Introduction

Typhoid fever remains a substantial cause of morbidity and mortality in many low- and middle-income countries, with an estimated 17.8 million new episodes annually [1]. By 2015, Oceania had fallen behind both Asia and sub-Saharan Africa to become the region with the lowest coverage of improved drinking water and improved sanitation [2, 3]. Pacific island nations including Fiji [4], Nauru [5], and Papua New Guinea [6] report high case counts of typhoid fever and frequent large outbreaks of the disease. Despite this apparent high incidence, studies to investigate sources and modes of transmission of typhoid fever in the Pacific, where unique socio-demographic, behavioral, and environmental conditions may exist, have been rare [4, 7]. A detailed understanding of local risk factors for typhoid fever is necessary to inform non-vaccine control measures. Furthermore, a robust understanding of the epidemiology of typhoid fever in Fiji is needed to inform decisions about the introduction of recently recommended [8] and prequalified typhoid conjugate vaccine.

Typhoid fever is endemic in Fiji with disease occurring among both rural and urban residents [4, 7]. Blood culture-confirmed typhoid fever infections have increased since the 1990s, rising rapidly since 2005 [9]. By 2010, the incidence of typhoid fever in Fiji, identified by passive surveillance, was 52 per 100 000 population [10]. However, given the limited sensitivity of blood cultures, patterns of health seeking behavior among Fijians, and low access to blood cultures services, the actual incidence of typhoid fever in Fiji is likely to exceed this rate [10].

Previous typhoid fever case-control studies in endemic countries have demonstrated considerable variation in major sources and modes of transmission by location [11, 12].This variation underscores the importance of local research, especially in distinct environments such as those that exist in the smaller island states of Oceania. To identify risk factors for typhoid fever relevant to the distinctive island ecology of Oceania, we conducted a case-control study to inform typhoid control efforts in Fiji’s Central Division.

Methods

Results

Of 39,775 blood cultures performed, 279 (0.7%) patients with blood culture-confirmed typhoid fever were identified. No Salmonella Paratyphi A, B, or C were isolated during the study period. Of those blood culture-confirmed typhoid fever cases, 175 (62.7%) were enrolled in the case-control study. Reasons for non-enrolment are shown in Fig 1. A total of 349 controls were enrolled in the study. Baseline characteristics of enrolled cases and controls are described in Table 1. Of enrolled cases 84 (48.0%) were from a rural area, while 59 (33.7%) and 32 (18.3%) were from urban and peri-urban areas respectively.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Fig 1. Flowchart illustrating recruitment of patients with Salmonella Typhi infection in Central Division Fiji, 2014–2017.

https://doi.org/10.1371/journal.pntd.0006571.g001

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 1. Baseline characteristics of 175 typhoid fever cases and 349 age, ethnicity, and neighborhood matched controls enrolled in study, Central Division, Fiji, 2014–2017.

https://doi.org/10.1371/journal.pntd.0006571.t001

In the univariable analysis (Table 2), the odds of a case being from a low SES index household was almost 3 times that of controls (matched OR 2.98 and 95% CI 1.64–5.44). Cases more frequently accessed their main water source from outside the house (OR = 2.96, 95% CI 1.20–7.29), had interrupted water availability (OR = 2.40, 95% CI 1.39–4.12), and drank water from an untreated source in the last two weeks (OR = 1.80, 95% CI 1.07–3.03). Among participants who reported drinking water from a source other than their main household water source, cases were more likely to have drunk water from a surface water source than controls (OR = 3.04, 95% CI 1.33–6.92). No difference in kava consumption was observed between cases and controls.

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 2. Univariable analysis of risk factors for blood-culture-confirmed typhoid fever among 175 cases and 349 age, ethnicity, and neighborhood matched controls, Central Division, Fiji, 2014–2017.

https://doi.org/10.1371/journal.pntd.0006571.t002

Compared to controls, cases were more likely to eat unwashed produce (OR = 3.48, 95% CI 2.06–5.89). Cases were also more likely to have eaten outside of the house in the past two weeks (OR = 1.56, 95% CI 1.04–2.34), and have attended a mass gathering in the past two weeks (OR = 1.59, 95% CI 1.05–2.40).

When categorized by latrine type, cases were more likely to have either an unimproved pit latrine (OR = 26.62, 95% CI 5.20–136.37), or no toilet or latrine in the household (OR = 14.17, 95% CI 1.97–102.14), or a damaged, improved municipal sewerage (OR = 5.28, 95% CI 1.23–22.73), or an improved pit latrine (OR = 4.40, 95% CI 1.46–13.26) compared to controls. Furthermore, cases were more likely than controls to report having a latrine built by someone within their own household (OR = 1.53, 95% CI 1.12–2.30). With respect to hygiene behaviors, a higher handwashing score after defecating (OR = 0.41, 95% CI 0.27–0.62) and using soap for handwashing (OR = 0.38, 95% CI 0.24–0.58) were associated with lower odds of typhoid fever.

In terms of distal environmental factors, compared to controls, cases were more likely to have experienced flooding of their nearest river or stream in the past two months (OR = 2.42, 95% CI 1.13–5.12). Cases were also more likely to have a dam located upstream from their closest river basin (OR = 2.18, 95% CI 1.22–3.91) than controls.

On multivariable analysis ten exposures remained independently associated with typhoid fever (Table 3) including drinking water from an alternative surface water source in the last two weeks (OR = 3.61, 95% CI 1.44–9.06), not having constant water availability (OR = 2.17, 95% CI 1.18–4.00), and eating unwashed produce (OR = 2.69, 95% CI 1.48–4.91). The population attributable fraction (PAF) for these exposures was 7%, 15%, and 19%, respectively. Having any unimproved sewerage system or a damaged improved sewerage system was associated with increased odds of typhoid fever (OR = 4.30, 95% CI 1.14–16.21). The summary PAF for these poor sanitation facilities was 72%. Frequent handwashing after defecating was associated with lower odds of Salmonella Typhi infection (OR = 0.57; 95% CI 0.35–0.93), as was using soap for handwashing (OR = 0.61, 95% CI 0.37–0.95).

Download:

• PPT
  PowerPoint slide
• PNG
  larger image
• TIFF
  original image

Table 3. Multivariable analysis of risk factors for blood-culture-confirmed typhoid fever among 175 cases and 349 age, ethnicity, and neighborhood matched controls, Central Division, Fiji, 2014–2017.

https://doi.org/10.1371/journal.pntd.0006571.t003

Discussion

To our knowledge, this is the first case-control study to investigate sources and modes of transmission for typhoid fever in Fiji. We demonstrate that unimproved sanitation facilities are a likely major source of Salmonella Typhi, with transmission occurring through drinking contaminated surface water and eating unwashed produce. We also show typhoid to be common among both rural and urban populations in Fiji (Table 1). Improving sanitation facilities and increasing access to safe water and clean produce for rural populations presents a major challenge in Fiji and more broadly for other island nations of Oceania.

The median age of cases in our study was 29 years, which is higher than the median age reported in several other typhoid endemic areas [11, 12]. While this may suggest a distinct typhoid epidemiology in Fiji, under-ascertainment by blood culture of typhoid fever among younger age groups is also a possibility as younger patients in Fiji have been reported to receive early empiric treatment for fever without blood culture [4]. Of cases in our study, 95% were among iTaukei. This finding is in contrast to a sero-epidemiologic survey conducted in Fiji in 2013, which found both iTaukei and Fijians of Indian ancestry to have similar sero-prevalence of antibodies against the Vi antigen of Salmonella Typhi [20]. Besides Vi serology being an inaccurate measure of Salmonella Typhi infection, a possible explanation for this difference is under-ascertainment by blood culture of typhoid fever among Fijians of Indian ancestry, who have been shown to preferentially present to private general practitioners and are again likely to receive early and empiric treatment for fever without blood culture [21].

Our results suggest that unimproved or damaged sanitation is a major source of Salmonella Typhi in Fiji. We found that people without access to improved sanitation facilities or with damaged improved sewerage systems were at particular risk. Those with typhoid fever were more likely than controls to have someone within their household build their toilet (Table 2). Others have shown latrines built by persons without expertise to be poorly constructed, built into permeable soil, and subject to flooding [7, 22]. Notably, households with improved pit latrines had greater odds of typhoid fever in our study, compared to undamaged, improved municipal sanitation (Table 3). In Fiji, a common ‘improvement’ is use of buried steel drums as the receptacle for sewerage [23]. Such receptacles are subject to flooding, corrosion, and leakage [23] leading to contamination of surface water and crops by human feces [24]. Interestingly, eating unwashed produce was an independent risk factor for typhoid in our study. Related research has shown that gardens of patients with typhoid fever were more often positioned closer to the household toilet or septic tank than in control households and the majority of cases propagated vegetables directly on or below the toilet drainage area [25]. In contrast to water-related factors, poor sanitation has seldom been reported as a risk factor for typhoid fever [11, 12, 26], highlighting the value of our local research.

Contaminated drinking water is commonly identified as a risk factor for typhoid fever in case-control studies [26]. We showed that cases were more likely to report having poor water availability than controls (Table 3). Intermittent access to water is a common problem for households in many low-resource countries [27] and can lead to increased risk of typhoid fever by a number of mechanisms. First, in Fiji, as in many other locations, poor water availability from a primary source results in households shifting periodically to alternative water sources, including unsafe surface water [28, 29]. Second, during periods of reduced water supply, households may rely on stored drinking water that is not disinfected. Although we did not demonstrate consumption of stored water as a risk factor for typhoid fever in Fiji, related research showed a significantly higher concentration of E. coli in stored drinking water in typhoid case households compared to control households, but not in source water [25]. Third, pressure drops associated with regular interruption in reticulated water supplies can result in negative pressure situations which, when combined with leaks in the distribution system, can result in inflow of environmental material [30]. While our study was not designed to confirm this mechanism, we hypothesize that this may occur in Fiji and warrants further investigation. Finally, poor water availability can affect the quality of sanitation facilities that rely on water as well as affect personal hygiene [31, 32]. However, our multivariable analysis showed poor water availability as associated with increased odds of having typhoid fever even after adjustment for unimproved sanitation and handwashing frequency.

Factors related to hygiene were also independently associated with typhoid in our study setting (Table 3). Frequent handwashing after defecating and using soap for handwashing were associated with lower odds of typhoid fever. Both of these handwashing behaviors have been associated with reduced risk of typhoid in other studies [12, 28].

Distal conditions within the water catchment have yet to be thoroughly evaluated in case-control studies of typhoid fever. However, geospatial studies are beginning to examine such relationships on broad spatial scales [7, 33, 34]. Although no distal environmental conditions were statistically associated with typhoid fever in our multivariable model, experiencing flooding of the nearest river or stream in the past two months and reporting dams upstream in the river basin were significant in univariable analysis. Descriptive accounts of dam construction and bursts have been linked to an increase in typhoid cases in Nigeria [35]. However, the basis for these risks is not well understood and could be the subject of future research. At a sub-catchment scale, typhoid infection and disease in the Fijian setting has been linked with forest fragmentation, increased erosion, rainfall, and flood risk [7, 34]. It is likely that such environmental factors result in overflow and damage to already poor sanitation facilities leading to contamination of produce and drinking water sources.

Our study had a number of limitations. First, recall bias may influence the reliability of potential exposures over the long incubation period of typhoid fever [36] and social desirability bias is a common concern for sanitation and hygiene questions [37]. We sought to control the former by making observations of sanitation facilities. However, observations and more objective measures of handwashing practices were not possible. Therefore, our findings on handwashing should be interpreted with caution. Second, since the interviewers in this study knew the case status of the participant, it is possible that they may have acquired differential exposure information from cases. However, our questionnaire was standardized and interviewers were trained to ensure cases and controls were questioned in the same way. Third, the relative homogeneity of sampled environments and the collinear relationship of many factors may have masked the detection of potential risk factors. However, by recruiting a second control in a distant neighborhood and by obtaining a large sample size we expected to address power concerns of ‘over-matching’ and have adequate statistical power to identify minor associations. Finally, since case detection was by passive surveillance at public healthcare facilities, we are unlikely to have identified all typhoid fever illnesses and we may have missed cases that were more likely to be treated empirically or preferentially access private healthcare services. This could have also resulted in us missing exposures associated with increased risk of typhoid fever in these populations. Future research should include case finding at private health care facilities.

In conclusion, our study demonstrates that unimproved and damaged sanitation facilities are an important source of Salmonella Typhi in Fiji. Transmission appears to be by drinking contaminated surface water and consumption of unwashed produce, and is common in both rural and urban populations. Poor hygiene practices also appear to increase odds for typhoid fever. Although not detected in this study, landscape factors contributing to flooding of poor sanitation facilities may also contribute to enhanced risk [7, 34]. The situation in Fiji may reflect sources and modes of transmission predominant elsewhere in Oceania, where typhoid incidence is also high [1] and similar socio-demographic and environmental circumstances prevail. Meeting the 2030 Sustainable Development Goals goals [2] to improve sanitation facilities and protect surface water and produce from contamination by human feces are likely to contribute to typhoid control in Fiji. Central or household based water disinfection would also help to render fecally contaminated water safe for consumption. Such long-term socioeconomic, land and water management, and sanitation infrastructure developments together with uptake of typhoid conjugate vaccination in the interim are likely to result in effective typhoid control in Fiji and Oceania.

Supporting information

Acknowledgments

Helen Thomson and the staff of Murdoch Childrens Research Institute provided administrative support throughout the project. We thank Dawit Mulholland for assistance in the field.