Non-Typhi Salmonella (NTS) is a common cause of bacteremia among children and adults in Africa [1]–[12]. NTS bacteremia in Africa is highly associated with other diseases, particularly HIV and malnutrition, and possibly malaria [2], [9], [10], [12]. As the epidemiology of these associated diseases evolves in Africa, the epidemiology and burden of NTS bacteremia may also change. Moreover, recent reports of a multi-drug resistant NTS clone, causing invasive disease in Africa suggests that the epidemiology of NTS bacteremia might also be in flux due to other selective pressures [13], [14]. Lastly, as more Africans migrate to urban areas, living in crowded, squalid conditions, these new environments might also lead to a different epidemiology of NTS bacteremia. Updated epidemiologic data on NTS bacteremia can serve to guide future policy decisions on treatment and prevention.

We evaluated three years of population-based surveillance on NTS bacteremia from an urban informal settlement in Nairobi and a rural area of western Kenya. Our surveillance highlights stark differences in invasive Salmonella epidemiology in urban and rural Africa and the close associations between NTS bacteremia and HIV and malaria.

Of 3,578 blood cultures done in the rural site, 155 (4.3%) had a pathogen isolated, of which 60 (38.7%) were NTS, the most common bacteria isolated, followed by Streptococcus pneumoniae (36.2% of isolates, Table 1). Twenty-three (38%) of rural NTS bacteremia patients were hospitalized. In the urban site, 2,138 blood cultures were done, with a pathogen isolated from 230 (10.8%), of which 7 (3.0%) were NTS; the most common bacteria isolated in the urban site were Salmonella typhi and S. pneumoniae. The median age of pediatric NTS patients in both sites was 2 years old and for patients ≥5 years was 32 years old. The sex distribution of NTS cases was comparable (rural–51% female; urban–57% female). Among rural children <5 years with positive blood cultures, 28/43 (65%) had NTS compared with 5/69 (7%) of urban children (Table 1, p<0.0001). In the rural area, NTS accounted for a greater proportion of bacteremia among febrile patients (72% of isolates) than among SARI patients (30% of isolates, p<0.005) or inpatients (36%, p = 0.008). In the urban area, the NTS isolation rate did not vary by indication for blood culture (SARI 2%, fever 5%, p = 0.3). Diarrhea was reported in 34% of patients with NTS bacteremia.

The predominant serotype of Salmonella causing bacteremia differed in rural and urban Kenya, with NTS more common in the rural site and Salmonella Typhi more common in the urban site (Table 2) [23].

There was a broader distribution of NTS serotypes in stool than blood (Table 3). S. Typhimurium was the predominant NTS blood culture isolate in both rural and urban Kenya accounting for 88% and 86% respectively. S. Typhimurium represented a greater percentage of stool isolates in the rural site (64%) than the urban site (21%, Table 3). Overall, 80% of S. Typhimurium isolates from blood were multi-drug resistant (MDR) in the rural site (Figure 1). The percentage of MDR among NTS blood isolates in the rural site did not change significantly during the course of the study – 2007, 11/13 (85%), 2008 9/10 (90%), 2009, 11/17 (65%). Of 14 S. Typhimurium blood isolates from the rural site that underwent PFGE, the predominant restriction pattern was found in 8 (57%) by Xba1 and 9 (64%) by Bln1 restriction enzymes, all of which were MDR (Figure 2). Three of six urban S. Typhimurium isolates were MDR, two of which had the same predominant PFGE pattern (by both Xba1 and Bln1) as the rural isolates.

There was no clear seasonal pattern to NTS bacteremia in either site. In the rural site, a concomitant increase in the number of NTS bacteremia cases and positive malaria cases was observed in 2008–9 (Figure 3, Spearman's correlation, p = 0.018 for children, p = 0.16 adults, p = 0.0001 overall). Positive malaria blood smears were more common among NTS bacteremia patients (20%) compared with patients who were blood culture positive for pathogens other than NTS (10%), although not statistically significant (p = 0.14). The number of blood cultures and isolation rate for bacteria other than NTS from blood culture did not change substantively during the surveillance period, suggesting the rising incidence of NTS bacteremia was not due to methodological changes. Similar temporal trends were not seen in Nairobi.

In the rural site, 18 (43%) of 42 patients with NTS bacteremia whose HIV status was known were HIV-positive (Table 1). The HIV prevalence among NTS bacteremic patients >18 years was 80%, which was substantially higher than the adult HIV prevalence in the community (17%, KEMRI/CDC unpublished data). The one adult patient for whom HIV status was known in the urban site was HIV-negative.

Overall crude and adjusted NTS incidence was higher in rural Kenya (78 per 100,000 pyo and 580 per 100,000 pyo, respectively) than in urban Kenya (13 per 100,000 pyo and 57 per 100,000 pyo, respectively) (Table 4, p<0.0001 for urban: rural difference for both crude and adjusted rates). In both sites, NTS incidence was highest in children <5 years. In the rural site, the crude rate among HIV-infected persons >18 years was 334 per 100,000 pyo (95% CI 274–394), with an adjusted rate of 1,555 per 100,000 pyo (95% CI 956–2,155).

Among the 7 NTS-associated deaths (2 children and 5 older persons), 1 occurred in the first 30 days after the date of the positive blood culture, 3 died from 31–60 days, and 3 died from 61–90 days afterwards. The 30-day case-fatality ratio (CFR) was 0% for children and 3.1% for older persons; the 90-day CFR was 7.1% for children and 15.6% for older persons. Compared with patients of the same age group who had a blood smear positive for malaria, but were not bacteremic, the 90 day mortality among NTS bacteremia cases in the rural Kenya site was 9 times higher for patients <5 years old and 68 times higher for patients ≥5 years old (Table 5). In contrast, the 90 day mortality among patients with NTS bacteremia was not significantly higher than that among patients with bacteremia due to other bacteria. Verbal autopsies on these 7 patients determined that 6 died of HIV/AIDS and 1 died of tuberculosis. No NTS-associated deaths occurred in the urban site.

NTS was the leading isolate from bloodstream infections in rural Kenya, with incidence in the higher range of those found in other rural African settings, where rates ranged from 88–330 cases per 100,000 pyo [1], [2], [4], [7], [9], [24]–[26]. In contrast to rural Kenya, NTS bacteremia rates were much lower in the urban informal settlement, where Salmonella Typhi predominated [23]. The most likely explanation for the rural-urban disparity in NTS bacteremia is that malaria is holo-endemic in the rural setting, while only occurring sporadically through importation in high altitude Nairobi (approximately 1,600 meters elevation). The association between malaria and NTS bacteremia is well-described [25], [27]–[32]. NTS bacteremia is higher in those with current or recent malaria and malarial anemia [2], [9], [26]–[30]. NTS rates were higher in the malaria season in Gambia [7] and decreased concomitantly with a decreasing trend in malaria prevalence over several years in Gambia, Tanzania and coastal Kenya [25], [31], [32]. The association between NTS bacteremia and malaria is likely multi-factorial, and may include hemolysis leading to dysfunctional macrophage phagocytosis and release of free iron promoting bacterial growth [2], [25], [27]. The risk of NTS bacteremia might be particularly high with malarial anemia, which is very common among children in western Kenya [2], [28]. The increasing rate of NTS bacteremia from 2008 onwards in western Kenya likely resulted, in part, from an increasing malaria prevalence in the area [33].

Besides malaria, there are several other reasons for the rural/urban difference in invasive Salmonella epidemiology. A possible reason for high rates of typhoid in the urban site could be related to the crowded living conditions and poor water and sanitation, conditions more similar to Asian mega-cities where typhoid predominates [23], [34]. The lower rates of NTS bacteremia in the urban site might also be related to the observation that those who are terminally ill tend to leave Nairobi to go back “up-country” to their ancestral homes where they would prefer to die. As NTS bacteremia occurs more frequently in end-stage AIDS, this migratory phenomenon would tend to decrease NTS bacteremia in the city and increase it in rural areas. Another possible explanation for the low urban NTS rates is that people living in cities have less exposure to risk factors for NTS. There is some evidence from developed countries that exposure to domestic animals, particularly chickens which are common in rural Kenya, might be a risk for NTS disease [6], [8], [27].

The high burden of NTS in adults in the rural site is also influenced by the high prevalence of HIV, as has been shown in other African settings [13], [27], [35]–[40]. The adjusted rate of NTS among HIV-infected adults we found of 1,555 per 100,000 pyo was slightly lower than that in Malawi (8,000 per 100,000), Uganda (1,900 per 100,000) and elsewhere in Kenya (1,600–7,500 per 100,000) [36], [38]–[40]. This lower rate in Kenya might be due to the institution of the home-based HIV counseling and testing program in 2008 that identified many HIV-infected persons in early stages of infection (median CD4 400 cells/mm³), when they would be less susceptible to NTS bacteremia, as well as leading to the initiation of cotrimoxazole prophylaxis among many HIV-infected persons, which has been shown to prevent NTS bacteremia [41]. Almost all deaths following NTS bacteremia in our study were HIV-related, according to verbal autopsy. This suggests that while NTS bacteremia might not be the proximate cause of death, it can serve as an indicator of end-stage AIDS and of imminent death from other HIV-related causes. Immune reconstitution with anti-retrovirals and/or cotrimoxazole prophylaxis can reduce the risk of NTS bacteremia [36], [38], [41]. Third, because we not only cultured febrile patients, but also those with severe acute respiratory illness, we might have identified more NTS bacteremia. NTS bacteremia commonly presents with respiratory symptoms in children, although its role in causing pneumonia based on lung aspiration studies is doubtful [2], [42]. Alternatively, patients with severe respiratory illness due to another cause can be simultaneously bacteremic with NTS, particularly among HIV-infected persons [2], [11], [13], [35].

We found a lower mortality from episodes of NTS bacteremia than other studies. Because free, high-quality care was offered to surveillance participants, ill persons likely came to the clinic at earlier stages of illness, so that milder cases of NTS bacteremia might have been captured, as suggested by the observation that 62% of NTS bacteremia patients in the rural site were treated as outpatients. Most studies have been done among hospitalized patients, where case-fatality rates for NTS bacteremia were much higher than we found, ranging from 4–27% for children and from 22–47% for adults [27].

S. Typhimurium was the dominant serotype causing bloodstream infections in both urban and rural sites, in contrast to a much broader serotype distribution among contemporaneous NTS stool isolates [2], [3], [6], [8]. Rises in multi-drug resistant S. Typhimurium have been temporally related to increased numbers of NTS bacteremia cases in Malawi from 2001–2004 [13]. Genomic sequencing of these invasive multi-drug resistant S. Typhimurium isolates from Malawi, and of contemporaneous isolates from coastal and urban Kenya, revealed a similar clone, ST313 [14]. It was postulated that genomic degradation in this NTS clone created characteristics more similar to Salmonella Typhi, changes that led to greater invasive potential [14]. Without genetic sequencing, we lack conclusive evidence that the predominant multi-drug resistant strain in our sites were the ST313 clone, although the predominant PFGE pattern had an antimicrobial resistance profile similar to previously identified ST313 isolates from Kenya [14].

Our study had several limitations. First, in making adjustments of bacteremia rates, we made certain assumptions – NTS isolation rates would be the same among those meeting indications for blood culture who did not get cultured and that those patients under surveillance who attended another clinic for their illness would also have similar isolation rates to those who went to Lwak and Tabitha. Both assumptions might be subject to biases. Nonetheless, the crude rates are clearly an underestimate of the true rate in the community, due to limited health-care utilization and insensitivity of blood cultures, and the adjustments provide reasonable upward adjustments, although should be considered the maximal rate. Second, only 60% of patients with NTS bacteremia had a confirmed HIV status and we assumed that those not tested had the same proportion HIV-positive as those tested. This assumption also might have been subject to differential bias. Third, NTS is a more robust and antibiotic resistant than the more fastidious Gram-positive bacteria, particularly Streptococcus pneumoniae, and therefore we can only claim that NTS was the most common isolate from blood culture in the rural site, rather than the most common cause of bacteremia.

Regardless of these limitations, NTS bacteremia clearly continues to cause a substantial burden of morbidity among children and HIV-infected adults in rural Kenya, warranting further investigation to define amenable risk factors, best treatment guidelines and new interventions, such as NTS vaccines currently under development [43].