In this study we have shown that Gambian patients with bacteraemia are more likely than those without to die in hospital and to have a raised peripheral blood WCC. Three organisms accounted for 73% of all bacteraemias; S. pneumoniae accounted for 45%. This hospital based study emphasizes the dominance of S. pneumoniae in the etiology of bacteraemia in The Gambia, consistent with previous etiology studies in this country [8–10].
S. pneumoniae has been identified as the dominant isolate in bacteraemia elsewhere in Africa. Berkowitz  identified bacteraemia in 315 (5.8%) of 5397 children admitted to a South African hospital; 23% died. S. pneumoniae accounted for 23% of isolates, S. aureus 6%, Salmonella species 20%, E. coli 13%, and H. influenzae 14%. Cotton et al  documented 132 episodes of community-acquired bacteraemia in hospitalized South African children; 12% died. S. pneumoniae accounted for 33% of pathogens isolated, S. aureus 14%, N. meningitidis 11% and other Gram negative organisms 18%. Berkley et al  found that 866 (5.9%) of 14,787 children over 60 days of age admitted to a rural hospital in Kenya had bacteraemia. S. pneumoniae accounted for 30% of isolates, S. aureus 7%, other Gram positive cocci 6%, NTS 19%, H. influenzae 15%, and E. coli 10%. In Malawi, Archibald et al  found blood stream infection in 70 (30%) of 233 febrile adults admitted to hospital, 70% of whom were HIV positive. S. pneumoniae accounted for 33% of isolates, Salmonella species 19% and Gram positive cocci only 4%. M. tuberculosis accounted for 29% of isolates, found exclusively in HIV positive patients.
Salmonella species have predominated in several other African studies, particularly in high HIV prevalence settings. Bahwere et al  identified 126 (15.9%) bacteraemias in 779 children admitted to a hospital in Congo;19.4% died. Salmonella species accounted for 44% of isolates. Gram positive bacteria accounted for only 10% of isolates. Ghiorghis et al  identified bacteraemia in 49 (7.7%) of 634 febrile children at an Ethiopian hospital. Of the organisms isolated, Salmonella species accounted for 57%, streptococci 16%, staphylococci 14%, and E. coli 6% and other Gram negative species 6%. Walsh et al  reported that of 365 positive isolated from children in Malawi, NTS accounted for 38% of isolates, H. influenzae 6%, other Gram negative species 29%, S. pneumoniae 16%, other streptococci 8% and S. aureus 2%. Gordon et al  identified 449 pathogens from the blood of adult patients in Blantyre: NTS accounted for 37% of isolates and 30% were S. pneumoniae. NTS accounted for 70–80% of all paediatric blood culture isolates in a hospital in Western Zaire . S. aureus was identified by Meremikwu et al  as the cause of bacteraemia in 48.7% of bacteraemias in children admitted to a Nigerian hospital, 44% were neonates. Differing criteria for patient selection, variation in the standardisation of methods and bacteriological techniques used may explain some of the differences seen between studies.
Others have identified risk factors for bacteraemia in African settings. Ghiorghis et al  found no difference in nutritional status, age or temperature between febrile bacteraemic cases and controls, but did find increased WCC in those with salmonellae and in those with E. coli. Bahwere et al , in their study dominated by Enterobacteraciae, also found no association with severe malnutrition. However, Cotton et al , Berkowitz et al  and Friedland  identified severe malnutrition as a risk factor for bacteraemia and death in South Africa. Archibald et al  identified HIV positivity as a risk factor for bacteraemia in Malawi. HIV and malnutrition have been shown to be independent risk factors for bacteraemia by Berkley et al .
In contrast to our study, seasonal variation in the etiology of bacteraemia has been demonstrated in sub-Saharan Africa. Bell et al , in a Malawi population with high rates of HIV positivity, found that NTSpredominated in the wet season while S. pneumoniae predominated in the dry season. High prevalence of NTS in the wet season in sub-Saharan Africa is thought to be associated with malaria [22–24], although other, often site-specific, factors should be considered as well. We had too few cases of NTS to assess seasonal variation properly in this study.
Other studies from The Gambia confirm that high resistance to penicillin is not yet prevalent here [10, 6]. In contrast, high resistance was found to co-trimoxazole, which is widely used at primary health care level. Evaluation of an alternative antimicrobial for community management of non-severe bacterial infections is now underway. Although numbers of isolates were small, S aureus was in general highly susceptible to cloxacillin, gentamicin and chloramphenicol. Penicillin was the least effective antibiotic. Unlike several other developing countries where resistance has been reported to these antimicrobials, cheap and widely available antibiotics remain effective in The Gambia. A larger study to give more precise estimates of antimicrobial susceptibility of isolates of E. coli and NTS is required.
While this study represents 'real life' clinical practice in this hospital and our data were prospectively gathered, our approach has several limitations. A more formal study would set strict criteria for taking blood for culture and standardize all clinical and laboratory procedures. Such an approach would enable certain tests to be routine, such as HIV testing, and may involve sending certain isolates to a reference laboratory for further identification. It is estimated that 5–10% of inpatients in our hospital are HIV positive (unpublished data). Despite having a standardized form, we did not ask the clinician to record his/her primary reason for taking a blood culture (we felt this would be difficult to categorize), although 47% of patients had a temperature of 37.5°C or more and 13% of these had a pathogen isolated. As 80% of our study subjects were under 5 years of age, interpretation of our findings in the older age groups should be done cautiously. We also cannot calculate an incidence rate of community-acquired bacteraemia as the source population of this hospital cannot be clearly defined and we did not obtain blood cultures from a representative sample of outpatient attendees .