Septicaemia in a population-based HIV clinical cohort in rural Uganda, 1996–2007: incidence, aetiology, antimicrobial drug resistance and impact of antiretroviral therapy


Corresponding Author Billy N. Mayanja, MRC/UVRI Uganda Research Unit on AIDS, PO Box 49, Entebbe, Uganda. Tel.: +256 414 320 042/+256 414 320 272; Fax: +256 41 4321 137; E-mail:


Objectives  To describe the incidence and aetiology of septicaemia, and antimicrobial drug resistance in HIV-infected and uninfected individuals, and the impact of antiretroviral therapy (ART) on septicaemia.

Methods  Between 1996 and 2007, we followed up a rural population–based cohort of HIV-infected and uninfected participants. The aetiology and incidence of septicaemia, and antimicrobial drug resistances were determined. ART became available in 2004, and its impact on the incidence of septicaemia was examined.

Results  The overall septicaemia incidence (per 1000 pyrs) was 32.4 (95% CI 26.2–40.6) but was only 2.6 (95% CI 1.3–6.2) in HIV-negative patients and 67.1 (95% CI 53.4–85.4) in HIV-positive patients not on ART. Among those on ART, the overall incidence was 71.5 (95% CI 47.1–114.3), although it was 121.4 (95%CI 77.9–200.4) in the first year on ART and 37.4 (95%CI 18.9–85.2) in the subsequent period. Septicaemia incidence was significantly associated with lower CD4 counts. The commonest isolates were Streptococcus pneumoniae (SPN, n = 68) and Non-typhi salmonellae (NTS, n = 42). Most SPN isolates were susceptible to ceftriaxone and erythromycin, while resistance to cotrimoxazole and penicillin was common. All NTS isolates were susceptible to ciprofloxacin, but resistance to cotrimoxazole and chloramphenicol was common.

Conclusions  Septicaemia incidence was higher in HIV-infected than in HIV-uninfected participants, and it remained high for some time among those who started ART. Starting ART earlier at higher CD4 counts is likely to lead to lower septicaemia incidence. Both SPN and NTS, the commonest isolates, were resistant to most commonly available antimicrobials. Blood culture laboratory surveillance systems to monitor antibiotic susceptibility and inform treatment guidelines are needed in Africa.


Septicémie dans une cohorte clinique VIH basée sur la population en milieu rural en Ouganda de 1996 à 2007: incidence, étiologie, résistance aux médicaments antimicrobiens et impact de la thérapie antirétrovirale

Objectifs:  Décrire l’incidence et l’étiologie de la septicémie, la résistance aux médicaments antimicrobiens chez les individus infectés et non infectés par le VIH et l’impact du traitement antirétroviral sur la septicémie.

Méthodes:  Entre 1996 et 2007, une cohorte basée sur la population rurale, de participants infectés et non infectés par le VIH, a été suivie. L’étiologie et l’incidence de la septicémie ainsi que la résistance aux médicaments antimicrobiens ont été déterminées. L’ART est devenue disponible en 2004 et son impact sur l’incidence de la septicémie a été examiné.

Résultats:  L’incidence globale de septicémie (pour 1000 personnes-année) était de 32,4 (IC95%: 26,2 -40,6) mais était seulement de 2,6 (IC95%: 1,3 - 6,2) chez les patients VIH-négatifs et 67,1 (IC95%: 53,4 - 85,4) chez les patients VIH positifs non sous ART. Chez ceux sous ART, l’incidence globale était de 71,5 (IC95%: 47,1 - 114,3), mais elle était de 121,4 (IC95%: 77,9 - 200,4) dans la première année sous ART et 37,4 (IC95%: 18,9 - 85,2) par la suite. L’incidence de septicémie était significativement associée à des taux faibles de CD4. Les isolats les plus courants étaient Streptococcus pneumoniae (SPN, n = 68) et salmonellae non typhi (SNRC, n = 42). La plupart des isolats de SPN étaient sensibles à la ceftriaxone et à l’érythromycine, tandis que la résistance au cotrimoxazole et à la pénicilline était fréquente. ous les isolats de SNRC étaient sensibles à la ciprofloxacine, mais la résistance au cotrimoxazole et au chloramphénicol était fréquente.

Conclusions:  L’incidence de septicémie était plus élevée chez les participants infectés que chez ceux non infectés par le VIH, et est restée élevée pendant quelque temps chez ceux qui ont commencé ART. L’ART commencée plus tôt, à des taux plus élevés de CD4 est susceptible de mener à une plus faible incidence des septicémies. Autant SPN que NTS, les deux plus communs isolats, étaient résistants aux antimicrobiens les plus couramment disponibles. Des systèmes de surveillance de laboratoire utilisant la culture d’échantillons sanguins pour contrôler la sensibilité aux antibiotiques et informer les directives de traitement sont nécessaires en Afrique.


Septicemia en una cohorte clínica de VIH basada en la población en Uganda rural, 1996 - 2007: incidencia, etiología, resistencia a antibióticos e impacto en la terapia antirretroviral

Objetivos:  Describir la incidencia y etiología de la septicemia, y la resistencia a antimicrobianos en individuos infectados y sin infectar con VIH, y el impacto de la terapia antirretroviral sobre la septicemia.

Métodos:  Entre 1996 y 2007, seguimos una cohorte basada en una población rural de participantes infectados con VIH y sin infectar. Se determinaron la etiología e incidencia de la septicemia, y las resistencias a medicamentos antimicrobianos. El TAR estuvo disponible a partir del 2004 y se ha examinado su impacto sobre la incidencia de la septicemia.

Resultados:  La incidencia total (por 1,000 personas año) era de 32.4 (95% IC 26.2 a 40.6), siendo solamente del 2.6 (95% IC 1.3 a 6.2) en pacientes VIH negativos mientras que en pacientes VIH-positivos que no estaban en TAR era del 67.1 (95% CI 53.4 a 85.4). Entre aquellos recibiendo TAR, la incidencia total era del 71.5 (95% IC 47.1 a 114.3), aunque era de 121.4 (95%CI 77.9 a 200.4) para pacientes en primer año de TAR y de 37.4 (95%CI 18.9 a 85.2) para pacientes en periodos subsiguientes. La incidencia de septicemia estaba significativamente asociada con un conteo más bajo de CD4. Los aislados más comunes eran Streptococcus pneumoniae (SPN, n=68) y salmonella no-typhi (SNT, n=42).La mayoría de los aislados de SPN eran susceptibles a ceftriaxona y eritromicina, mientras que la resistencia a cotrimoxazol y penicilina era común. Todos los aislados de SNT eran susceptibles a ciprofloxacino, pero la resistencia a cotrimoxazol y cloranfenicol era común.

Conclusiones:  La incidencia de septicemia era mayor entre los infectados con VIH que entre los pacientes no infectados, y se mantuvo alta durante algún tiempo entre aquellos que comenzaron TAR. El comenzar TAR antes, cuando aún se tienen conteos altos de CD4, podría llevar a una menor incidencia de septicemia. Tanto los SPN como NTS, los aislados más comunes, eran resistentes a los antimicrobianos disponibles más comunes. Se requieren sistemas de seguimiento en los laboratorios para los cultivos de sangre con el fin de monitorizar la susceptibilidad antibiótica e informar sobre las guías de tratamiento en África.


Septicaemia is an important cause of morbidity and mortality worldwide. Before introduction of antiretroviral therapy (ART) for HIV-infected patients, its incidence was increasing globally, with a shift from the predominance of gram-negative organisms in the 1970s to gram-positive organisms in the 2000s (Bearman & Wenzel 2005). HIV-induced immunodeficiency increases the risk of septicaemia, especially that owing to Non-typhi salmonellae (NTS) and Streptococcus pneumoniae (SPN), which are important causes of morbidity and mortality among HIV-infected individuals (Gordon et al. 2001; Gordon 2008). NTS septicaemias are usually severe, invasive and recurrent and are the commonest cause of hospital admissions owing to bacteraemia in sub-Saharan Africa (Gordon et al. 2002; Kankwatira et al. 2004). Invasive pneumococcal disease tends to occur earlier in HIV infection than gram-negative septicaemia (Grant & De Cock 2001).

Prompt use of appropriate antibiotics can dramatically improve the outcome of patients with septicaemia (Nobre et al. 2007). However, in developing countries, many patients with septicaemia do not fully benefit from antimicrobial therapy. Antimicrobial resistance is a worldwide phenomenon but particularly common in developing countries, where human and animal antimicrobial use is often unregulated, and antibiotics can be purchased over the counter without prescription (Hart & Kariuki 1998; Okeke et al. 1999; Byarugaba 2004). Natural calamities, poverty, poor adherence to the prescribed regimen, substandard antimicrobial formulations and antibiotic misuse by physicians and unskilled practitioners all encourage antimicrobial resistance (Hart & Kariuki 1998; Okeke et al. 1999; Byarugaba 2004).

Although the impact of ART in decreasing the incidence of HIV-related septicaemia has been well documented in developed countries (Tumbarello et al. 2000; Pedro-Botet et al. 2002), there is little information so far from developing countries in sub-Saharan Africa. Because health facilities in these settings often lack microbiological laboratories, and new and more effective antimicrobials are often unavailable or too expensive for use in standard management, knowledge of prevailing causes of septicaemia and antimicrobial sensitivity patterns is important in guiding treatment.

We describe the incidence and aetiology of septicaemia, and antimicrobial resistance patterns, in HIV-infected and uninfected outpatients in rural Uganda over 12 years (1996–2007). The introduction of ART in 2004 in a population-based HIV clinical cohort enabled an assessment of its effect on the incidence of HIV-related septicaemia.

Materials and methods

Study setting and participants

The cohort in rural southwest Uganda is drawn from a stable, homogeneous population mainly from the Baganda tribe and subsistence farmers. The majority are Christians, and 50% are under 15 years (Nakibinge et al. 2009). The main domestic water sources are boreholes used by 45% of the population, shallow wells used by 32% and protected springs used by 23%, and pit-latrine coverage stands at 75–80% (unpublished data). Adult HIV-1 prevalence declined from 8.5% in 1990/1991 to 6.2% in 1999/2000 but thereafter rose to 7.7% in 2004/2005 (Shafer et al. 2008).

Participants in the clinical cohort were enrolled from a larger open general population cohort (GPC), which was established in 1989 to describe the population dynamics of HIV-1 infection (Mulder et al. 1994) and now comprises about 20 000 residents of 25 neighbouring villages. In 1990, one-third of HIV prevalent cases from the first GPC survey were randomly selected and enrolled into the clinical cohort. In addition, all adult HIV seroconverters (13 years and older) identified during subsequent annual GPC surveys were invited to enrol in the clinical cohort as HIV incident cases. To facilitate comparisons with the background morbidity and mortality in the study population, HIV-negative controls were randomly selected from the population to match the age and sex of the HIV prevalent and incident cases. HIV-negative participants who seroconverted during follow-up remained in the clinical cohort.

The clinical cohort has been described in detail elsewhere (Morgan et al. 1997). Study clinicians were initially blinded to the participants’ HIV serostatus and were un-blinded in 2004 when ART was introduced. Because participants in the GPC HIV serosurvey may opt not to know their HIV serostatus, some of those invited to enrol in the clinical cohort are unaware of their HIV serostatus and are therefore encouraged to undergo voluntary HIV counselling and testing at centres provided in the study area and at the study clinic. Eligibility for ART (introduced in January 2004) was in line with Ministry of Health criteria (MOH Uganda 2003a): CD4 count of 200 cells/μl or below; or WHO clinical stage 4; or advanced stage 3 with persistent or recurrent oral thrush and invasive bacterial infections regardless of CD4 count; or CD4 count of 250 cells/μl or below in pregnancy.

Clinical methods

Participants gave informed written consent to participate in the study and were seen in the clinic every 3 months (routine visits) and in the event of inter-current illness (interim visits). CD4 cell count was measured every 3 months for participants on ART, every 6 months for HIV-positive participants not on ART and annually for HIV-negative controls. At routine visits, a medical interview and clinical examination were performed, and the findings entered onto pre-coded questionnaires. The clinical diagnosis and laboratory results at a preceding interim visit were also recorded. Home visitors attempted to trace participants who missed their appointments and those who emigrated from the study area.

Between January 1996 and December 2007, study participants with fever (axillary temperature of 38 °C and above) but with no detectable malaria parasites had a blood sample taken under aseptic conditions for culture and sensitivity testing. Our definition of septicaemia was isolation of pathogenic bacteria in the blood culture of a patient with a temperature of 38 °C and above, in the absence of malaria parasitaemia.

Laboratory methods

Thick and thin blood smears were prepared, dried and stained with the Leishman stain and examined to estimate the malaria parasite density and determine the malaria species, respectively. Two independent Enzyme immunoassay (EIA) tests were used to determine HIV-1 serostatus using parallel test algorithms, with external quality assurance from the College of American Pathologists (325 Waukegan Road, Northfield, IL 60093-2750, USA). Western Blot testing was used to resolve discordance between the EIA tests. (Nunn et al. 1993). CD4 T-lymphocyte cell counts were performed using the FACSCount method (Becton Dickinson, San Jose, CA, USA) with external quality control from the United Kingdom External Quality Assurance Scheme (UKNEQAS).

Between January 1996 and April 2004, brain heart infusion and nutrient broth bottles were used for blood cultures, and subcultures were performed under 5–10% carbon dioxide on chocolate agar and aerobically on blood agar. From May 2004, BD BACTEC PLUS aerobic and anaerobic blood culture bottles were used and then processed with a Bactec automated fluorescent blood culture system (BD BACTEC 9120; Becton Dickinson, BD Biosciences). Serological, enzymatic or the Analytic Profile Index (API 20E) biochemical galleries (Biomerieux S.S, Mercy-l’Etoile, France), or any combination of the above were used to identify the culture organisms. We used the McFarland standard tube in the preparation of a standard turbidity of test organisms. Disc diffusion method and E-test minimum inhibitory concentration method (AB BIODISK, Solna, Sweden) were used for antimicrobial sensitivity testing, with external quality control from the College of American Pathologists (325 Waukegan Road, Northfield, Illinois, 60093-2750, USA). Standard operating procedures were followed, and all laboratory results were recorded on standard pre-coded laboratory forms.

Statistical methods

Up to 2004, we used FoxPro, and since then, Ms Access databases (Microsoft Corp., USA) for data entry and Stata version 10 (Stata Corporation, College Station, Texas, USA) for data analysis. Patients were classified into three categories: HIV- negative, HIV- positive not on ART and HIV- positive on ART. Sixteen participants had un-interpretable HIV results and were excluded from the analysis. Some participants moved from one category to another, owing to seroconversion or being started on ART, and the time spent by each participant in each category was calculated, and the person-years (pyrs) of observation summed over the time spent in each category. Person-time started at enrolment in the cohort and was censored at death or loss to follow-up. The participants’ characteristics and time at risk were examined for each of the three categories. Incidence rates (IR) of septicaemia with 95% confidence intervals (CI) were calculated for each category. To obtain incidence rate ratios (IRR) and 95% CI for the comparison between categories and by age and sex, we used robust standard errors analysis to account for multiple septicaemia episodes in the same individual. CD4 counts (cells/μl) taken within 12 months prior to the blood culture were categorised into three groups: under 200, 200–499 and 500 and above; and the septicaemia incidence was calculated and compared between these groups; 95% CIs appear in square brackets after relevant values.

Ethical considerations

The study was approved by the Science and Ethics committee of the Uganda Virus Research Institute and the Uganda National Council of Science and Technology.


Between January 1996 and December 2007, we followed 759 participants, 199 (26.2%) of whom died. We observed a total of 4909 person-years (pyrs) of follow-up; 2668 pyrs (54.3%) among 286 HIV-negative participants (15 of whom seroconverted during follow-up), 1835 pyrs (37.4%) in 387 HIV-positive participants not on ART, (99 of whom later started ART) and 406 pyrs (8.3%) among 200 participants on ART (Table 1).

Table 1.   Description of study participants’ characteristics
CharacteristicsStudy participantsTotal n (%) P-value
HIV negative n (%)HIV positive not on ART n (%)HIV positive on ART n (%)
  1. ART, antiretroviral therapy.

  2. *Number includes 15 participants who subsequently sero-converted to HIV positive.

  3. †Number includes 15 who seroconverted and 99 who accessed ART while in the cohort.

  4. ‡Number includes 99 who accessed ART while in the cohort.

Total contributing286*387†200‡759 
 Men149 (52%)185 (48%)80 (40%)363 (48%) P = 0.036
 Women137 (48%)202 (52%)120 (60%)396 (52%) 
Age at start (years)
Median (IQR)
 Men35.2 (29.0–48.5)33.7 (27.6–44.1)38.7 (33.1–47.2)35.2 (28.9–44.9) 
 Women32.5 (25.5–44.5)28.6 (22.9–36.7)35.5 (30.3–41.6)31.8 (24.5–40.9) 
 Overall33.5 (27.9–46.5)31.4 (24.9–41.2)36.9 (31.6–43.0)33.4 (26.3–43.2) 
Died:n (rate: per 100 person years)26 (0.97)142 (7.74)31 (7.64)199 (4.05) P < 0.001
Person-years of follow-up2668 (54.3)1835 (37.4)406 (8.3)4909 
Median follow-up time (years) (IQR)12.0 (6.6–12.0)4.2 (2.2–7.0)1.9 (0.8–3.4)5.4 (2.6–12.0) P < 0.001

From the 703 blood cultures collected, 159 (22.6%) grew a bacterial isolate, 152 (95.6%) of which were from HIV-positive patients and 7 (4.4%) from HIV-negative patients. Cryptococcus neoformans was isolated from seven cultures (excluded from this analysis). Among HIV-negative participants, 87% of CD4 counts were above 500 cells/μl, and the proportions did not differ between those with growth of a bacterial isolate (83%) and those without bacterial growth (87%). In the 12 months prior to the blood culture, the mean CD4 count (cells/μl) was 877 among HIV-negative participants, 312 among the HIV-positive participants not yet on ART and 355 among those on ART.

Septicaemia incidence per 1000 person-years at risk (pyrs) was 32.4 (26.2–40.6) overall and 2.6 (1.3–6.2) in HIV-negative patients. Septicaemia incidence was significantly higher among HIV-positive patients regardless of ART status than in HIV-negative patients and among those aged 25–34 and 35–44 years than among 13–24 year olds. Septicaemia incidence was 67.1 (53.4–85.4) in HIV-positive patients not on ART. Among those on ART, overall septicaemia incidence (per 1000 pyrs) was 71.5 (47.1–114.3) and was 121.4 (77.9–200.4) in the first year on ART and 37.4 (18.9–85.2) in the subsequent period (17-fold higher than among HIV-negative patients (Table 2). Septicaemia incidence was significantly associated with lower CD4 counts, regardless of whether or not the person was on ART, with a sevenfold higher incidence in those with a CD4 count (cells/μl) under 200 compared to those with 500 or more.

Table 2.   Incidence of septicaemia per 1000 pyrs, by HIV serostatus, age, sex and CD4 counts
SepticaemiaNo.IR (95% CI)IRR (95% CI) (unadjusted)IRR (95% CI) (adjusted for age, CD4 counts and HIV/ART)
  1. IR, incidence rate; IRR, incidence rate ratio; ART, antiretroviral therapy.

Overall15932.40 (26.18–40.55)  
HIV and ART status   P < 0.001 P < 0.001
 HIV negative72.62 (1.29–6.22)11
 HIV positive – not on ART12367.05 (53.36–85.38)28.97 (13.29–63.16)28.80 (13.42–61.82)
 HIV positive – all on ART2971.50(47.10–114.25)33.33 (14.32–77.58)29.90 (12.79–69.91)
  First year on ART10121.43 (77.88–200.40)58.39 (24.16–141.10)52.29 (21.64–126.30)
  Subsequent period1937.37 (18.88–85.17)17.05 (6.28–46.33)15.07 (5.44–41.74)
Sex   P = 0.45 P = 0.45
 Men8735.10 (25.80–49.0)11
 Women7229.60 (22.30–40.20)0.85 (0.55–1.30)0.86 (0.58–1.27)
Age group (years)   P < 0.001 P = 0.011
 13–241619.0 (11.20–34.80)11
 25–346634.30 (24.10–50.40)1.80 (0.94–3.45)1.80 (0.94–3.42)
 35–445450.20 (35.50–73.10)2.68 (1.41–5.09)2.46 (1.31–4.60)
 45+2321.60 (13.20–37.60)1.13 (0.54–2.40)1.74 (0.83–3.63)
CD4 count group (cells/μl)   P < 0.001 P < 0.001
 < 2006278.0 (60.80–100.00)11
 200–4996952.60 (41.60–66.60)0.71 (0.50–1.00)0.61 (0.43–0.87)
 500+2811.0 (7.57–15.90)0.14 (0.09–0.22)0.33 (0.21–0.53)

Streptococcus pneumoniae (SPN, n = 68) and Non-typhi salmonellae (NTS, n = 42) were the commonest isolates contributing 42.8% and 26.4%, respectively, to the total number. Of the other isolates, 9 (5.7%) were Escherichia coli, 6 (3.8%) Salmonella typhi and 6 (3.8%) Staphylococcus aureus (Table 3). Coagulase-negative staphylococci were isolated from 10 cultures (6.3%). The seven isolates from HIV-negative individuals included four SPN, one NTS, one E. coli and one S. aureus. The overall incidence of SPN septicaemia was 13.9 per 1000 pyrs (10.1–19.5). SPN septicaemia incidence was significantly higher among HIV-positive patients regardless of whether or not they were on ART, those with low CD4 counts, male patients and those aged 25–34 and 35–44 years (Table 4). The overall incidence of NTS septicaemia was 8.6 per 1000 pyrs (6.2–12.2). NTS septicaemia incidence was also significantly higher among HIV-positive (regardless of whether they were on ART or not) than in HIV-negative patients and those with lower CD4 counts, but there was no significant difference by gender and age (Table 4). There was a fall in overall septicaemia among all participants from a peak in 2004/05 to 2006/07, which was more marked among HIV-positive participants overall and for SPN and NTS (data not shown).

Table 3.   Aetiology and incidence rates (IR) of septicaemia
OrganismsAll study participants
Number (%)IR (95% CI) (per 1000 person-years)
  1. *Coagulase-neg. staphylococcus = Coagulase-negative staphylococcus.

  2. †Unidentified GPB = Unidentified gram-positive bacilli.

  3. ‡B haem. Streptococcus Gp. F = beta haemolytic streptococcus group F.

  4. §Unidentified GNB = Unidentified gram-negative bacilli.

All blood cultures taken703 
Positive blood cultures159 (22.6)32.40 (26.20–40.60)
Streptococcus pneumoniae 68 (42.8)13.86 (10.10–19.50)
Non-typhi salmonellae 42 (26.4)8.55 (6.17–12.23)
Escherichia coli 9 (5.7)1.83 (0.77–5.42)
Coagulase-neg. staphylococcus* 10 (6.3)2.04 (1.12–4.11)
Salmonella typhi 6 (3.8)1.22 (0.43–4.97)
Staphylococcus aureus 6 (3.8)1.22 (0.56–3.22)
Haemophilus influenzae3 (1.9)0.61 (0.19–3.00)
Alpha haemolytic streptococcus 2 (1.3)0.41 (0.09–4.08)
Corynebacterium species2 (1.3)0.41 (0.09–4.07)
Unidentified GPB†2 (1.3)0.41 (0.09–4.07)
B haem. Streptococcus Gp. F‡1 (0.6)0.20 (0.03–1.45)
Non-haemolytic streptococcus 1 (0.6)0.20 (0.03–1.45)
Unidentified GNB§1 (0.6)0.20 (0.03–1.45)
Mycobacterium sp.1 (0.6)0.20 (0.03–1.45)
Other identified organism5 (3.1)1.01 (0.43–3.04)
Table 4.   Incidence of Streptococcus pneumoniae and Non-typhi salmonellae septicaemia by HIV serostatus, age, sex and CD4 counts
 No.IR (95% CI) per 1000 pyrsIRR (95% CI) (unadjusted)IRR (95% CI)
  1. IR, incidence rate; IRR, incidence rate ratio; ART, antiretroviral therapy.

S. pneumoniae    (adjusted for age, sex , CD4 countsand HIV/ART)
 Overall6813.86 (10.11–19.52)  
 HIV and ART status   P < 0.001 P < 0.001
  HIV negative41.50 (0.57–5.31)11
  HIV positive – not on ART5228.35 (20.08–41.37)22.78 (7.81–66.48)21.81 (7.76–61.30)
  HIV positive – on ART1229.59 (14.58–69.80)23.84 (7.24–78.47)21.64 (6.75–69.40)
 Sex   P = 0.007 P = 0.008
  Men4919.78 (13.48–30.21)11
  Women197.82 (4.59–14.43)0.40 (0.20–0.78)0.43 (0.23–0.80)
 Age group   P < 0.001 P = 0.002
  13–24 years22.37 (0.52–23.34)11
  25–34 years2915.06 (9.21–26.32)6.19 (1.41–27.30)5.09 (1.16–22.30)
  35–44 years2926.98 (16.75–46.20)11.22 (2.62–48.10)8.20 (1.99–33.80)
  45+ years87.51 (3.47–19.20)3.09 (0.63–15.30)3.66 (0.75–17.90)
 CD4 count group (cells/μl)
  <2002734.0 (23.30–49.50) P < 0.001 P = 0.005
  200–4992821.30 (14.70–30.90)0.67 (0.39–1.33)0.55 (0.32–0.94)
  500+135.09 (2.95–8.76)0.15 (0.08–0.29)0.33 (0.16–0.64)
Non-typhi salmonellae    (Adjusted for HIV/ART and CD4 count)
 Overall428.55 (6.17–12.23)  
 HIV and ART status   P < 0.001 P < 0.001
  HIV negative10.37 (0.05–2.66)11
  HIV positive – not on ART3820.70 (14.80–29.80)60.04 (8.00–450.70)60.04 (8.00–450.7)
  HIV positive – on ART37.39 (2.37–35.33)29.79 (3.07–289.20)29.79 (3.07–289.2)
 Sex   P = 0.21 P = 0.16
  Men176.86 (4.14–12.23)11
  Women2510.29 (6.70–16.62)1.54 (0.78–3.07)1.62 (0.83–3.19)
 Age group   P = 0.11 P = 0.30
  13–24 years67.13 (3.28–18.50)11
  25–34 years178.83 (5.33–15.72)1.24 (0.50–3.06)1.30 (0.53–3.16)
  35–44 years1513.95 (7.89–27.08)2.07 (0.77–5.52)2.12 (0.79–5.68)
  45+ years43.75 (1.42–13.32)0.54 (0.15–1.93)0.94 (0.27–3.28)
 CD4 count group (cells/μl)   P < 0.001 P = 0.01
  < 2001923.90 (15.20–37.50)11
  200–4991713.00 (8.06–20.90)0.61 (0.31–1.17)0.57 (0.29–1.10)
  500+62.35 (1.06–5.23)0.10 (0.04–0.26)0.27 (0.10–0.68)

Ceftriaxone sensitivity testing was introduced only in recent years, and all SPN isolates tested were susceptible. Almost all SPN isolates were susceptible to erythromycin. However, they showed varying degrees of resistance to all other antibiotics tested, in particular to cotrimoxazole to which 91% were resistant. All NTS isolates were susceptible to ciprofloxacin, while 75% were resistant to cotrimoxazole and 67% resistant to chloramphenicol (Table 5). There was no significant change in antimicrobial resistance over the study period (1996–2007), except regarding SPN resistance to tetracycline, which increased over time (based on few data). Data were not collected on cotrimoxazole prophylaxis received by HIV-infected participants, so its effect on the incidence of septicaemia and the antimicrobial resistance patterns was not examined.

Table 5.   Numbers and percentages of antimicrobial resistant Streptococcus pneumoniae and Non-typhi Salmonellae over three 4-year periods
Organism and antimicrobialStudy time periodTrend test
  1. N/A, not available.

S. pneumoniae 16242868 
  n (%) n (%) n (%) N (%) 
 CotrimoxazoleNot performedNot performed20/22 (92)20/22 (91)N/A
 Tetracycline3/16 (19)4/6 (67)Not performed7/22 (32)0.03
 AmpicillinNot performed5/14 (36)0/1 (0)5/15 (33)0.46
 Chloramphenicol1/14 (7)3/24 (13)0/3 (0)4/41 (10)0.7
 Penicillin6/16 (38)8/24 (33)8/28 (29)22/68 (32)0.8
 Erythromycin0/16 (0)0/24 (0)1/28 (4)1/68 (1)0.9
 CeftriaxoneNot performedNot performed0/110/11N/A
Non-typhi salmonellae 8201442 
 Cotrimoxazole5/7 (71)13/19 (68)12/14 (86)30/40 (75)0.5
 Ampicillin4/7 (57)13/20 (65)7/9 (78)24/36 (67)0.4
 Chloramphenicol4/7 (57)13/19 (68)5/7 (71)22/33 (67)0.9
 Ciprofloxacin0/8 (0)0/20 (0)0/13 (0)0/41 (0)N/A
 CeftriaxoneNot performedNot performed1/8 (13)1/8 (13)N/A

The numbers of participants treated with particular antimicrobials for SPN septicaemias were 21 (30.9%) (amoxicillin), 14 (20.6%) (ciprofloxacin), 13 (19.1%) (chloramphenicol) and 10 (14.7%) (erythromycin). One participant with SPN septicaemia and treated with amoxicillin before ART introduction died, giving a case fatality rate (CFR) of 1.5% and a pre-ART introduction CFR of 2.5% (1 death/40 cases). The numbers of participants treated with particular antimicrobials for NTS septicaemias were 23 (54.8%) (ciprofloxacin) and 7 (16.7%) (chloramphenicol). Six participants with NTS septicaemia died, giving an overall CFR of 14.3%. Before ART introduction, the CFR was 17.9% (5 deaths/28 cases) and after it was 7.1% (1 death/14 cases).


In this prospective cohort study, the incidence of septicaemia in the HIV-negative adults was 2.6 per 1000 pyrs. Despite a thorough Pubmed search, we did not find any study reporting on the incidence of septicaemia among the general population or in HIV-uninfected adults in developing countries in sub-Saharan Africa. A study in Kilifi, Kenya, found an incidence of septicaemia in children younger than 5 years of 11.9 per 1000 pyrs with 50% owing to invasive pneumococcal infection (Brent et al. 2006).

The incidence of septicaemia was significantly higher in HIV-positive than in HIV-negative participants. In HIV-positive participants, the septicaemia incidence was higher among those on ART than those not yet on ART. Participants in the first year on ART had a significantly higher septicaemia incidence than those not yet on ART, but even among those receiving ART for more than 1 year, the septicaemia incidence was still higher than among HIV-negative participants. This is no surprise as patients initiating ART with low CD4 counts still remain at a higher risk of septicaemia than those not yet eligible for ART (who still have a better immunity). We noted a significant decline in septicaemia incidence (from 121.4 per 1000 pyrs in the first year on ART to 37.4 per 1000 pyrs in subsequent years), as did studies in Spain (Pedro-Botet et al. 2002), Italy (Tumbarello et al. 2000) and France (Meynard et al. 2003). Longer follow-up is necessary to show the long-term impact of ART on HIV-related septicaemia incidence once patients have recovered immune competence.

We found a higher incidence of overall and SPN septicaemia among participants aged 25–34 and 35–44 years (compared to those aged 13–24 years) and in men, but this was not observed in NTS septicaemia. A review of epidemiological studies of sepsis in industrialised countries also concluded that men are more likely to develop sepsis than women (Moss 2005). However, this association has not been studied in developing countries in sub-Saharan Africa. Male sex and age are risk factors for sepsis in studies in industrialised countries (Hodgin & Moss 2008), with those aged 65 years and above at higher risk (Martin et al. 2006).

Our finding of SPN and NTS as the most frequent causes of septicaemia confirms those of hospital-based studies in Malawi (Gordon et al. 2001) and Uganda (Ssali et al. 1998). In the study in Uganda, the prevalence of SPN was similar among HIV-positive and HIV-negative in-patients (Ssali et al. 1998), but we found a significantly higher SPN septicaemia incidence among HIV-positive than among HIV-negative participants. Our population-based study had less selection bias (because of the close proximity of our study clinic to where the study participants live, provision of free medical care and transport to the clinic for those who needed it) than hospital-based studies, which usually recruit seriously ill patients.

The Uganda National treatment Guidelines (MOH Uganda & Uganda NDA 2003b) recommend that before sensitivity results are known, adult patients with suspected septicaemia should start treatment with intravenous gentamicin and either cloxacillin or chloramphenicol. We did not test for sensitivity to gentamicin, but we observed substantial resistance of SPN and NTS (the two commonest isolates) to chloramphenicol, one of the recommended antimicrobials. Our findings are in contrast to those of a hospital-based study in the Gambia (Hill et al. 2007), where SPN isolates were highly sensitive to penicillin, ampicillin and chloramphenicol. The reasons for these differences may be because of differences in antibiotics availability and use in the two countries. Similar high or increasing prevalences of NTS resistance to commonly used antimicrobials have been reported in Kenya (Kariuki et al. 2005) and Malawi (Gordon et al. 2008). Based on our findings, a combination of ceftriaxone or erythromycin and ciprofloxacin could be recommended for the management of adults with suspected septicaemia in our area.

The widespread use of cotrimoxazole for treating common ailments may explain the high resistance we found for this drug in both SPN and NTS isolates. Another explanation may be the use of cotrimoxazole prophylaxis for opportunistic infections in HIV-infected patients in line with WHO and Ugandan national guidelines (MOH Uganda 2005). However, we could not examine its effect on septicaemia incidence or antimicrobial drug resistance patterns as these data were not collected. Similar high levels of resistances to cotrimoxazole have been reported (Mootsikapun 2007). In Uganda, cotrimoxazole is still beneficial in preventing infections such as toxoplasmosis, Isospora belli diarrhoea and Pneumocystis carinii pneumonia (Watera et al. 2006).

The strong points of our study were: the comparison of septicaemia in HIV-infected participants at various stages of immunosuppression and in HIV-uninfected controls; the long follow-up period that enabled assessment of trends in septicaemia IR over time; changes in antimicrobial resistance; and the effect of ART introduction. We have not found other published studies corroborating the decreased CFR of NTS septicaemia after ART introduction. Although we changed blood culture method in May 2004 by introducing the Bactec fluorescent method (which enables rapid bacteria detection and therefore early changes in patient management), both methods have similar sensitivity. This change in the blood culture method applied equally to the three participants’ groups and is unlikely to have led to a selection bias in our findings.

The proportion of blood cultures successfully showing bacterial growth in our study (22.6%) was similar to that in other studies in Malawi (Gordon et al. 2001) (16.1%) and in Uganda (Ssali et al. 1998) (24%). The proportion of positive blood cultures depends on the criteria used for taking blood cultures from a sick patient and will be higher when blood cultures are only taken from patients with a higher pre-test probability of septicaemia. On the other hand, taking blood cultures from all patients with fever will result in a lower proportion of positive cultures, but fewer cases of septicaemia will be missed. Therefore, in resource-constrained settings, clinicians need guidelines on the criteria for blood culture, not only for individual patient care but also for the most cost-effective use of the limited available resources. Using the cut-off temperature of 38 °C for blood cultures, we might have underestimated the incidence of septicaemia as some patients with septicaemia present with temperatures below 38 °C, and furthermore axillary temperature is a weak predictor of infection (Brent et al. 2006).

We may have missed cases of malaria and septicaemia co-infection, as we restricted blood cultures to only those with negative malaria blood smears. We did not calculate the proportion of positive malaria smears among routine specimens from participants with and without a fever. A study in Nigeria found concomitant bacteraemia in 44 (35.2%) of 125 malaria cases, with the highest frequency of co-infection in the 31–40 and 6–10 years age groups (Ukaga et al. 2006). Although we did not collect data on NTS and malaria co-infection, studies of children in Kenya (Oundo et al. 2002), adults and children in Nigeria (Akinyemi et al. 2007) and adults in Thailand (Ellis et al. 2006) showed an association between malaria and Salmonella septicaemia.

Specimen contamination can obscure blood bacterial growth. In our study, the growth of 10 Coagulase-negative staphylococci and one Staphylococcus aureus (1.6% of all cultures) may have been owing to contamination with dermal flora during blood collection. This was lower than the 103 (43.3%) found in a hospital-based study in Ethiopia (Asrat & Amanuel 2001), and exclusion of these 11 contaminants from the analysis did not substantially change the overall septicaemia IR and IRR.

In conclusion, the association between septicaemia and lower CD4 counts regardless of ART means that clinicians and patients should be aware that HIV-infected patients continue to have a high risk of septicaemia after starting ART, until after achieving immune recovery. Starting ART earlier at higher CD4 counts is likely to lead to lower septicaemia incidence. Investment in laboratory surveillance systems to monitor antimicrobial resistance would help improve patient care.


We thank the clinical team of the MRC/UVRI Rural Clinical Cohort study, the data and statistics team, the laboratory staff and the study participants without whom this report would not have been possible. This study was supported by the Medical Research Council (MRC) of the UK.