An acute undifferentiated febrile illness is characterized by a fever of fewer than 2 weeks for which no localizing signs or etiology is found after a full history and physical examination (1, 2). It may be the only symptom of a mild self-limiting illness or antecedent of a serious infection caused by a bacterial pathogen (3, 4). Most cases of acute undifferentiated fever (AUF) are caused by viruses in South and Southeast Asia while malaria parasitemia has frequently been identified in parts of sub-Saharan Africa (5–7).

In most resource-poor settings, polymicrobial infection with viruses, bacteria, and malaria is common and could be clinically indistinguishable (8–10). Although malaria could be rapidly diagnosed in most of these settings, clinicians in these settings are often challenged by the limited access to bacterial culture facilities to aid in accurate diagnosis of bacteremia (11). Also, white blood cell parameters which are available are inadequate to differentiate between malarial, viral, and bacterial pathogens. Hence, children with acute fever are often presumptively prescribed antibiotics, and this could increase the risk of antibiotic resistance (12, 13).

Unlike those with AUF, children with demonstrable focus of infection such as pneumonia and meningitis are rationally prescribed antibiotics according to existing guidelines (14, 15). However, there is currently a dearth of information on the accurate burden of bacteremia and its clinical predictors among children with AUF in most resource-poor settings, especially in sub-Saharan Africa, and this could jeopardize the rational deployment of antibiotics. The Pediatric Early Warning Score (PEWS) and the pediatric quick Sequential Organ Failure Assessment (qSOFA score) have been used to identify patients at risk of poorer outcomes (16–19). The qSOFA have been mostly used to predict mortality from sepsis rather than identifying subjects who had sepsis and needed prompt initiation of antibiotics. Besides, the sensitivity in predicting mortality and severe sepsis has been reportedly poor in the pediatric population (18, 20). Also, the qSOFA included blood pressure measurement, which is often not done routinely in most resource-poor settings. Similarly, the PEWS was designed to identify hospitalized children with heterogeneous disease conditions at risk for clinical deterioration and may not be applicable to children with AUF. Studies identifying the bacterial etiology of childhood AUF as well as its clinical predictors would thus be crucial, and this is the thrust of the current study. This study was initiated on a platform of ongoing surveillance for bacteremia in young children to identify clinical predictors of bacteremia among under-five children with acute undifferentiated fever.

Between November 2015 and March 2016, 140 under-five children who presented with acute undifferentiated fever were recruited into this study. Of this number, we identified 24 (17.1%) with a positive blood culture and identified an organism in 22 (15.7%) of them. Salmonella Typhi was the most frequently cultured organism (40.9% of isolates; Table 1), and this was followed by non-Typhi Salmonella (31.8% of isolates; Table 1). Co-trimoxazole had the highest resistance among antibiotics as 77.8% and 85.7% of Salmonella Typhi and non-Typhi Salmonella were resistant to the antibiotic, respectively (Table 1).

We investigated the burden and factors predictive of bacteremia among under-five children with acute undifferentiated fever seen at the emergency unit in northwestern Nigeria. We found that one in six participants had bacteremia, majority from Gram-negative enteric pathogens. Inpatient care, longer duration of fever, lethargy, tachypnea, and lymphopenia were all associated with the increased odds of bacteremia.

The prevalence of bacteremia among under-five children with acute undifferentiated fever was 17.1%. This is similar to the rate reported among children with similar characteristics in Benin, Nigeria (5). However, a higher prevalence rate (25.7%) was reported among Indian children (33). The diagnosis of bacteremia in the current study was based only on blood culture as opposed to the Indian study where the Widal test, a serologic test fraught with high false-positive rates, complemented blood culture in the diagnosis of Salmonella infection; this may have overestimated the burden of bacteremia (34). In contrast, the prevalence of bacteremia was significantly lower (1.4%) in a previous study conducted in Brazil with a similar study population to ours (35). The higher positive culture rate found in our study could be because of the utilization of a more sensitive automated culturing system, unlike the Brazilian study where the conventional manual bacteriological culture method was used.

Gram-negative pathogens were the predominant cause of bacteremia among our cohort, accounting for over 80% of the isolates. The preponderance of Gram-negative pathogens in childhood bacteremia had been described previously in Northern Nigeria and in other low- and middle-income countries (5, 7, 33, 36). The preponderance of enteric Gram-negative pathogens in the study location, like in most resource-poor settings, could be related to the high presence of predisposing factors such as high level of poverty, limited access to clean water, poor sanitation, and malnutrition (37). In contrast, some studies have reported predominantly Gram-positive pathogens, especially Staphylococcus aureus among febrile children (38–40). While this difference might reflect geographical variations of bloodstream bacterial pathogens, the characteristics of the study participants and the study design could also account for the differences in the organisms isolated. The current study is prospective; hence, it has very minimal selection bias, unlike the previous studies, which were retrospective in design. Also, our subjects are those with acute undifferentiated febrile illnesses and are also devoid of peculiar risk factors such as HIV infection that could predispose to Staphylococcus aureus infection (41).

Salmonella infection accounted for most of the cases, consistent with the previous reports from the same study location (27, 36). Bacteremia caused by Salmonella is most prevalent in the dry season (October–March), and the conduct of the present study during this peak period may have accounted for the predominance of the organism (36). Meningococcal bacteremia was not found in the current study despite the timing of the study, which coincided with the seasonal peak of Neisseria meningitides. This observation may, however, not reflect the burden of meningococcal bacteremia in the study location because subjects with overt clinical features of meningitis such as neck stiffness and abnormal tone were excluded from the current study.

In vitro antibiotic susceptibility testing revealed a high prevalence of co-trimoxazole resistance, especially by Salmonella infection, but with high sensitivity to gentamicin, ciprofloxacin, ceftriaxone, and amoxicillin-clavulanate. A previous multicenter study in Northern Nigeria has also shown high resistance of non-typhoidal Salmonella and Salmonella Typhi to co-trimoxazole (36). This finding has some implications; first, amoxicillin-clavulanate would seem a rational antibiotic on an outpatient basis, but it is quite expensive compared to co-trimoxazole. Second, ciprofloxacin is cheap and available in both oral and parenteral forms and could be used in both inpatient and outpatient settings, but the widespread use of ciprofloxacin has been associated with a high incidence of extended-spectrum β-lactamase enzyme-producing pathogen, hence increasing antibiotic resistance (42). Third, like ciprofloxacin, ceftriaxone is a broad-spectrum antibiotics, and its use as a first-line drug could also promote resistance. Also, it is expensive and available only in a parenteral formulation and would be less affordable in resource-poor settings where the majority have no form of health insurance.

In this study, the duration of fever rather than the degree of fever was associated with increased odds of bacteremia, and this is similar to previous studies (6, 43). It was further identified that subjects were four (4) times more likely to have bacteremia if fever was present beyond 7 days. Given that the majority of subjects without bacteremia had fever for <7 days, which may have been due to undiagnosed viral illnesses, the increased likelihood of bacteremia after 7 days may thus represent a bacterial superimposition. Other studies have also failed to demonstrate a correlation between the degree of fever and positive bacterial culture (44, 45). Fever results from a cytokine-mediated elevation of the hypothalamic set point and is perhaps the earliest systemic feature of microbial invasion of the bloodstream (46). An increasing duration of fever could therefore reflect an intense bacterial multiplication and an increased tendency of positive culture as obtained in our study. Furthermore, participants who required hospitalization were four times more likely to have bacteremia when compared to outpatients, while those admitted with lethargy and fast breathing were six times more likely to have bacteremia, and this is consistent with a previous study (35). Febrile children presenting with lethargy and fast breathing often appear ill, necessitating inpatient care. In this study, a subjective assessment of “ill/toxic look” made by the attending emergency clinician was one of the eligibility criteria for hospitalization; the discriminatory power of the criteria thus gave it credibility and could thus be adopted in triaging febrile children in other resource-poor settings where facilities and resources for inpatient care are suboptimal.

Among all hematological parameters, a low lymphocyte count remained predictive of bacteremia, and participants with lymphopenia were four times more likely to have bacteremia compared to those who did not have lymphopenia. This finding could be due to the predominance of Salmonella species among the organisms isolated. Previous studies, including a controlled human infection model, have consistently demonstrated an association between Salmonella infections and lymphopenia (47, 48).

We have described the burden and the predictors of bacteremia among under-five children with acute undifferentiated fever. Such children pose a diagnostic challenge, and clinicians are often in a dilemma whether to use antibiotics empirically or not, unlike previous studies on bacteremia conducted among children with focal signs of infection, e.g., meningitis or pneumonia, in whom the deployment of empirical antibiotics is justified. In addition, our study design was prospective, and this enabled the recruitment and the collection of data from all our participants using preset clinical criteria, thus minimizing selection bias as observed in a previous study (7).

However, about 13% of the study participants were lost to follow-up, and hence, the mortality rate as presented may not reflect the actual death attributable to bacteremia among the study participants. In addition, the potential impact of the pathogens causing bacteremia on the identified predictors could not be assessed due to the preponderance of Salmonella among the isolates. Therefore, this picture may not apply in other settings where pneumococcal or other forms of bacteremia are most prevalent.

We have shown that one in six under-five children presenting to a typical emergency unit in Northwestern Nigeria with AUF had bacteremia, and this was mainly due to Salmonella. Participants who required hospitalization, those that presented with lethargy, had fever beyond 7 days, had tachypnea, or had lymphocytes counts <3 × 10⁹/L, were at increased odds of bacteremia. We, therefore, recommend that in settings similar to ours, the presence of these clinical predictors could form the for a rational ordering of blood culture. Also, where facilities for blood cultures are not available, febrile under-five children without focal signs but possessing the highlighted clinical features should be treated empirically for bacteremia using the most sensitive antibiotics.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The study was reviewed and approved by Ethics Committees of the University of Nebraska Medical Center, Aminu Kano Teaching Hospital and the Kano State Hospital Management Board. Written informed consent to participate in this study was provided by the participant's legal guardian/next of kin.

TO conceived the study, recruited, and obtained clinical data. TO, TA, SE, and FH-H designed the study protocol. AI analyzed the data. TO and AI prepared the first draft of the manuscript. TO, RO, and SO are guarantors for the paper. All authors contributed to the first major revision of the manuscript, final manuscript revisions, and approval of the final version.

Research reported in this publication was supported in part by the National Institute of Allergy and Infectious Diseases of the National Institutes of Health (Grant Number. R01AI097493); the Bill & Melinda Gates Foundation (Grant Number. OPP1034619); internal grants from the University of Nebraska Medical Center.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.