Discussion

This is the first study describing the epidemiology of CO-BSIs at CMJAH EDs. The positivity rates in our study ranged between 14% and 20% and were higher than those described in other studies. ^{Reference Lamy, Dargère, Arendrup, Parienti and Tattevin23} Contamination rates in our study (ie, 5%–14%) were higher than the internationally acceptable standard of 2%–3%. ^{Reference Wilson24} Gram-positive bacteria were the predominant pathogens in our pediatric cohort. Gram-negative bacteria dominated among adults and the elderly. E. coli was the most prevalent gram-negative pathogen across all age groups. Despite S. aureus showing preponderance across all age groups, the prevalence of MRSA was low. Although the prevalence of ESBL-producing Enterobacterales and CREs was low, high rates of resistance of gram-negative bacteria to currently prescribed empiric antibiotics were noted among neonates, adults, and the elderly.

Limiting collection of blood cultures to patients with a high likelihood of BSI and adherence to blood culture collection guidelines can assist in decreasing contamination rates. ^{Reference Ntusi, Aubin, Oliver, Whitelaw and Mendelson25,Reference Doern, Carroll and Diekema26} The possible reason for the higher positivity rate in this study could be that the patients who present to this institution usually have a greater severity of illness and are often bacteremic. The high volume of patients and inadequate resources possibly results in suboptimal aseptic techniques, thus increasing the contamination rate.

In our study, the majority of pediatric BSIs occurred among infants. As reported in a South African study that reviewed community-acquired BSI in children aged 0–14 years, gram-positive bacteria predominated in our pediatric cohort, too. ^{Reference Dramowski, Cotton, Rabie and Whitelaw12} More than half the isolates responsible for BSI among children of all age groups were gram-positive bacteria. Similar to other studies, S. aureus was a predominant pathogen across all age groups in our study. ^{Reference Laupland, Lyytikäinen and Sgaard4,Reference Dramowski, Cotton, Rabie and Whitelaw12,Reference Friedman, Kaye and Stout27,Reference Young, Chen and Yen28} S. pneumoniae, reported as the most common pathogen associated with CO-BSI among African children, ^{Reference Reddy, Shaw and Crump2} was prominent among infants and older children. South African surveillance data have shown that invasive pneumococcal disease (IPD) in children aged <5 years is predominantly due to nonvaccine serotypes, whereas in older individuals it includes both vaccine and nonvaccine serotypes. ²⁹ S. pneumoniae BSI was higher among adults than among the elderly, which could be related to the higher incidence of HIV among adults. ³⁰ Children and adults at risk for IPD should be adequately vaccinated to protect against severe disease. Although S. agalactiae was predominantly seen among neonates, it was also identified among infants, adults, and the elderly. This finding contrasts with those of other studies in which it was identified exclusively in neonates and adults aged >60 years. ^{Reference Moon, Ko, Park, Hur and Yun9}

Enterobacterales were the commonest group of gram-negative pathogens with E. coli predominating across all age groups, consistent with the findings of several studies. ^{Reference Dramowski, Cotton, Rabie and Whitelaw12,Reference Friedman, Kaye and Stout27,Reference Young, Chen and Yen28} The most prevalent isolate in CO-BSI among adults in Africa, nontyphoidal Salmonella, ^{Reference Reddy, Shaw and Crump2} was uncommon in our study, similar to another study from South Africa. ^{Reference Dramowski, Cotton, Rabie and Whitelaw12} This difference is likely due to the improved management of HIV in South Africa, a known risk factor for nontyphoidal Salmonella infections. The high prevalence of HIV (19%) among adults in South Africa, a well-established predisposing factor for invasive cryptococcal disease, could be the reason for the high rate of recovery of C. neoformans from BSI in this group. ^{Reference Lin, Shiau and Fang31,32}

Although gram-positive bacteria (excluding S. aureus) had good susceptibility rates to ampicillin among neonates, the susceptibility of the most common gram-negative bacteria to the current empiric therapy of ampicillin in combination with gentamicin was low (68%). For this reason and the high prevalence of MSSA (17%), cefotaxime monotherapy is recommended with the addition of amikacin to be considered in neonates with high risk for MDR pathogens. This risk includes prior exposure to healthcare or antibiotics, particularly in very low birth-weight and premature neonates.

The high prevalence of MSSA in older children supports the current South African empiric therapy guideline of ceftriaxone. Treatment should be changed to cloxacillin if culture confirms S. aureus BSI. When meningitis is not the reason for admission, amoxicillin- clavulanate is a suitable alternative.

In this study, gram-negative bacteria were the predominant cause of BSI among adults and elderly patients. Ceftriaxone is recommended for the empiric treatment of CO-BSI among adults. Consistent with previous reports, increasing AMR with increasing age was evident in our study. ^{Reference Lee, Wang and Lee7} The elderly had higher proportions of AmpC, ESBL-, and CRE-producing Enterobacterales compared to adults and children. This finding could be related to increased comorbidities, exposure to health care, and antimicrobial use among the elderly. The addition of amikacin to ceftriaxone as a second agent is recommended in this population due to the high risk of MDR pathogens.

To provide a conclusive recommendation on empiric therapy for surgical patients, further studies focusing on this cohort with additional clinical information is required.

Although MRSA was isolated in several age groups, the prevalence was lower (0%–3%) than that described in other studies, in which it varied from 6% to 37%. ^{Reference Naidoo, Nuttall, Whitelaw and Eley5,Reference Moon, Ko, Park, Hur and Yun9,Reference Young, Chen and Yen28} Empiric antibacterial cover for MRSA is not recommended in this population.

The prevalence of ESBL-producing Enterobacterales showed wide variation (1%–9%) between the different EDs, with the highest prevalence in our study surpassing levels of 7% to 8% described elsewhere. ^{Reference Blomberg, Manji and Urassa6,Reference Moon, Ko, Park, Hur and Yun9}

The potential source of infection was not identified for most BSIs. Although cultures from sites other than blood were collected in 75% of neonatal cases, <40% of BSIs had concomitant cultures from the other EDs. Meningitis was the most common focus among children, and UTI was the most frequently identified source among adult, elderly, and surgical patients. Identification of the focus of infection is important to guide empiric antimicrobial therapy because the prescribed drugs must achieve adequate tissue concentrations at the site of infection. ^{Reference Strich, Heil and Masur33} Efforts should be made to submit samples from EDs for microbiological investigation from the most likely source of sepsis prior to the initiation of antibiotics. Due to the frequent presentation of nonspecific symptoms in neonates, identification of a source can be challenging. Cultures from cerebrospinal fluid and urine are advisable in the absence of an obvious source in this group. However, the collection of samples from other sites should not delay the initiation of antibiotics in critically ill patients with sepsis or septic shock. ^{Reference Strich, Heil and Masur33}

The findings of this study provide evidence for continued pathogen surveillance at hospital EDs, and they support the need for current guideline revision for empiric treatment of CO-BSI among adults and children.

This study had several limitations. It was conducted in a single tertiary-care center; the findings may not be universally applicable. Due to the retrospective nature of the study, the current epidemiology and AMR might not be adequately reflected. Although all blood cultures analyzed were collected in the EDs, recent contact with other healthcare facilities and other risk factors for MDR pathogens could not be determined; thus, healthcare associated infections could have been inadvertently included. Adequacy of blood culture collection techniques, such as collection prior to initiation of antimicrobial therapy and volume of blood collected, both of which could lead to underestimation of BSI, could not be ascertained. Patients charts which could provide additional information regarding source of infection and predisposing factors for BSI (diabetes, immunosuppression, etc) were not reviewed. Lastly, the outcomes associated with empiric therapy were not available. The recommendations made above should thus be applied judiciously based on the clinical context.

Ongoing local and national surveillance of the epidemiology and AMR in CO-BSI is recommended to ensure that guidelines for empiric therapy remain appropriate. Follow-up studies are needed to determine whether implementation of these recommendations affects the clinical outcomes of patients with CO-BSI.