There are many fevers: Communities’ perception and management of Febrile illness and its relationship with human animal interactions in South-Western Uganda

Diagnosing the causative agent of febrile illness in resource-limited countries is a challenge in part due to lack of adequate diagnostic infrastructure to confirm cause of infection. Most febrile illnesses (>60%) are non-malarial, with a significant proportion being zoonotic and likely from animal origins. To better characterize the pathways for zoonotic disease transmission and control in vulnerable communities, adequate information on the communities’ experiences and lexicon describing fever, and their understanding and perceptions of risk pathways is required. We undertook an ethnographic study to understand behaviors, exposures, and attitudes toward fever at the community level. Our hope is to better elucidate areas of priority surveillance and diagnostic investment. A focused ethnography consisting of participant observation, informal conversations, 4 barazas (community meetings), and formal ethnographic interviews (13 Focus group discussions and 17 Key informant interviews) was conducted between April and November 2015 in Kasese and Hoima Districts in Uganda. Perception of illness and associated risk factors was heavily influenced by the predominant livelihood activity of the community. The term “fever” referred to multiple temperature elevating disease processes, recognized as distinct pathological occurrences. However, malaria was the illness often cited, treated, or diagnosed both at the health facilities and through self-diagnosis and treatment. As expected, fever is as an important health challenge affecting all ages. Recognition of malarial fever was consistent with a biomedical model of disease while non-malarial fevers were interpreted mainly through ethno etiological models of explanation. These models are currently being used to inform education and prevention strategies and treatment regimens toward the goal of improving patients’ outcomes and confidence in the health system. Development of treatment algorithms that consider social, cultural, and economic contexts, especially where human-animal interaction is prevalent, should factor animal exposure and zoonotic illnesses as important differentials. Ethical approvals were obtained from the University of Minnesota Institutional Review Board (IRB) study number 1502S64201, the Joint Clinical Research Centre (JCRC) IRB in Uganda and the Uganda National Council of Science and Technology (UNCST HS 1726). All tools and approaches used were reviewed and approved by UMN and JCRC IRB boards. Informed consent was obtained from all key informants and focus group participants. The study was conducted in two districts (Kasese and Hoima) in western Uganda which are located within the Albertine Rift Valley, an approximately 500-kilometer-long geo-formation that is part of the northern sector of the Western Rift. It stretches from the Kivu volcanic province in the south (1°S) to the border between Uganda and South Sudan in the north (3°N), straddled between latitudes 29°E and 31°E with an estimated area of 14,098.9 km2 [46,47]. The districts are characterized by both pastoral and agro-pastoral livelihood systems. The Albertine Rift Valley has rich biodiversity, high livestock density, a tropical climate with highly fragmented forest areas and rapidly growing urban centers. Kasese has a population rate of over 6.4% and Hoima has 4.3%, making both among the fast growing and highly populated districts (Hoima population 572,986; Kasese population 702,029), in Uganda [48,49]. Kasese’s major pastoralist groups are the Basongora and Bahima who own majority of the 110,000 cattle in the district. While in Hoima the Bahima own most of the district’s 140,000 cattle. Both districts also have large agro-pastoralist populations which include the Bakhonzo, Batooro in Kasese and, Bakiga, Banyoro and Lugbara in Hoima. The pastoralists inhabit the low-lying regions of Hoima and Kasese and are characterized by high reliance on livestock for economic and social wellbeing. These pastoralist areas are often semi-humid to semi-arid regions (annual rainfall ≤1000mm) and thus pastoralists employ various types of targeted mobility to access adequate pasture and water for their livestock. The vegetation cover is made of bushy-acacia trees, shrubs, and sparse grassland. Majority of cattle kept are Ankole cattle, a hybrid between Zebu (Bos-taurus indicus) and long-horn cattle (Bos-taurus taurus). Apart from Ankole cattle, few Boran and Zebu cattle are also kept with different herds mixing freely, and occasionally interacting with wild ruminants [50]. Families live in closely knit communities where all family members have roles and their interaction within this larger network of community members is governed by set of cultural rules and expectations that are adhered to closely. Most of the areas they inhabit are often remote, have limited infrastructure and border large water bodies or wildlife reserves. A lot of their former communal grazing land has been transformed into wildlife protection areas by government policies. This has led to increased human wildlife interaction and conflict, and a sense of marginalization [51]. Housing is often temporary to semi-permanent in nature (S3 Fig), the area sparsely populated apart from centroids of habitation next to water bodies or trading centers and access to healthcare or education is limited. In addition to livestock production, some pastoralists engage in limited auxiliary activities such as crop production or firewood collection to help complement their pastoral proceeds [51,52]. The agro-pastoralists live in areas that are higher in altitude; receive more rainfall (>1000mm) and have lower daily mean temperatures (18°C—30°C). The vegetation is savannah grassland interspersed with bushy shrubs and sometimes fragmented forests. The area is densely populated with better access to schools and health care centers (compared to pastoralist regions), most farmers practice mixed crop-livestock farming for both subsistence and commercial purposes. Increasingly, these farmers are adopting intensive production methods such as zero-grazing (stall-feeding system) and paddocking systems, while improving their indigenous breeds by crossing them with high yielding dairy breeds such as the Friesian cows. Land ownership is individual, and houses are mostly permanent [53,54]. The lakeshores of both districts are home to fishing communities (e.g., the Langi, Alur, Baganda) and, in some cases, pastoralist communities as well (Banyankole/Bahima, Basongora). The homesteads are often lined up in neat rows adjacent to the lake shore. Most have mud walls and grass thatched roofs with a few brick-walled and tin roofed houses. Amenities are often limited, temperatures high and access to healthcare poor. There is a lot of small scale trading because of the fishing industry, high rate of human-animal interaction because all converge around their need for water, high rate of in and out-migration, significant economic and social exchange between transient occupants of these communities, and as a result higher prevalence of HIV compared to other parts of the country [55]. The soils are sandy and hence cannot sustain crop agriculture well. These sandy soils make construction of toilets (pit latrines) difficult resulting in frequent waterborne disease outbreaks. There are many boats, people often use motorcycles to commute in and out of the village and a few vehicles (most probably belonging to local traders) can also be spotted picking or dropping merchandise. These communities are engaged in limited crop and livestock production for subsistence and small-scale commercial purposes. In addition to agricultural production, Kasese and Hoima also have protected and non-protected forests with a substantial wildlife population. Hoima is also part of a biodiversity hotspot–the Albertine Rift–having one of the largest portions of unprotected forests in Uganda with high rates of deforestation averaging about 2.27% annually, three times higher compared to protected environments [48,56,57]. To appreciate the current challenges faced by the peripheral public health system in these areas, particularly in the diagnosis and management of febrile illness (curative services), it is important to look briefly into the past and understand why this current need exists. Uganda’s health system has undergone tremendous changes over the last 60 years, driven by colonial policies, post-independence political turmoil, and subsequently, health sector reforms. During the 1960’s and 1970’s Uganda had a remarkable health system and one of the best network of health services in the continent, allowing its citizens to enjoy free access to health care services [58]. The political instability of the late 1970’s and 1980’s took a toll on the social, economic and health systems within the country, resulting in physical deterioration of health facilities and massive exodus of trained personnel. During its recovery period in the 80’s and influenced by global reform policies, Uganda invested in a primary health care community based approach, focused on maternal and child health and implemented through vertical programs that had little integration into the health system’s governance structure [59]. This was subsequently followed by the fee-for-service model for all curative services, and reduced government investment into the public health infrastructure particularly that which was located at the lower administrative level [58]. Despite subsequent efforts and polices to try and ameliorate the situation, these global and national polices of the 1980s and 1990s have resulted in persistent disparity in health access particularly among rural communities given the limited diagnostic and curative infrastructure in many of these peripheral public health units. The current health care delivery system mirrors the government administration system with health care delivery centers from the national to the village level, complemented by private-not-for profit facilities, private for profit health facilities and traditional medicine practitioners [59] (S1 Table). In these frontier communities, most individuals seek treatment from a local health center II, or in some cases, health center III, and upon referral, the regional or district health centers if they have the resources required for this. The health center (HC) II is normally staffed by an enrolled nurse, a qualified midwife, and at least two nursing assistants. It is often built and staffed to handle out-patients with common illnesses and support maternal and childcare. The HC III has a bigger pool of professionals and is often led by a senior clinical officer and has a maternity ward and functional laboratory (S1 Table) [60]. Sometimes community members resort to local pharmaceutical shops or traditional healers for quick treatment alternatives as well. Kasese has 46 HCIIIs, 4-HCIVs, and four hospitals while Hoima has 33-HCIIIs, 3-HCIVs and one regional hospital. These health centers often have a shortage of drugs and qualified personnel to cater for the existing need on ground. Accessibility to these formal health services in Kasese is 78% and 94% in Hoima. Some of the barriers that result in this disparity include; geographical barriers, lack of drugs in public facilities, limited qualified staff and long wait times, as well as education and sociocultural barriers especially among the pastoralist community that is larger in Kasese [61]. The recorded leading cause of morbidity and mortality is malaria, followed by respiratory tract infections in both districts [62]. Most patient NMFI management at the peripheral health center level is syndromic given the limited diagnostic infrastructure at this level. Treatment is often guided by WHO IMAI for first level health facilities and focuses on malaria management. Greater diagnostic facilities are available at the district level where management is guided by the WHO IMAI District Clinician Manual. Yet even these resources could benefit from occupational and contextual (sociocultural) patient information to help improve clinicians’ diagnostic and treatment strategies [63]. Ethnography is a discipline that seeks to describe a people’s way of life informed by a deep desire to understand their view points and interpretation of happenings in their environment (proverbial native’s perspective) [41]. It involves extensive data collection from multiple sources. The data sources used in our study were focus group discussions, key informant interviews, participant observations, archival data review (reports, district health information), community meetings and informal conversations. Our focus group meetings involved; i) a group of carefully selected individuals (preferably not known to each other hence selected from different households), ii) a moderator gently guiding the participant’s conversation down the questioning route (using questions and probes from the guide as needed) without dominating the conversation, thus allowing for group interaction that produced data and insights that would not have been accessible if it were not for this carefully set up context that was both permissive yet planned, iii) an observer whose work was to capture the salient non-verbal features of the interactions, and the community setting as well [64]. The in-depth interview targeted individuals who were knowledgeable about the communities’ way of life, health experiences and had some expertise in human, animal, or community health. Carefully selected questions were used to draw from their specialized knowledge and triangulate the FGD data. Participant observation was used to help contrast informants’ narratives with daily practices of the communities. It involved a close follow up on livelihood and cultural practices, human-animal interactions, level of health services available, available infrastructure and its influence on community well-being. By so doing it allowed us to make explicit, the rules, unspoken ideals, norms and values that are critical for the functioning of these communities [41]. Using these streams of data, this study employed a cross-sectional, qualitative, exploratory design and a focused ethnography (emic) approach to describe the syndrome of fever and important biosocial pathways associated with NMFIs. This method was appropriate given our desire to understand communities’ point of view, their vision and interpretation of happenings in their environment, social interactions and behavior around a narrow focus on health [39,65]. Selection of counties experiencing high malarial and high NMFI was guided by the district surveillance (archived) data. Additionally, communities that accounted for diverse livelihood strategies within the selected counties were purposively selected to explore unique community-based understanding of common febrile illness, their causes, their management, and their lived experience. Given our desire to understand the spatial distribution of febrile illness occurrence in the two districts in order to guide selection of counties, surveillance data (archived) was obtained through the District Biostatistician’s office in Hoima and Kasese with permission from the District Health Officer of each district. This was health facility data routinely collected from public and private health facilities within the sub counties in each district and summarized using the District Health Information Software (DHIS2). The data collected included information on key indicators on communicable and non-communicable diseases. Data extraction was limited to communicable (febrile) diseases; Malaria, Tuberculosis, Typhoid fever, Severe Acute Respiratory Infection (SARI), Epidemic prone diseases (meningitis, dysentery, measles, Yellow Fever, VHFs, plague), Neglected tropical diseases (e.g., Schistosomiasis, Leishmaniasis, Lymphatic Filariasis, Onchocerciasis, Sleeping Sickness, Bacterial Zoonoses). All laboratory confirmed cases of Malarial and Non-malarial febrile illnesses among children above 5years old and adults during 2012 and 2013 were summarized and the average tabulated using MS Excel. In Hoima district, 50% of rural sub-counties, especially those located near the lake, had high malaria prevalence, while 75% of those located in peri-urban areas and areas next to natural reserves had high prevalence of NMFI especially the neglected tropical diseases (NTDs). In Kasese district all the sub-counties recorded high levels of malaria and some typhoid infections (S1 and S2 Figs). Data was collected over seven months in twelve sub-counties with evidence of moderate to high malarial and NMFI occurrence in Kasese and Hoima district (S1 and S2 Figs) that were identified using district surveillance data. In each district selection of these sub-counties was done to capture the diverse livelihood practices within each district. Data were collected using multiple methods, including participant observation, informal conversations, baraza (community meetings), focus group discussions (FGD) and key informant interviews (KII). This allowed for robust data collection and extensive cultural context from the researchers’ personal experience living and interacting with members of these communities. The FGD and KIIs guides had open ended questions that were used more as discussion prompts and guides. Development of the guides was based off the One Health principle that posits human, animal, and environmental health are dependent and inextricably linked [66]. We also gleaned from some of the preliminary studies we had conducted in the area, and other relevant research regarding health defining human-animal interactions to help inform the design of the guides (S4 and S5 Texts). The guides were designed to start with a set of “grand tour” descriptive questions to get an idea about the cultural scene followed by “mini tour” questions to narrow down to more specific items, while probing for “native” terms and phrases [32,66,67]. All sessions were recorded to facilitate easier and more accurate transcription. Although not from this community, the first author is of East African origin, has a working knowledge of the Runyoro/Runyankole language. He is a veterinarian who has lived in and has extensive knowledge of these agricultural communities. He therefore uses his knowledge of the region and these communities to facilitate better connection and dialogue with the communities without losing the in-depth appreciation of the cultural context and how it informs the emerging themes. Allowing him to effectively draw upon communities’ interpretation of their lived experience, and to examine how communities’ livelihoods and culture influence risk, perception, and management of febrile illness. As a result, his epistemological assumptions are consistent with that of constructionism. Key informants included human, animal and environmental health professionals and administrative officials from Hoima and Kasese, selected based on their technical knowledge and cultural understanding of these communities individual informed consent was obtained before each KII was conducted (S7 Text). FGDs were conducted at a site normally used for village meetings away from other community activities or distractions and comprised of male and female community members from multiple social strata, cultural backgrounds, occupations, and adult age groups (Table 1). aFocus groups and IDI were conducted in 6 sub-counties in Kasese {Kyarumba, KatweKabatoro, Lake Katwe, Munkunyu, Kahendero(Muhokya), Kasese Central b Focus groups and IDI were conducted in 6 sub-counties in Hoima{Kigorobia, Kiziranfumbi, Bugambe, Bujumbura, Buhimba, Kahoora} Two Community Outreach Meetings (Barazas) were held in Hoima (Kiziranfumbi and Kigorobia) Summary of Kasese participants in grey. Prior to holding FGDs in any community, we first met with local council officials (these are the gate keepers in these communities) and a few community members. We would then proceed to explain the basis of the study and its potential benefits, inclusion criteria for participants and answer questions they had. After which we agreed on a suitable day for the focus group to be held. We then left them with translated explanations of the study and its objectives. This small group of opinion leaders reached out to other community members using a snowball approach and extended our invitation without any form of coercion or undue pressure. Participation was voluntary and recruitment was supported and guided by these local council chairpersons from selected communities. Only willing individuals who were 18 years or older were invited to the focus group. Before each session verbal consent from the group was sought (S8 Text). This included an explanation of the study objectives, any risks/benefits, and a commitment to ensuring confidentiality of all that was discussed. During the discussion no names or personal identifiers were used. Only the moderator was allowed to tape the proceedings and recordings were downloaded unto a secure laptop that was encrypted for extra security. Participants were allowed to excuse themselves from the group before or at any point during the interview if they did not consent to participation. FGDs lasted about 60–90 minutes and were led by one of the team members trained in ethnographic methods and fluent in the local language. Selection of livelihood groups (pastoralist, agro-pastoralists, fishing) was based on their distribution and availability in the two districts. Balance between male and female participants was desired but dependent on the community. Combined sessions were deemed wise by the community leaders to reduce any suspicion among some of the communities targeted. Therefore, interview teams paid particular attention to create space to hear from the women during the interviews. This was particularly necessary in the pastoralist community. Among the agro- pastoralists and some of the fishing villages the women were more vocal than the men. In this case effort was made to get balanced perspectives from the men as well. Data saturation was assessed through evaluation of the FGDs and KIIs for increased repetition of emerging ideas across the groups. Participant observation was conducted during formal interviews and informal interactions with community members to help make more general observations about the communities and glean additional information on the health behaviors and human animal interactions. Observations were made in several settings: 1) fishing villages and fish landing sites, 2) pastoralists’ grazing and cattle watering areas 3) within the homesteads of study participants 4) trading centers 5) health centers 6) at village meetings. First author, SN, CN, CK, LM, and other study staff conducted the observations. A total of 400 hours of observation were completed from January 2015 to August 2015 (SI: Field notes). First author undertook the role of a veterinarian and adopted an observer stance as participant-as-observer because it facilitated an in-depth understanding of the various human livestock interactions and its possible influence on health. Interpreters were research team members who were; fluent in the local languages of Runyoro, Rutooro and Runyankore (these three languages are similar and are all part of a larger cultural group called Runyankitara), trained in ethnographic methods and research ethics involving human subjects. The research team involved in conducting interviews had 2 veterinary technicians, 3 veterinarians, 1 clinical officer, 3 social scientists all with experience in community health and social/behavioral research. The inclusion criteria included a willingness to participate, residence in the study area for more than a year and a minimum age of 18 years and from different households within the village. Permission to access the community was also obtained from the districts’ chief administration officer, the sub-county administrators, and local council leaders at the community level. Given their gatekeeper role these leaders at the community level guided our approach into the community. In some communities the trusted gatekeepers were the local nurses and village health team leaders, and thus they were also consulted and coopted into our community entry teams. Often entry into the community was initiated by having a local village leaders meeting (baraza) where the goals and value of the research were clearly laid out and their collaboration requested. Preliminary community concerns around the issues of health and livestock productivity were also addressed as they emerged. This helped the research team get a sense of the felt human and animal health concerns within these communities. All recordings were independently and directly translated from Runyakitara to English and transcripts from the FGDs and the KIIs compared to the recordings for consistency and accuracy of translation. Initial data analysis and summary was done using an inductive content analysis approach [64]. To achieve this, a coding framework (S2 Text) was developed by MM with CR guidance and input by reading through the transcripts to get a sense of the whole and identifying the ideas (codes) consistently emerging through the transcripts using an inductive in-vivo approach. This first stage of open coding was both data driven and theory (One Health) based [65]. MM and JW reviewed the framework and made improvements to ensure consistency and integrity of the data collected. The analysis involved coding, categorizing of related codes and identification of themes. Regular check-ins with some of the team members involved in the interview process was done to ensure that coding and resulting patterns reflected the experience of the interviewees. The entire transcript was reviewed, and categories relevant to the research question described. Further ethnographic analysis was guided by our research objectives and the “relational theory of meaning” [38]. Spradley’s levels of analysis, domain and taxonomic analysis, were used to identify patterns in the data that make explicit, implicit cultural meanings of febrile illness and their relationships as previously described. This was achieved first by organizing the copious amount of data into categories (domains) and identifying terms that describe these categories (included terms). We identified the domain focus as the different aspects of febrile illness as experienced by community members (the nature, pathways/causes, and management of fever). After which the semantic relationships between these categories were assessed using the domain analysis framework [40]. The following universal semantic relationships; strict inclusion, cause-effect, attribution, were used to unpack the different aspects of the term omuswijja as used by community members [40] (Table 2). Field notes from participant observation were used to contextualize and guide interpretation of our findings (S3 Text). We used different sources of data (FGD, KII, and PO) to complement and clarify the emerging themes and patterns of data. To ensure rigor, regular briefing with CR, JW, KP, LM, DT and SN was done to identify and address potential biases. Additionally, regular reflection on the transcripts and audiotape recordings to ensure all emerging themes were captured and use of analytic memos during the coding process was also done. We used Atlas.ti (Atlas.ti Scientific Software Development Product GmbH version 7.0.82.0), a qualitative analysis software for data management.