Etiology, Antibiotic Susceptibility and Prognostic Factors of Pediatric Community-Acquired Sepsis in Addis Ababa, Ethiopia

Background: There is a scarcity of data on pediatric community-acquired sepsis (CAS) in Ethiopia. We sought to determine the etiology, antibiotic susceptibility pattern, and prognostic factors in children with CAS in Addis Ababa, Ethiopia. Methods: A prospective cross-sectional study of 101 children aged 0-15 yrs with suspected CAS was performed. Blood culture, antibiotic susceptibility testing, Polymerase Chain Reaction (PCR) of the autolysin (lytA) gene from whole blood samples and typing Streptococcus pneumoniae by sequencing the cpsB gene and using Quellung reaction were performed. Data were analyzed using descriptive statistics and binary logistic regression. Results: The prevalence of culture-positive CAS was 14.9 % (15/101). S. pneumoniae (26.7%) and Klebsiella pneumoniae (26.7%) were the most common causes of CAS. The four isolated pneumococci belonged to serotypes 19A (n = 2), 33C and 12F. Half of K. pneumoniae isolates were resistant to gentamicin and ceftriaxone. The case-fatality rate was 11.9% (12/101). In univariate analysis, age of 28 days - 1year (odds ratio (OR), 0.1; 95% condence interval (CI), 0.01-0.98; P, 0.048), body temperature of 37.5 0 C - 38.5 0 C (OR, 0.2; 95% CI, 0.05-0.82; P, 0.026) were negatively associated with mortality. Presence of underlying comorbidity (adjusted odds ratio (aOR), 6.8; 95% CI, 1.59-28.7; P, 0.009) was an independent predictor of mortality. Conclusions: 0.05. Results with 95%

Conclusions: Streptococcus pneumoniae and Klebsiella pneumoniae were the major causes of CAS and there was a substantial level of antibiotic resistance among isolates. Presence of underlying comorbidity was a predictor of mortality. Large scale studies on etiology, antibiotic susceptibility pattern and prognostic factors of CAS in Ethiopia are warranted.

Background
Sepsis remains a huge burden and a leading cause of childhood morbidity and mortality worldwide. Although Africa is likely to account for a signi cant proportion of the global burden of sepsis, there are limited reports on the epidemiology, management and outcomes of sepsis in African countries. 1 Information about etiological agents involved in causing community-acquired bacteremia and their antibiotic resistance pattern will help to improve empiric antibiotic therapy and consequently patient outcomes. 2 In developing countries, Staphylococcus aureus, Klebsiella spp. and Escherichia coli account for 55% (95% con dence interval (CI) 39-70%) of culture positive sepsis among neonates, while in infants the most prevalent pathogens are S. aureus, E. coli, Klebsiella spp., S. pneumoniae and Salmonella spp., which account for 59% (95% CI 26-92%) of culture-positive sepsis. 3 The distribution of etiological agents of sepsis in the pediatric population is changing considerably as a result of introduction of conjugate vaccines such as Haemophilus in uenzae type b (Hib) vaccine and pneumococcal conjugate vaccines (PCVs) and it has become important to understand the role of childhood vaccinations not only on the rate and distributions of causative organisms but also on the risk factors and long-term outcomes of sepsis. 4 Case-fatality rates due to CAS range from 6% to 13% and factors such as age, severity of illness at onset of sepsis, presence of comorbidities, bacteremia and etiology have been associated with mortality. [4][5][6][7] PCV10 was introduced in Ethiopia in October 2011 as a three dose primary series (3p+0) without any booster dose 8 . A study by Muhe and colleagues on CAS in children younger than three months of age in Addis Ababa Ethiopia performed two decades before the introduction of PCV10 in the country has identi ed S. pneumoniae as the predominant cause of sepsis. 9 There is however a scarcity of data on the etiology of sepsis and the role of S. pneumoniae as a cause of sepsis after the introduction of PCV 10 in the country.
Most of the previous studies on childhood sepsis in Ethiopia, have been on neonatal sepsis and do not distinguish clearly between community-acquired (CA) and hospital-acquired (HA) sepsis. [10][11][12] Besides, S. pneumoniae infections including sepsis are often quite rare in neonates. 13 The aim of this study was therefore to determine the etiology, role of S. pneumoniae as a cause of sepsis, antibiotic susceptibility pattern of isolates and prognostic factors in children with CAS, ve years after introduction of PCV10 in Ethiopia.

Study design and setting
We carried out a prospective cross-sectional study from September 1, 2016 to August 30, 2017. The study was carried out in pediatric emergency departments of two large hospitals in Addis Ababa, Ethiopia; Tikur Anbessa Specialized Hospital (TASH) and Yekatit 12 Hospital.

Participant selection and inclusion criteria
Among children between the ages of 0 to 15 years presenting to the two pediatric emergency departments (PED) during the study period, the ones that were included in the study were, those presenting with suspected CAS and who had not taken antibiotics within two weeks prior to presenting to the hospital. CAS was de ned as a case of suspected sepsis in children with no hospital or health care admissions in the two weeks prior to the current admission 14 and identi ed from samples taken within 48 h of admission. 15 Suspected sepsis was categorized based on the clinical decision of the attending physician and was de ned as meeting the systemic in ammatory response syndrome (SIRS) criteria. 16 Data collection and outcome measurement Trained research nurses approached parents/guardians of 101 children, suspected with CAS during the study period. After obtaining informed consent, a structured questionnaire was used to obtain sociodemographic data and other relevant clinical data. List of antibiotics that were used for management of patients was extracted from medical records. Final outcome (discharge or death) was registered and length of stay in the hospital was recorded in days. The microbiological outcomes assessed were: culture positive CAS, pneumococcal CAS and antibiotic susceptibility pattern among isolates.

Laboratory procedures
Sample collection At enrollment, venous blood (1 mL for < 1 month-olds and 2-5 mL for > 1 month-olds) was drawn aseptically and transferred into 20 ml Brain Heart Infusion (BHI) broth (Oxoid, Cheshire, England) bottles and mixed gently by inverting the bottle. In addition, 1-2 mL of blood, collected using Ethylenediaminetetraacetic acid (EDTA) vials was available for 69 children and was frozen at -80 °C for PCR.

Culture and identi cation
The inoculated BHI broth was cultured aerobically. After 24 hrs of incubation, Gram stain was done followed by subculture on Blood agar, Chocolate agar and MacConkey agar (Oxoid). Culture bottles that did not show growth were further incubated for 7 days and subcultured before being reported negative. Initial identi cation of bacteria was made by Gram stain, hemolytic activity on sheep blood agar plates, optochin sensitivity, bile solubility, coagulase test, colony morphology on MacConkey agar and growth on Mannitol salt agar at the Armauer Hansen Research Institute (AHRI), Addis Ababa, Ethiopia. Further identi cation was then performed on all isolates by Matrix-assisted laser desorption/ionization time of ight mass spectrometry at Ghent University, Ghent, Belgium. Coagulase negative staphylococci (CoNS), Micrococcus spp, and viridans streptococci were considered as contaminants when identi ed in the blood cultures.

Antibiotic susceptibility testing
Antibiotic susceptibility testing on a selected panel of antibiotics that are used locally was done using the Kirby-Bauer disk diffusion method. 17 For pneumococcal isolates, penicillin resistance was initially measured using oxacillin discs and for isolates with zones ≤ 19 mm, minimum inhibitory concentrations (MICs) were determined using E-Test strips (bioMerieux, Marcy-l'Etoile, France). Test results of both disc diffusion and MICs were interpreted according to the Clinical and Laboratory Standard Institute (CLSI) criteria. 18 American Type Culture Collection (ATCC) strains: Escherichia coli ATCC 25922, Staphylococcus aureus ATCC 25923 and Streptococcus pneumoniae ATCC 49619 were used for quality control. DNA extraction and ampli cation of lytA genes DNA extraction was performed on available whole blood samples obtained from 69 children. Initial pretreatment of the samples was performed as described previously. 19 DNA extraction was then carried out with the DNeasy® 96 Blood & Tissue Kit (Qiagen, Venlo, The Netherlands), following manufacturer's instructions. Ampli cation of a 101-bp fragment of the lytA gene was carried out as previously described. 20 PCRSeqTyping and serotyping DNA extraction from S. pneumoniae isolates was carried out by alkaline lysis as described previously. 21 Ampli cation and sequencing of the 1061 bp of the cpsB-gene (encoding for phosphotyrosine phosphatase) were done using PCRSeqTyping as previously described. 22 Twenty μl of the amplicons were sent for sequencing to GATC Biotech (Constance, Germany) and sequencing was performed using the Sanger sequencing technique. cpsB sequences were used to interrogate GenBank database (http://www.ncbi.nlm.nih.gov/blast). Serotype was assigned based on a BLAST bit score of > 99% sequence identity of the query cpsB nucleotide sequence with the reference sequences from GenBank. Because correct serotype could not be assigned by PCRSeqTyping for two of the isolates, serotyping was performed on all isolates by the Quellung reaction 23 using pool and group antisera, obtained from the Statens Serum Institut (Copenhagen, Denmark).

Statistical analysis
Data were initially entered in ReDCap (Vanderbilt, USA), exported into Excel and analyzed using PASW Statistics 20 software (SPSS Inc., Chicago, Ill. USA). Age was strati ed into four groups: < 28 days, 28 days-1 year, 2-5 years and ≥ 6 years. Continuous variables were reported as median and interquartile range (IQR). Descriptive statistics was used to analyze sociodemographic, clinical and laboratory characteristics. To identify prognostic factors of CAS, bivariate analysis using binary logistic regression was performed. Variables which were signi cant at P < 0.1 were then used in multivariable model. Variables in the multivariable analysis were considered as signi cant when P < 0.05. Results from the binary logistic regression analyses were reported as ORs and aORs with 95% CIs.

Ethical approval
The study procedures were in accordance with the Helsinki Declaration. The study was approved by the AHRI/All Africa Leprosy Rehabilitation and Training Hospital (ALERT) Ethics Review Committee, Addis Ababa University Institutional Review Board, Yekatit 12 Medical College Ethics Committee and the National Research Ethics Review Committee (No. 310/194/2017). The parents and/or guardians of all participants gave written informed consent.

Characteristics of the study participants
Among 377 children with suspected sepsis admitted to TASH and Yekatit 12 hospitals during the study period, 101 (26.8%) met the study inclusion criteria and were enrolled ( Table 1). The median age of the enrolled children was 2 months IQR (1 -5.5) with most children (65.3%) aged between the age of 28 days and 1 year; and 58 (57.4%) were boys.
The most common sign was fever (> 37.5 °C), which was seen in 91.1% of the children, followed by irritability (74.3%) and vomiting (70.1%) ( Table 2). A total of 32.7% of the children had comorbidities, with gastrointestinal comorbidities (8.9%) being the most frequent. The focus of infection was not determined in 35.6% of the cases and, in those with a focus; lower respiratory tract infections (44.6%) were the most common. In most children who presented to the two emergency departments (53.5%), the number of days since onset of symptoms reported by parents/guardians was < 3. Based on the data obtained from vaccination cards or words of parents/guardians, 69.3% of the children have received at least one dose of PCV10. Malnourishment (moderate or severe) was seen in 39.6% of the children.

Use of antibiotics
The most common antibiotics used for the management of children with CAS were a combination of ampicillin with gentamicin, 47.5% (48/101) followed by ceftriaxone, 16.7% (17/101) (Figure 1).
Combinations of ampicillin and gentamicin and /or along with an additional antibiotics were used in 59.4% (60/101) of the cases. Ceftriaxone alone or along with other antibiotics were used in 41.6% (42/101) of the cases. Combinations of ampicillin/gentamicin or ceftriaxone alone or along with other antibiotics were used in 96% (97/101) of the cases overall.

Antibiotic susceptibility
All of the four S. pneumoniae isolates were susceptible to penicillin while two were resistant to trimethoprim/sulfamethoxazole and chloramphenicol, while one was resistant to erythromycin and none to tetracycline ( Table 3). Half of the K. pneumoniae isolates were resistant to gentamicin and ceftriaxone.
Pneumococcal serotypes and lytA PCR Using PCRSeqTyping, only two of the isolates were typed to serotype level (19A (n=2)) while the other two were only typed to subtype level (44, 12B/F, 32A/F and 35A/B/C/F, 33A/C/F). All isolates were then serotyped using the Quellung reaction. The ones initially identi ed to subtype level by PCRSeqTyping were identi ed as serotypes 12F and 33C and there was 100% concordance between PCRSeqTyping and Quellung for the remaining two isolates. All the four pneumococcal isolates were non-PCV10 types. Three of the four children had been vaccinated with PCV10; one of the three has been fully vaccinated while the other two have received two doses. Whole blood was available for only two of the four cases from which pneumococci was isolated and only one was lytA-PCR positive.

Factors associated with mortality
The overall case-fatality rate was 11.9% (12/101). The length of stay for the children who died was 5 (IQR: [3][4][5][6][7][8][9][10][11][12][13][14] days. Among 29 variables in Table 1 and Table 2 which were used for bivariate analysis using binary logistic regression, three were signi cant at P < 0.1 and were further used in the multivariable analysis (Table 4). Seven of the children who died (i.e. 58.8% of mortality cases), were aged between 28 days and 1 year, which is also the age group of most of the children in the study (65.3%). However, in bivariate analysis, compared to children aged > 6 years, children within this age group were less likely to die (OR, 0.1; 95% CI, 0.01-0.98; and P, 0.048) ( Table 4). Children with body temperature of 37.5 °C -38.5 °C were also less likely to die (OR, 0.2; 95% CI, 0.05-0.82; P, 0.026) compared to those with > 38.5 °C. In a multivariable analysis, children with comorbidities were 6.8 times more likely to die than children without comorbidities (aOR, 6.8; 95% CI, 1.59-28.7; P, 0.009). A pathogenic bacterium was isolated from only one of the children who died and it was a multi-drug resistant K. pneumoniae. There was no association between culture positivity and mortality (P, 0.459).

Discussion
Results from our study indicate that the prevalence of culture-positive sepsis was 14.9%. This nding is lower than results obtained in other studies in Gondar (32.1%) and Addis Ababa (27.9% and 44.7%), Ethiopia 10,11,24 , in which most of the cases concern neonatal sepsis with nosocomial origin. Our nding is however higher than a study on CAS in Philippines (3.4%). 5 S. pneumoniae (26.7%) and K. pneumoniae (26.7%) were the two most common causes of sepsis. The prevalence of pneumococcal sepsis was 3.96% (4/101). Similarly, in a previous study in Addis Ababa, Ethiopia long before the introduction of PCV10; S. pneumoniae was identi ed as the most common cause of sepsis in children younger than 3 months of age while K. pneumoniae was uncommon. 9 A recent report from Switzerland has indicated that after widespread vaccination with PCV7 and a transition to PCV13 in 2011, incidence of pneumococcal sepsis still remained substantial and was responsible for 25% of the CAS episodes in 2015. 25 All four S. pneumoniae isolates from this study were non-PCV10 serotypes with two of them being serotype 19A. In countries that have introduced PCVs, there is an increase in diseases due to non-vaccine serotypes. 26 In Brazil, after introduction of PCV10, the proportion of serotype 19A among invasive pneumococcal disease strains increased from 2.8% in 2005-2009 (pre-PCV period) to 16.4% in 2016-2017 (6 to 7 years post-PCV).27 Lack of comprehensive pre-vaccine serotype data in Ethiopia and the limited sample size and isolates in the current study means that we could not make comparisons. Our ndings however warrant further large scale studies in Ethiopia to assess the role S. pneumoniae as a cause of sepsis in the post-PCV period.
Although they did not clearly distinguish between community-acquired and hospital-acquired cases, various studies on neonatal sepsis in Africa 28-31 have identi ed Klebsiella spp. as the most common cause. Kabwe and colleagues who analyzed 13 studies on neonatal sepsis in Africa, have reported that Klebsiella spp. account for 32% (323/1009, range 0-59%) of the isolates identi ed. 32 We performed lytA PCR for detection of S. pneumoniae from whole blood samples that were available from 69 patients. Enhanced diagnosis of pneumococcal bacteremia by using lytA PCR which resulted in up to 10.7% increased yields compared to blood culture has been previously reported. 19 In our study lytA PCR did not increase sensitivity. Whole blood was available for two of the samples from which S. pneumoniae were isolated and only one was positive for lytA PCR. Possible reasons for low sensitivity could be the small number of samples and limitations in transport and storage of samples.
The Ethiopian standard treatment guideline recommends ampicillin and gentamicin as a rst line and ceftriaxone as a second line antibiotic therapy for pediatric sepsis. 33 Among 96% of the cases either the rst line antibiotics with or without additional antibiotics or the second line antibiotics with or without other antibiotics were used for management of the patients. Among K. pneumoniae isolates in this study, half were resistant to gentamicin and ceftriaxone. A recent review by Williams and colleagues revealed a high prevalence of Klebsiella spp non-susceptibility to gentamicin (median 49%, IQR 48-58%) and ceftriaxone (range 33-50%) in sub-Saharan Africa. 34 In developing countries, amikacin, which is effective against most MDR Klebsiella spp. and has comparable cost to gentamicin, has been recommended as an alternative to gentamicin as second-line treatment in combination with penicillin. 3 Our results indicate the need for continued surveillance of antibiotic resistance in Ethiopia and assessment of treatment options.
In the pediatric consensus de nition of SIRS, abnormal core temperature is de ned as a temperature of < 36 °C or > 38 0C measured by rectal, bladder, oral, or central catheter probe and a temperature of > 38.5 °C is used as it increases speci city 16 In our study, we used the local de nition of fever (> 37.5 °C) measure via axillary route. We compared the nal outcome of children with local de nition of fever (> 37.5 0C) and fever as per the SIRS criteria (> 38.5 °C). In a univariate analysis, we were able to see that children with body temperature between 37.5 °C and 38.5 °C were less likely to die than children with body temperatures of > 38.5 °C. In a similar study performed in Dar es Salaam, Tanzania, George and colleagues found that children with a temperature of > 38.5 °C and > 2 SIRS criteria were seven times more likely to die (OR, 7.05; P, 0.01). 35 The case-fatality rate in this study (11.9%) was higher than ndings from European childhood lifethreatening infectious disease study report (6%, increasing to 10% in the presence of septic shock) 6 and was slightly lower than a report from Philippines. 5 One of the noteworthy ndings of this study is that children with underlying comorbidity were 6.8 times more likely to die than those without. Similarly, in a study on pediatric severe sepsis in the US, case-fatality rate was reported to be signi cantly higher in children with underlying comorbidity. 4 We acknowledge some limitations to our study. The sample size was small and more patients and data would have been useful for our analysis. Because the study hospitals are tertiary care facilities with shortage of beds, a selection bias might have occurred due to the possibility that children with comorbidities are preferentially admitted while those without comorbidities are referred to other health care centers. Since we did not perform serial blood cultures, the rate of culture positivity in the present study could be an underestimation. Limitations in proper transport and storage of whole blood might have also affected the results of lytA PCR.

Conclusions
The study identi ed that S. pneumoniae and K. pneumoniae are the main etiological agents of communityacquired sepsis among children at the two pediatric emergency departments. All the pneumococci isolates were non-PCV10 serotypes. There was a high rate of antimicrobial resistance to rst and second line    Legend: a Among 29 variables in Table 1 and Table 2 used for bivariate analysis using binary logistic regression, 3 were significant at P < 0.1 and were further used in the multivariable analysis.
b P-value of bivariate and multivariable logistic regression analyses, those in bold are with P < 0.05. Figure 1 Types of antibiotics and their frequency of usage for the management of CAS in Ethiopian children aged 0-15 yrs.