, by direct observation, etc.) as well as when and where they obtained it. If the source was a written one (e.g., local newspaper article) they kept the document if possible and submitted it.
Key informant interviews provided another source of data. Individuals from different sectors and social groups of the village knowledgeable about certain aspects of the village or of village life were interviewed. A balance between men and women, different age groups, and individuals of different positions was attempted. A sample matrix with a list of ideas was provided to the ERAs (e.g., village leaders, religious leaders, commercial sex workers, health workers, school teachers, women leaders, youth leaders). In these in-depth interviews with key informants researchers did not use structured questions, although ERAs were provided with ideas for guiding questions. The total number of interviews was 221, ranging from 14 in the smallest village to 40 in the largest village (an average of 22 per village). The interviews were conducted by the pair of ERAs. Most interviews were taped and subsequently transcribed. When this was not the case, extensive notes were taken.
The ERAs also conducted 96 focus groups in the 10 villages, ranging from 5 in the smallest village to 20 in the largest village (an average of almost 10 focus groups per village). Focus groups were conducted with different male and female groups such as agricultural laborers, drivers, nonagricultural laborers, married lower caste women, young men, and so on. The focus groups were conducted by the pair of ERAs. Again, most group discussions were taped and subsequently transcribed, and when this was not the case, extensive notes were taken. The topics varied in part according to the groups. For instance they addressed issues such as practice of child marriage, reasons for migration, caste relations within the village, entertainment activities of villagers, and so on.
The PCT was responsible for collecting secondary source data for each village, from sources such as local NGOs, community based organizations, village Panchayat administrations, India census department, and so on. Examples of data were population patterns, social stratification systems (caste), land holdings, main agricultural products, and changes in economic situation.
The field data were all transcribed and translated from the local language (Kannada) into English. The data were analyzed by the PCT and the project investigators, with support of some ERAs.
Several data analysis workshops were held by the research team. In addition, a few months after the fieldwork was completed, a one-day meeting (including a celebratory meal) was organized in a regional city with approximately 20 Panchayat leaders from the 10 villages (2 from each village). One purpose of the workshop was to share with the village leaders some of the findings identified by the rapid ethnographies and to listen to their feedback, and another was to thank them for agreeing to have their villages be a part of the study and for their support to the project.
The coding and summarization of qualitative data for triangulation was done manually using Excel spreadsheets. The method employed was comparative contextual analysis, a method for comparative research in which contextual interrogation precedes any analysis of similarities and differences (Henham & Findlay, 2001). Contextual analyses typically address the effects on individual behavior or attitudes vis-à-vis contexts (geographic, socioeconomic, social environment). The observation guide data were analyzed horizontally (item by item across villages). The diaries, interviews and focus group data were analyzed first vertically (ERA by ERA) and then horizontally (identified themes compared across villages). Secondary data sources were comparatively analyzed across the 10 villages. The recursive data analysis operated at two levels. First, by testing and reformulating the relationships of the initial model, the researchers refined the final domain variables and their plausible associative mechanism as causes or markers of HIV risk. Second, the analyses identified potentially meaningful measures of these variables and their units of data and data sources.
A refined version of the domains and variables framework (Table 1) and a pilot version of the VRAT were developed. For variables to be included in the final framework, the evidence had to have identified the plausible causal mechanisms at play or have established a explicit rationale for the variable to be a possible marker of risk. The PCT then field-tested the tool in 10 villages (different from the original 10 villages), and final revisions were completed.
Conceptual framework: Village-level risk environment domains and variables
Note: NGO = nongovernmental organization; CBO = community-based organization.
Village-level rapid assessment tool
The main outcome of the rapid ethnographic study was the development of the village-level rapid assessment tool (VRAT), which consists of a comprehensive document detailing all its aspects, a data collection workbook for field research assistants (FRAs), a workbook for FRA supervisors, a booklet for the project coordinating team, and a database for data input and scale conversion. The VRAT requires that one team of FRAs (a male and female pair) visit a village with a population of 5,000 or less for 2 full days, or one FRA team for 3 full days for villages with a population of more than 5,000. Based on the field-testing, the tool appears to be more appropriate for medium and small villages. For villages larger than 8,000 people, the tool might become problematic in terms of the reliability and validity of the data.
The variables in the risk-environment framework are considered potential markers of village-level risk for HIV infection, not necessarily causes. Several assumptions underlie these markers: first, that they are attributes of villages, not of individuals; second, that the villages have discrete boundary delimitations; and third, that as community variables they would be stable for approximately 5 years (i.e., trait measures). The risk environment framework developed by the rapid ethnographic study guided the development of the VRAT. The measures were constructed based on clear conceptual distinctions and were assessed in terms of their meaningfulness as village level variables (for an example of a variable see Table 2). The study's ethnographic evidence within and across the villages had to support this plausible linkage to constitute a domain indicator within the risk environment framework. Furthermore, this evidence was the basis for assigning an initial score to a particular variable. Nonetheless, there was not enough evidence to assign weights to the variables. Finally, for the variable to remain in the tool, the field test had to demonstrate that comparable and reliable data at the requisite area level were possible to be collected within a short time period in all villages. The measurement model is a “formative” indicator model (also known as causal or composite), meaning that the indicators determine the domains. The tool results in a standardized composite index of village-level indicators. A validation study of the VRAT is being planned in approximately 250 villages from a sampling frame of 7,500 villages in 15 districts to assess its reliability and predictive validity.
Example of VRAT variable
Analysis and discussion
The goal of our study was to better understand village dynamics, their plausible connection with HIV transmission (as causes or markers), and to develop a framework and a rapid assessment tool. We called our study design rapid ethnographic assessment. Let us examine the validity of this terminology.
Comparisons between the two study designs
Rapid Ethnographic Assessment Study
Village-level Rapid Assessment Tool
Although our study commenced with a tentative framework of potential variables related to village dynamics and HIV, our programmatic experience indicates that the study was purposefully designed to challenge these assumptions and discover new or different relations. A number of traditional ethnographic methods were used. The immersion of the ERAs in the villages (albeit for only 2 weeks) and the use of diaries as data-gathering technique can be understood as such. Furthermore, mapping and observation, in-depth interviews and focus groups, were the other central sources of data.
The analysis of this information was essentially of an emic nature. An emic approach consists of primary methods such as
interviewing, in-depth and in the local language; intent to seek the categories of meaning, as nearly as possible in the ways the locals define things; the people's definitions of meaning, their idea systems, are seen as the most important causes or explanations of behavior; the methodology begins with particular observations and uses them as a foundation for understanding. (Pelto & Pelto, 1978, p. 62)
In this sense, our use of diaries and, to a large extent, the other sources had this intent. Even the most structured data technique (the observation guide) intentionally directed the ERAs to capture information outside of the categories suggested by the guide itself. Nonetheless, it is here where our study also can be seen as using an etic approach. The observation guide was developed based on our preliminary notions of plausible variables of interest. If etic is related to the use of the researcher's categories, then our study also incorporated etic components. In a sense, we can talk of having used an iteration of emic and etic approaches consistent with our recursive analyses. The etic and emic distinction parallels that of inductive and deductive or positivist and naturalist paradigms. Our study used ethnographic approaches to develop categories, and in that way was using an inductive (or emic) paradigm. However, the outcome of the study was to obtain a tool closer to a deductive (or etic) paradigm.
Continuing to examine our methods, an argument can be made that our “ethnography” was limited in a number of ways. One is the short time (as compared to traditional ethnography) spent by the ERAs in the villages. Furthermore, the use of teams of researchers (ERAs) rather than one ethnographer doing all of the data collection, analysis, and interpretation can challenge our study's being considered ethnographic. Finally, the abovementioned iteration between emic and etic approaches also puts this into question. On the other hand, our study clearly shared many traditional ethnographic methods and approaches, as explained above. Nonetheless, some authors prefer to use the term qualitative inquiry rather than ethnography to denote this rapid methodology, “out of respect for those who have helped define ethnography and argue it always requires prolonged fieldwork” (Beebe, 2001, p. xv). However respectful this change of denomination, it begs the question as to whether these rapid designs are essentially qualitative. In line with Handwerker's (2001) thinking, ethnography itself should not be confused with qualitative methods. Beebe's terminology appears to perpetuate a dualism that is not reflected in the reality of rapid designs. The essential difference is not in the qualitative or quantitative data but, as mentioned above, in the etic or emic analytical strategies.
In our approach, qualitative data were central to our efforts and analyses. Nonetheless, we used quantitative information as well, mostly but not exclusively from secondary sources. It would seem that defining our work as a qualitative inquiry would miss out the potential value of no qualitative information. The notion of ethnography appears to capture better the idea of holistic inquiries, and following Miles and Huberman (1994), that its task is to “reach across multiple data sources (including but not limited to observations, interviews, artifacts, and documents), and to condense them” (p. 8). If so, it would seem more valid to consider these type of inquires as ethnographies, even if they are rapid. Simultaneously, the rapid assessment tool that our study created (VRAT) offers a clear example of a study design that in our opinion should not be considered ethnographic, and for good reasons.
As noted, the main distinction is not between qualitative and quantitative data collection but with the analytical approach of the study; that is, emic or etic. Our study was essentially emic (rapid ethnography) in approach, but the tool created by our study is essentially etic (rapid assessment tool). Despite the fact that the latter used mostly qualitative information and that the former produced largely quantitative data, both should work with this distinction as a continuum, not as categorical. The VRAT is etic because it is a structured tool with a set of variables for which data need to be collected. Furthermore, these data are transformed into scores that will then be analyzed using statistical techniques. In fact, part of the data collected by the VRAT is initially qualitative, but the analytical approach is deductive, not inductive. This is the main basis to not consider this approach ethnographic, which brings us back to the common issue of “rapid” and “quick.”
As compared to other study designs, both our ethnographic study and the VRAT are quite intense and rapid approaches. In that sense, both can be understood as rapid designs, one essentially a rapid ethnography, the other a rapid assessment. In relative terms, 15 days is “rapid” as ethnographies go, and two days is “rapid” in comparison to many structured studies. The particular advantage of this “quickness” is the intense fieldwork and the momentum of the work. This can be particularly advantageous in several aspects. From a resource perspective, it may be a more efficient use of human and logistical resources, and the focus of the study may be easier to sustain. In cases with community involvement or studies more linked to action, intense fieldwork can better articulate the “study” portion of the inquiry with the “intervention” portion, particularly when community mobilization is necessary. It can also minimize research fatigue from both the research team and the community's perspective. Our study suggested that intense work for a relatively short period of time paid off with rich data, while at the same time not overstaying our welcome in the villages. Ultimately, the above considerations relate to the appropriateness of study designs to the purposes of the study.
Beebe (2001) considers rapid designs appropriate when the topic needs to be explored (e.g., when there is insufficient information available to identify variables), when the question begins with “how” or “what,” when there is a need for a detailed view, when there is a need to study individuals in their natural settings, and when there is a need to emphasize the researcher's roles as a partner instead of as an expert. Our research objectives matched well with most of the conditions outlined by Beebe. Nonetheless, the last condition relates to the notion of participatory research, which requires examination.
A number of rapid designs have been defined as participatory. For instance, a main development in this area was that of participatory rural appraisal, mainly systematized by Chambers (1994a, 1994b, 1994c). There has been a tendency to automatically classify rapid enquiries as participatory, in large part because of the idea that they “listen to people” and also because many such studies use the emic approach. Unfortunately, this seems to misunderstand the real nature of participation, assigning to certain methodologies innate participatory qualities. At its best this is a mistake, at its worst a misrepresentation.
Let us return to our study to examine this point. As part of a large intervention program on HIV/AIDS, a particular need was identified, and consequently specific study goals were defined. Given that accurate data on HIV infection rates existed on 10 villages, these were selected to be part of the study. To this point, we do not observe any participatory involvement from the villages. Because of ethical standards and practical reasons, we approached the Panchayats, explained the study purposes, asked for their permission to conduct the study in their villages, and requested some basic assistance in its implementation. The rapid ethnographies did not commence until all approvals were formally obtained. At all times during the study, ongoing dialogue with the village leaders was maintained. Further, sometime after the fieldwork was completed, basic findings from the study were shared with Panchayats leaders, their responses to the findings were heard, and a formal acknowledgement for their support was provided with a celebratory meal. In a superficial sense we could talk of “participation” of the villages in the study. Nevertheless, for a number of reasons it would be misguided to consider our study as participatory.
From our point of view, true participation requires taking part in making the major decisions (Mignone, 2005). The “you participate, I decide” mode of doing participatory research is disingenuous. Our ethnographic study, despite satisfying the ethical requirements of community permission, as well as counting with some level of village participation in knowledge about the study and assistance, was not participatory. The major decisions (of study goals, of methodology, of use of findings) were not made by the villages nor did they have a role in these decisions. They were participants in the study because the Panchayats generously agreed that their villages be “studied,” but that does not imply participatory research. Rapid ethnography or rapid designs of any type are not innately participatory. As with any research design, it might or might not be participatory research based on who are the main decision makers and users of the information.
Aside from potential limitations and weaknesses of rapid designs common to most research strategies, there is one of particular concern. Beebe (2001) has pointed out that “despite the large number of rapid research reports that have been done in the last several decades, there has been very little criticism of methods by practitioners” (p. 108). He reiterated Chambers's lament that calling research methods rapid “has been used to justify and legitimize sloppy, biased, rushed, and non-self critical work” (p. 108) Rapid designs require the same methodological rigors of other research strategies. Further, the appropriateness of rapid designs to particular studies is a key consideration related to the validity of this approach. We finalize this analysis by examining the appropriateness of our study to community and public health research in general, and to the prevention of HIV transmission in particular.
Health researchers have criticized public health research for the lack of explicit theories in their studies (Krieger & Zierler, 1997). Empirically driven research can provide interesting data but not necessarily information and knowledge. Worse, not formulating the theoretical underpinnings of the studies blinds them to their conceptual biases. Among one of the central debates about social science is whether its mandate is to achieve universal knowledge or should it be mostly context specific. This relates to the longstanding dispute of whether the natural sciences model is applicable to social sciences, and in words of Flyvbjerg (2001) “is theory possible in social science?” (p. 25). He is asking if we can speak of theory in the study of social phenomena in the same sense that theory is used in the natural sciences; that is, if the key concepts of explanation and prediction can be based on context-independent theories, as they are in natural science. Bourdieu (1992) suggested that context cannot be excluded because context in social science defines the type of phenomenon which theory encompasses. If a theory must necessarily be context free to be a general theory, then this presents a dilemma for social science.
In a certain sense, our study was seeking to formulate a theory of village-level HIV risk environment, based on which a predictive tool would be developed for preventive interventions. Prior experience and literature provided us with a tentative model. Nonetheless, the rapid ethnographic study enabled us to formulate a model grounded in village dynamics. The emic approach and recursive analyses allowed us to identify variables potentially associated as causes or markers of village-level risk. The ethnographic approach was central to the understanding of context. The rapid assessment tool developed by the study seeks more generalizability in its application, although it should not be considered context independent in two senses. First is the study design, because its application requires a careful understanding of the sociocultural norms of the particular regions where it will be used. Our findings suggested the crucial importance of establishing a formal process of engagement with village leaders, research teams integrated by men and women of the region, being properly informed of caste issues, political dynamics, and so on. These aspects are expected to vary according to regions or countries, and designing rapid studies without rigorous consideration of the particular context may not only compromise the study but, worse still, might negatively affect the villagers. Second, although its etic approach speaks of a level of universality and the validation will seek to make it applicable to all Karnataka villages (stratified based on the seven agro-climatic regions of the state as well as by village population size), it will be at that contextual level that it may be generalizable. There is nothing against the extrapolation of many of its attributes to other states in India and potentially to other countries, but that would require grounded reformulations both of the framework itself and the tool.
The strength of the rapid ethnographic design was that the model that guided the development of the village-level HIV risk assessment tool was formed iteratively “from the ground up.” It enabled the study to capture numerous village dynamics, unforeseen aspects, and plausible social environmental pathways of HIV transmission. This led to the reformulation of the initial model, the identification of variables of potential interest, and the development of meaningful measures. This study design allowed for a more valid approach to formulating a contextually appropriate conceptual framework, rather than a generic theoretical based approach or a data-driven nontheoretical approach. Nonetheless, further studies are needed to validate the VRAT. In a study soon to follow, we will pilot-test the VRAT in 250 villages from the seven agro climatic regions of the state of Karnataka to empirically refine the scales of the tool and to assess its predictive validity. The use of the tool in other Indian states, as well as other countries, might require new rapid ethnographic studies to properly adapt the VRAT to other contexts.
Footnotes
1.
“Risk environment is an environment in which the chances of disease transmission are increased as a result of social, economic, and cultural factors” (Barnett & Whiteside, 2002).