Why don’t segregated Roma do more for their health? An explanatory framework from an ethnographic study in Slovakia

We used structural-constructivist relational theories of human action as the conceptual framework of our study (Archer 2000; Bourdieu 1998). According to these theories, everyday practices are driven mostly by the actors’ practical reasoning, which is spatiotemporally contingent, partially implicit and subconscious, and significantly shaped by historically evolving structures and social constructions. The structures represent the environmental, social and bodily conditions in which the selected actors operate. The social constructions represent how the actors interpret these conditions. Actors acquire their specific practical reasoning gradually, through the process of socialization. In this process, the actors’ bodies, inner drives, motivations and interpretative repertoires become practically attuned to their specific conditions. Both acting according to any reasoning in practice and specific socialization patterns continue to depend on enabling environments. What specific structures support which specific practices and how they do so can be best examined by exploring related social constructions that the specific actors use, i.e. their related practical reasoning and their socialization in such reasoning.

This study was part of a larger longitudinal study exploring the social root causes of poor health status of segregated CEE Roma through the case of a segregated settlement in Slovakia. The larger study spanned 2004–2014 and consisted of four methodologically distinct phases combining ethnography (Reeves et al. 2008) and systematic medical-anthropological interviewing (Hausmann-Muela et al. 2003): (1) a socio-graphic survey, aimed at the selection of a single segregated place; (2) ethnography, aimed at gaining close personal access to and extensive primary data regarding the local everyday health-endangering settings and practices; (3) systematic interviewing, aimed at increasing local representativeness of the collected material; and (4) follow-up communication, aimed at obtaining locals’ reflections on preliminary interpretations and obtaining additional data on long-term outcomes. All fieldwork was carried out by the first author. All aspects of the larger study methodology, relevant also for this study, have been reported in more detail elsewhere (Belak et al. 2017). This regards a description of the setting of the Roma in Slovakia, our procedure for selection of the locality and informants, the characteristics of the selected settlement, the observation and elicitation procedures that we used in all phases of the study, our initial coding of the study data, the first author’s embeddedness in the settlement and the study’s potential biases.

In the terms of the theories outlined above, the primary focus of this study was on what local structures supported the actors’ eventual nonadherence to clinical and public health recommendations over the long term (cf. Frohlich et al. 2001; Singer et al. 2016). Drawing on the theories, we started with an exploration of the reasoning of local Roma regarding nonadherence (cf. Cockerham 2005; Frohlich et al. 2001). Then, we explored what experiences contributed to the adoption of such reasoning (cf. Singh-Manoux and Marmot 2005; Williams 1995). Finally, we explored local-level mechanisms that supported both the adoption of such reasoning and the everyday practice of pro-nonadherence reasoning, i.e. the local structures that systematically enabled such adoption and such practices (Hedström and Ylikoski 2010).

In phase 1 (July 2004), we selected a single segregated settlement with a growing population of approximately 260 people (230 in 2004, 300 in 2014) on the outskirt of a village with a declining non-Roma population of about 530 (580 in 2004, 470 in 2014). In 2004, approximately half of the Roma settlement’s inhabitants were children under age 15 years, and only 5 people were older than 60. In phase 2 (September 2004–October 2005), we obtained data on approximately 90 people belonging to one of the three then largest extended families in the settlement. In phase 3 (October 2005), we visited a sample of 10 of the settlement’s 48 households. The sample was representative according to the households’ social ranking (low, medium and high), based on the combination of ascribed affluence and prestige, and kinship affiliations (to the three largest extended families). In these households, we interviewed 28 people, 22 of them adult women. Locally, men were considered less competent regarding health-related issues both by themselves and by women, and most of them also showed less interest in discussing health spontaneously. None of the people approached refused to participate in the interviews. Phase 4 follow-up observations and elicitations (November 2005–November 2014) were limited to approximately 15 Roma personally closest to the first author.

The data consisted of field notes on direct observations and written and audio records of elicitations obtained during phases 2–4. We collected observations, spontaneous declarations and replies in elicitations regarding: why individual Roma did not adhere to selected clinical and public health recommendations—as data on reasoning for nonadherence; what experiences individual Roma considered important for their adoption of such pro-nonadherence reasoning—as data on the adoption of pro-nonadherence reasoning; and how and what local circumstances supported the recurrence of such contributing experiences and nonadherence practices—as data on local-level supporting mechanisms.

To gain data specifically and exhaustively on contemporary clinical and public health recommendations, an encyclopaedic practitioner’s handbook covering both clinical and public health knowledge and recommendations was used throughout all phases of the study to guide observations and elicitations in terms of topics (Sasinka et al. 2003).

For coding the data we re-used transcripts from previous analyses (Belak et al. 2017). In these transcripts, all field notes and audio recordings relevant regarding health-related settings and practices were already merged and coded for relevance in relation to household social levels, kinship affiliations, genders, ages, time periods, and domains of exposures and their core elements, as defined in a widely used eco-social framework on social determinants of health (WHO 2010). The first author then in steps added new axial codes signifying “pro-nonadherence reasoning”, “experiences contributing” to the adoption of such reasoning and local-level “mechanisms”.

Next, we performed a qualitative content analysis using recurrent abstraction (LeCompte and Schensul 2013). We repeatedly read and in steps summarized all text sequences coded as relating to pro-nonadherence reasoning, experiences contributing to the adoption of such reasoning, and local-level mechanisms. We focused on capturing local variability and the most salient kinds, as follows. First, we summarized sequences on local reasoning, yielding 13 kinds of reasoning. We have reported these in Appendix 1 in electronic supplementary material. Second, we summarized sequences of experiences contributing to the adoption of pro-nonadherence reasoning, yielding four such kinds. We have reported these in Appendix 2 in electronic supplementary material. Then, we summarized sequences on local mechanisms, yielding seven mechanisms in total. We report these mechanisms in our results below. During this analysis, we realized that some of the mechanisms were controlled by and operating through Roma actors in the settlement, while others more by local non-Roma actors outside the settlement. We report according to this distinction, as it informs on which actors need to be prioritized in interventions and regarding what mechanisms.

During the analyses, we also realized that the identified mechanisms supported each other. Given that most of the identified individual mechanisms have been previously described with respect to CEE Roma (see “Introduction” section), we focused mainly on reporting their mutual interrelations. For more information and illustrative examples, we refer the reader to Appendices 1, 2 and 3 in electronic supplementary material and to our previously published detailed descriptive report on other parts of the same study (Belak et al. 2017).

The identified mechanisms formed an interlocked system, as schematically summarized in Fig. 1.

The main strength of our study is its sociologically well-theorized and applied approach. The use of ethnographic methods enabled intimate access to local everyday settings, local people and their practical reasoning. The systematic interviewing across several local stratifications allowed for local representativeness as well as for topical omissions in the previous less-systematic phase to be considered. The follow-up communication enabled additional reflections of preliminary interpretations by local core informants. Choosing structural-constructivist sociological theories and the WHO framework on social determinants of health (WHO 2010) strengthens both the sociological and public health significance of our results.

Our research design and our reporting also had some limitations. First, the fieldwork, coding and most of the analyses were performed by a single researcher, limiting the potential for inter-personal corroboration. However, this is standard in ethnographic research, given the logistic difficulties connected with embedded research. Second, the researcher conducting the fieldwork was a male, which may have influenced the reporting by women due to existing local gender power-imbalances. This may have resulted in, e.g. underreporting of gender-sensitive issues. However, we think that this bias is rather limited, as the first author also experienced numerous intimate conversations with local women across generations which included strong criticisms of local male and female roles. Third, it was impossible to remain personally embedded in the settlement during the full study-period. Nevertheless, throughout the research period the first author did stay personally very close to members of one of the three major local extended families which he kept visiting. Fourth, given the author’s embeddedness within the Roma settlement, data on local non-Roma views and practices were much more mediated compared to the data on local Roma. However, all assertions implying non-Roma social norms or practices were still based only on real and documented cases. Despite all the reassuring circumstances just mentioned and the above discussed general good match of our findings with other published research, the presented framework cannot be generalized.

Future research and interventions aiming at behavioural changes of segregated Roma towards adherence to biomedical recommendations should include discussions with the Roma about whether some of their identity-related preferences for nonadherence might not present a case of symbolic violence (Bourdieu 2000), i.e. values historically imposed on their communities by powerful non-Roma. Second, interventions need to address the local anti-Roma prejudices and malpractices of local non-Roma, as well as the influences of more distal actors and processes that support such local mechanisms (e.g. poor education and media coverage in CEE regarding racism and Roma history and conditions). Third, the research and interventions should always carefully examine the health-related outcomes and social side effects of eventual alternative Roma care practices.

Segregated Roma might be doing less for their health due to interlocked systems of local-level mechanisms, controlled by and operating through both local Roma and non-Roma. Racist non-Roma ideologies, internalized by Roma into a racialized ethnic identity through socialization, and drawbacks in adherence might represent powerful, yet neglected, mechanisms supporting segregated Roma nonadherence.