In normal conditions, blood is sterile. Severe localized or systemic infections can cause micro-organisms to enter the bloodstream through the lymphatic system. This presence of bacteria in the bloodstream is called “bacteremia.” Most of the time, these bacteria are cleared quickly by the immune system. In the case of overwhelming infections or intravascular focus of infection, the immune system may be unable to clear the bacteria from the blood, resulting in a bloodstream infection (BSI) (3). The micro-organisms responsible for this infection can be identified by blood culture. A blood culture consists of a blood sample from a patient, suspected to have a BSI, which is inoculated into a specialized blood culture bottle containing a broth (i.e., liquid) medium that supports optimal growth of bacteria. The concentration of bacteria in the blood of patients with BSI is very low (4), therefore direct culture on an agar plate cannot detect the presence of bacteria in the patient's blood. Once the blood is inoculated into the blood culture bottle, further amplification of the bacteria can take place, ultimately leading to visible bacterial growth. When growth of bacteria is detected in the blood culture bottle, a Gram stain of the blood-broth mixture is done to confirm presence of micro-organisms and distinguish between Gram-positive, Gram-negative bacteria and yeasts (Table 1). This and other microscopic information, such as shape and configuration of the bacteria, can orient the clinician to the identification of the bacteria causing the BSI and hence to the most appropriate antibiotic treatment. A subculture of the blood-broth mixture on an agar plate is then done to obtain colonies of the pathogen, on which further identification and antibiotic susceptibility testing can be performed. See Figure 1 for a visualization of the blood culture workflow and Table 2 for definitions regarding blood cultures.

Blood cultures must be obtained whenever there is a clinical suspicion of BSI. However, it is currently not clear which clinical signs are good predictors for BSI. Many predictive models to optimize the yield of blood cultures have been proposed, but so far only two of these have been shown to reliably distinguish between high (>30%) and low (<3%) risk of BSI, according to a 2015 systematic review (8). Of these, one model is a computerized system using a causal probabilistic network with input of many different variables (9), compromising its possible use in LMICs, and the other has been validated only for patients with community-acquired pneumonia (10). Moreover, none of these validated models are used in routine clinical practice, possibly because they are too complicated to calculate at the bedside and because of their reliance on laboratory data that are not available at the time of blood sampling (8).

Other well-studied clinical predictive models are the SIRS criteria (systemic inflammatory response syndrome) and the Shapiro criteria (11, 12) (see Figure 2). Both these models have demonstrated high sensitivity but low specificity (14–16) and rely on laboratory parameters that are not readily available in most LMICs. Their usefulness in daily practice is therefore rather limited.

In 2016, the Third International Consensus Definitions Task force described an easy clinical score to identify patients at risk of sepsis, the quick Sequential Organ Failure Assessment (qSOFA) score (13) (see Figure 2). Sepsis is defined as life-threatening organ dysfunction caused by a dysregulated host response to infection (17). It can occur with or without bloodstream infection. The predictive validity of the qSOFA for in-hospital mortality outside of intensive care units was higher than that of the SOFA and SIRS criteria (which are more elaborate). However, as qSOFA was validated on patients already suspected of infection, it cannot be used to differentiate patients with infection from those without infection.

In Figure 3, we propose a set of clinical indications for sampling blood cultures in LMICs. These indications were based on known common causes of BSI (3), the qSOFA criteria (13) and experience within our research network (18, 19). For neonates, blood culture indications are different and more complex; we refer to other resources for more information (20–22).

The process of collection of blood for blood culture is very important for the quality of the results (23). It requires applying a tourniquet on the arm, palpating the vein that will be used for the sampling, and applying appropriate antiseptics at the place of sampling. After antisepsis, the vein should not be touched anymore unless wearing sterile gloves. Next, the vein is pierced with either a needle and syringe or a butterfly needle. A sufficient volume of blood is aspirated either directly into the blood culture bottle (with certain bottle types and use of butterfly needle) or into a syringe and next divided over the blood culture bottles.

Further processing of cultures is described in this text with an emphasis on so-called “manual” blood culture systems (i.e., not making use of automated equipment). These systems rely on the use of appropriate blood culture bottles, which are placed in a conventional static incubator and are inspected daily to visually detect signs of growth of micro-organisms. This contrasts with the automated systems, which over the past decades have largely replaced manual blood culture systems in HICs and are the current standard (23). In these systems, the carbon dioxide production by micro-organisms in the bottles is continuously monitored, either by colorimetric or fluorescent detection. During incubation, the bottles are continuously agitated in the automated equipment. The current automated systems show better performance than manual systems in terms of yield and especially speed of growth, as recent studies from LMICs such as Egypt, Pakistan, and India have shown (24–26). However, these automated systems are costly, require regular maintenance and are not adapted to tropical, dusty environments, impeding the sustainable implementation of this technique in many LMICs outside of study sites, large reference laboratories or private laboratories in the capital cities (1, 27). Therefore, manual blood culture bottles are still most frequently used in LMICs and contribute worldwide to approximately the double of market share compared to automated systems (27). These manual blood culture bottles will be discussed and evaluated keeping the World Health Organization's (WHO) ASSURED criteria for diagnostics in LMICs in mind; i.e., they must be affordable, sensitive, specific, user-friendly, robust, rapid, equipment-free, and delivered to those who need it (28). Recently, the same research group proposed to add two criteria to this list: the REASSURED criteria put an additional emphasis on real-time connectivity and ease of specimen collection (29).

The most important feature of a blood culture bottle is that it should adequately support bacterial growth. For this purpose, a variety of peptide broths and additives are available (see Table 3). There is no single “optimal” broth medium; most broths contain dextrose and animal-derived complex peptide molecules. Broth media which support growth of a wide range of bacterial species include tryptic soy broth, also known as soybean-casein digest broth, supplemented peptone broth and brain heart infusion broth. Other commonly used broths such as thioglycolate broth, thiol broth, Columbia and Brucella (hypertonic) broth are also adequate for bacterial recovery (69, 88). For anaerobic bacteria, Columbia broth, pre-reduced peptone broths, thioglycolate broths, and thiol broths are advantageous, supposedly based on their low redox potential (76, 77, 89–91).

Regarding the volume of broth, a blood-broth ratio of 1:5 to 1:10 should be respected to optimize growth (92). In children, higher blood-broth ratios (e.g., 1:50 to 1:100) are acceptable (93).

Supplementation of the broth medium can further promote growth. A number of additives have been defined as growth enhancers in blood cultures, but the effect of additives is limited to the organism targeted for growth (see Table 3).

One of the most important additives is sodium polyanethole sulfonate (SPS). This is an anticoagulant, which in addition has been shown to have a stabilizing effect on microbial growth in blood cultures (69, 88). The typical concentration of SPS ranges from 0.025 to 0.05% (3, 77, 88, 94). The utilization of SPS has greatly reduced time-to-detection for many bacteria, and it is generally agreed to be an indispensable component of blood cultures.

Another commonly used additive is saponin, which is added in some blood culture systems as a lytic agent with subsequent centrifugation (Isolator system from Abbott Laboratories, Chicago, USA). A 1998 study also showed an increased recovery rate with shorter time-to-detection with the addition of saponin to the blood culture media of an automated system (95). Saponin is also widely used in anaerobic bottles in the BACTEC automated blood culture system, in combination with resin-supplemented aerobic bottles (96)³.

All guidelines on sepsis management stress the importance of sampling blood for blood culture before administration of antibiotics (37). The negative impact of antibiotic use prior to sampling on the yield of blood cultures is substantial, as the presence of antibiotics in the serum can inhibit the growth of bacteria (97–100). However, patients in LMICs are often already under antibiotic treatment before presenting to the hospital (101), since antibiotics are often readily available over-the-counter in these settings (102–104). To counteract the effect of antibiotics on growth of bacteria in the broth, antibiotic removal devices have been in use since 1982 (105). Nowadays, this role is largely taken up by resins.

Antimicrobial-binding resins and charcoal are generally considered effective in increasing microbial recovery rates. More microorganisms, particularly staphylococci and yeasts, are recovered from formulations with these additives, compared to non-supplemented formulations (106–109). There is no firm clinical evidence that the increased microbial recovery rates are due to inhibition or removal of antibiotic substances, although in-vitro research has extensively shown this to be the case (69, 110–113). Most resins are formulated as highly porous polyvinyl and benzene within a spherical bead; there are cationic ion-exchange resins and polymeric adsorbent resins (77). In addition to binding of antimicrobial agents, resin beads provide additional surface area for bacterial growth and help filter and bind components of the complement cascade (69). The exact formulations and balance between ratios of nutrients and resins is often proprietary and thus unknown.

The drawbacks of adding resins to a blood culture medium are the non-specific neutralizing and binding properties of the resins that can result in the removal of nutrients and other substances required for bacterial growth. Moreover, addition of resins to the blood culture broth changes the medium's dynamics, shortens its shelf life and impacts the visual reading of growth in the bottles (90). The addition of resins may even increase the time-to-detection for some bacteria, such as Pseudomonas species (113).

However, in automated systems, resin-supplemented broths are superior to general broths and broths containing charcoal (114–118). Importantly, resins also do not interfere as much with Gram staining and reading as does charcoal (114).

For aerobic blood culture bottles, the headspace contains ambient atmosphere to which different amounts (5–10%) of carbon dioxide (CO₂) have been added. For anaerobic blood culture bottles, headspace contains CO₂ and nitrogen dioxide (NO₂). Actual amounts of CO₂, O₂ and redox potential vary widely, mostly depending on the manufacturing practices of the blood culture bottles (90).

A portion of the headspace atmosphere is evacuated to create a partial vacuum; thus, blood culture bottles contain an atmosphere in the bottle headspace that has a lower pressure than the atmosphere, enabling easy sampling (88). Not all commercially available bottles contain a vacuum; depending on the sampling techniques, this can create problems during sampling.

For detection of mycobacterial, fungal or yeast infections in the blood, other broth types and incubation times may be needed. For detection of yeast, such as Candida species, regular blood culture bottles and incubation times are advised (74), although Candida has a slightly longer time-to-detection than most bacteria (119, 120). Specialized bottles for detection of fungi in blood exist and are more effective than regular aerobic blood culture bottles (121–123), but are not routinely used.

In case of suspicion of invasive filamentous or dimorphic fungi infection, such as Histoplasma spp., the use of lysis-centrifugator tubes (Isolator system, Abbott Laboratories) is advised (74, 124). In this method, blood cells are lysed by saponin and the sample is centrifuged. The resulting sediment is then cultured directly on blood agar (69). Lysis-centrifugation can also be used for detection of bacteria, for which a study from India showed lower sensitivity but shorter time-to-detection than conventional blood culture using trypticase soy broth (125). Older studies had already shown good results for detection of bacterial or fungal BSI, but with the drawback of higher contamination rates (126, 127). Moreover, it is more time-consuming than conventional blood culture.

For mycobacteria, lysis-centrifugation has also been used to good effect, including in some LMICs (128, 129). Currently, the use of specific bottles for mycobacteria is advised (74). Longer incubation times are necessary for both mycobacterial infections and filamentous fungal infections (74).

Growth detection of manual blood culture bottles depends on visible changes in the broth such as turbidity, hemolysis, puff balls, and gas production (see Figure 4). Inspection of the blood culture bottles is done by the laboratory technicians with the naked eye. Therefore, the blood culture bottle must be designed to optimize this visual detection. For example, see-through plastic or glass must be used. The addition of resins or charcoal, described above, can create a haziness in the broth, compromising visual detection of growth (90), while addition of charcoal renders visual detection of growth nearly impossible due to the dark coloration of the broth.

Biphasic bottles can be an option to facilitate growth detection. They consist of a liquid phase (the broth) and a solid phase in the form of an agar slant (see Figure 5). These bottles were first developed by Castañeda in 1947 for the isolation of Brucella (130). Brucella species grow slowly with a need for extended incubation times. Because of hemolysis and debris of dying red blood cells, turbidity can no longer be reliably assessed beyond 7 days and repeated subcultures are needed to detect growth. By obviating the need for frequent subculture and its associated infection risks for the laboratory staff handling the blood cultures, biphasic bottles are useful for the isolation of this pathogen. Biphasic media also facilitate the detection of growth of other bacteria for non-expert users, as growth of colonies on the agar may be easier to detect than subtle changes in the broth. Additionally, there is the theoretical advantage of bypassing the initial subculture step in the process of identification and antibiotic susceptibility testing when using the colonies growing on the agar slant, resulting in a potential 18–24 h reduction of the turnaround time in comparison with the conventional blood culture bottles (131).

Published literature on the effectiveness of biphasic blood culture bottles for bacteria other than Brucella is scarce and there are no recent studies (131–136). Overall, those studies indicate superior performance of biphasic bottles for Gram-positives in comparison to other, non-biphasic, manual blood culture bottles. With regards to the speed of growth, a study by Brook et al. showed slower growth in the biphasic bottle as compared to a monophasic bottle (135). Weckbach et al. used a special biphasic bottle design (Figure 6A), physically separating the agar from the broth, and found faster recovery of yeasts (Candida spp.) and Pseudomonas spp. with the biphasic bottle compared to the conventional bottle. Because of easy subculture, isolated colonies were also available much sooner with the biphasic bottle compared to subculture from the regular bottle (131). Some other studies also found that detection of growth on the agar slant preceded detection of growth in the broth (134, 137). However, the few studies done suffer from sample sizes and a lack of comparability which precludes sound conclusions on the advantages of the biphasic blood culture bottles.

A variation to this type of biphasic blood culture bottle is the connection of a cylindrical “paddle,” containing solid agar plates, to a blood culture bottle, effectively separating the agar and broth phase (Figure 6B). Subculturing can then be done by inverting the bottle and flooding the agar plates with the broth-blood mixture. This can be done at various intervals to optimize isolation. A well-known example of this type of bottle was the Septi-Chek blood culture system by Becton-Dickinson Diagnostics (Franklin Lakes, New Jersey, United States), which was based on a conventional blood culture bottle to which a transparent plastic chamber was connected, which contained a panel of three solid agar media allowing to visualize growth and preliminary identification. This system had promising results and was widely used, but it is no longer commercially available (138–141).

The concentration of micro-organisms in blood of patients with BSI ranges between <1 colony forming unit (CFU)/ml to 10 CFU/ml of blood in adults (4). To avoid missing BSI with a low bacterial load in the process of culturing blood, it is recommended to sample as large a volume as possible for culture. Multiple studies have already shown that higher volumes of cultured blood lead to higher rates of detection (6, 142–147). Therefore, the volume of sampled blood is an important quality indicator for blood cultures and should be monitored (3, 148, 149). Blood volume inoculated in a blood culture bottle can be estimated by weighing the bottle before and after sampling, translating the weight to volume by correcting for the density of blood (1.06 g/ml) (150).

However, sampling large volumes of blood is not without risk of iatrogenic anemia in children (151). Moreover, it is not culturally acceptable to patients and healthcare workers in many LMICs (152, 153). Although this has not been widely studied, it is telling that even large clinical studies conducted in Africa using blood cultures sample only one bottle of 5–10 ml of blood in adults (154, 155).

For adults, both the Clinical Laboratories Standards Institute (CLSI) and Cumitech recommend to sample two blood cultures of 20–30 ml of blood each (3, 88) over 24 h, thus adding up to sampled volume of 40–60 ml. One blood culture is defined in these, and most other, guidelines as the volume of blood sampled through one venipuncture (70, 88). These recommendations are based on two older studies using manual blood cultures, where two consecutive blood cultures were found to detect 88 and >99% of BSI episodes (106, 156).

More recent studies suggest that even higher volumes of blood should be cultured. Cockerill et al. found that with an automated blood culture system, 65.1% of BSI episodes were detected with the first blood culture (consisting of 20 ml of blood) (6). By adding a second blood culture, 80.4% of episodes were detected, and up to 95.7% were detected when three blood cultures were used. A study by Lee in 2007 indicated similarly that two blood cultures of 20 ml each in a 24-h period detected 89.7% of BSI (147). In this study, three blood cultures of 20 ml detected 98.2% of BSI and only with four blood cultures, >99% of BSI episodes were detected. Another recent study confirmed the need for at least three blood cultures (or 60 ml) to be sampled, as they found that 7.5% of BSI episodes would have been missed without a third blood culture (157).

A particular situation in terms of number of cultures to sample is suspicion of endocarditis. In this case it is generally recommended to sample three blood cultures of 20 ml, as this has showed to detect 96 to 98% of BSI in the context of endocarditis (158).

In many studies, a distinction is made between optimal volume per blood culture and optimal number of cultures to be sampled per BSI episode. However, as it has been established that drawing blood cultures at intervals or at the time of a fever peak does not contribute to a higher recovery of pathogens (159, 160), the key variable is the total volume of blood that is cultured, irrespective of time and frequency of sampling. For example, if it is recommended to sample 60 ml of blood for culture, it may not make much difference whether this 60 ml is cultured as two blood cultures of 30 ml or as three blood cultures of 20 ml (23, 145, 161).

In children, the situation is more complex. The volume of blood that can safely and comfortably be sampled in children is related to the child's age and weight. It was long believed that very small volumes of blood were sufficient for blood culture in children, as bacterial concentrations in children with BSI were thought to be much higher than in adults (162). However, some studies have shown proportions of low-level bacteremia (< 10 CFU/ml) in 23 to 69% of infants and children (163–165). In line with these, sampling higher volumes has been shown to increase sensitivity of blood cultures in children (163, 166, 167). Given the high proportion of low-level bacteremia that was found by Kellogg et al. (163), the authors recommended to sample up to 4–4.5% of the patient's total blood volume.

The latest guidelines from the Infectious Disease Society of America and American Society of Microbiology, recommend sampling 2.5 to 4% of total blood volume from children (74) (Table 4). However, a review from 2011 suggests that only 3.8% of total blood volume can be safely sampled (for all laboratory analyses combined over a 24 h period) in children beyond the neonatal period, and this safety value is further challenged by higher rates of severe anemia in children in LMICs, particularly caused by severe Plasmodium falciparum malaria, which predisposes to Gram-negative BSI (43, 151, 168, 169). Given the necessity in many seriously ill children to sample blood for other analyses as well, sampling 4% of total blood volume for blood culture alone can therefore not be recommended in LMIC. More in line with these concerns, CLSI recommends sampling of maximum 1% of the total blood volume (for infants and younger children) (88).

Many articles however recommend an age-based sampling strategy illustrated in Table 5 (166, 170, 171). This age-based sampling is convenient and safe, as the volumes are noticeably smaller than with weight-based guidelines (see Figure 7 for an example in LMICs). However, no upper limits are defined for any of the age categories, leaving quite some room for interpretation. For children over 36 months of age, this guideline risks sampling insufficient volume. It is unlikely that the same volume is appropriate for a 4-year old child as for a 13-year old child, but this distinction is not made. Unfortunately, as CLSI remarks, “There are no published data for determining when volumes considered to be appropriate for adults can be used for older children” (88). As was pointed out by Dien Bard et al. in 2016, there is no consensus between different guidelines on which volume to sample in children (172), so controversy remains.

In conclusion, it is recommended that for adults at least 40 ml of blood should be drawn, divided over 4 blood culture bottles, to obtain sufficient sensitivity. If feasible, it is recommended to sample an additional 20 to 40 ml of blood. It is unnecessary to sample more than 80 ml of blood, as >99% of pathogens will be detected at this point (147). In case of suspicion of endocarditis, 3 blood cultures of 20 ml each must be sampled. For children, recommendations are less straightforward. Because of simplicity, concerns regarding iatrogenic anemia, widespread use in hospitals and comparability across published literature, we recommend the use of the age-based simple algorithm depicted in Table 5 (e.g., sampling ≥0.5 ml for neonates <1 month of age, ≥1 ml for children age 2–36 months and to sample ≥4 ml in children ≥36 months of age), despite our awareness of the limitations of this strategy.

It can be difficult to make the distinction between a contaminant and a pathogen, as some typical blood culture contaminants such as coagulase-negative Staphylococcus species can cause catheter infections or other foreign body infections. The distinction can be made either by clinical assessment (i.e., review of medical records), or by the number of blood cultures that show growth for this particular organism. Often, such an organism is only regarded as clinically relevant if it is isolated in at least 2 separate cultures (and venipunctures), because the odds of having contaminated both cultures with the same pathogen are very small (174). However, this approach cannot be used in settings where only one blood culture is sampled. Time-to-detection can also be of help in the interpretation, as it has been shown that contaminants show slower growth than true pathogens (6, 119, 175).

Corynebacterium species, Bacillus species (other than Bacillus anthracis), Micrococcus species, Lactobacillus species and Cutibacterium species are rarely associated with clinical infections and are almost always contaminants (176). Coagulase-negative staphylococci can cause true infection but are much more often implicated as contaminants. Isolation of Enterococcus species, non-fermenting Gram-negative species (e.g., Acinetobacter species or Stenotrophomonas maltophilia) and viridans streptococci is often of uncertain clinical significance, complicating their role in the interpretation of blood culture results (70, 176).

Contamination of blood cultures remains a substantial problem, with contamination rates as high as 10% even in many HICs (177). Contamination is suspected to be more frequent in LMICs, with some settings (i.e., South Africa, Ghana, the Gambia, Malawi) indeed reporting very high contamination rates (178–181). The spectrum of contaminants in LMICs differs slightly to HICs; for example, more Bacillus species are seen in LMICs (178). This finding suggests that contamination of blood cultures in LMICs commonly has the environment as a source, as Bacillus species are known to be present in dust and have been described in outbreaks of pseudo-bacteremia originating from the environment (182, 183). More confusingly, some organisms representing “true pathogens” in HICs, such as Pseudomonas aeruginosa or other non-fermenting Gram-negative organisms, are sometimes considered as contaminants in LMICs (181, 184).

Laboratory work-up of these contaminants demands time and money, both of which are not in large supply in resource-limited settings. Furthermore, contaminants lead to longer hospital stays, increased prescription of antibiotics and related morbidity (185–188), and may hamper growth of pathogens. Therefore, avoiding contamination is of utmost importance and can lead to substantial cost savings and improvements in patient management. Current guidelines advocate a target of <3% contamination rate and ideally <1% (70, 88, 173). In the following paragraphs, some strategies are described to avoid contamination.

Because of the challenges related to visual detection of growth, other strategies were employed before the widespread use of automated monitoring of blood culture bottles. For manual systems, “blind subcultures” are often recommended. These are subcultures from the blood culture broth on agar plates, in the absence of visual signs of growth. The optimal timing of this subculture varies according to the consulted source. Depending on the moment of blind subculture during the incubation [for instance early at day 1 vs. at day 7 at the end of the incubation period (so-called “terminal” subculture)], blind subculture can be used either as a way to shorten the time-to-detection or as a final check for growth.

CLSI for example recommends subculture after 24 to 48 h of incubation to facilitate early detection of microorganisms (88). The American Society of Microbiology “Clinical Microbiology Procedures Handbook” by Leber et al. (70) recommends blind subculture at 72 h. Cumitech advises routine blind subcultures after 12–18 h of incubation for aerobic bottles (3, 70). Routine blind subcultures are not advised for anaerobic bottles (3).

Early subculture (within the first 24 h of incubation) seems to be a successful strategy for rapid detection indeed. Szymczak et al. detected 85% of pathogens with blind subculture within 24 h, whereas in only 48% of positive cultures visual signs of growth could be detected within 24 h of incubation (238). Even earlier subculture, within four to 14 h after incubation, led to detection of 50% (239) and up to 85% of pathogens (240). Studies performing subculture between 6 and 17 h after incubation detected 48% (241) and 63% (242) of pathogens. Subcultures performed within 6 h yielded only 10% of pathogens according to the study by Sliva et al. they therefore do not recommend subculture within 6 h of incubation (242). A drawback of blind subcultures is that by opening the bottle and introducing a syringe or pipette, contamination may be introduced.

Terminal subcultures (blind subcultures of blood culture bottles at the end of the incubation period) are considered unnecessary by many and may increase contamination rates and needle stick injuries (243–245). However, they may still be considered in systems which do not perform a routine blind subculture at an earlier time or in situations with prolonged incubation (245). They are also useful at the validation phase of a newly introduced blood culture system. In settings with many immunocompromised patients, where higher numbers of Pseudomonas aeruginosa and yeast BSI are anticipated, terminal subculture could also be of value, as these organisms tend to show slower growth and more subtle visual signs of growth, presumably due to smaller initial concentrations of bacteria (246).

It may be impossible to increase the speed of detection of manual systems up to the performance of the automated systems, as the algorithm-driven continuous monitoring will prove difficult to mimic in an equipment-free manual system. However, another feature of automated systems is continuous agitation during incubation. Agitation is thought to improve microbial recovery by increasing the oxygen concentration in the broth medium and will therefore speed up detection of aerobic bacteria (69, 256–258). The “Clinical Microbiology Procedures Handbook” recommends continuous agitation of manual systems as well (70). Continuous agitation, however, impacts ease of visual growth detection, as turbidity will increase due to mixture with the red blood cells.

Another strategy to speed up growth is venting, i.e., inserting a filtered needle or other device that allows air to enter into the bottle headspace. The need for venting of aerobic blood culture bottle depends mainly on the manufacturer of the blood culture bottle. For automated systems, it is agreed that venting is not necessary (69, 259), but some commercial manual blood culture bottles may still need venting. Most studies evaluating the impact of venting stress the role of appropriate amounts of oxygen; this seems especially important for Pseudomonas, Candida and other strictly aerobic species, both for rate and speed of growth (77, 144, 241, 260).

It is currently unclear whether venting is still required with the newer commercial manual bottles; there are no recent studies on this. Of course, venting involves an extra procedure to be carried out, thereby potentially contaminating the sample. Despite this theoretical possibility, there is no documented evidence for increased contamination with vented bottles.

Clinical importance of increased speed of detection is also influenced by laboratory organization. Notification of the Gram stain result to the clinician has been shown to have the highest clinical impact (56, 58, 59). Laboratories not providing 24-h service will probably see less benefit from small differences in time-to-detection, as they may be unable to translate this into quicker results to the clinician. Furthermore, the most important actionable result is growth in the first bottle of a BSI episode; growth in the other bottles of that episode are of lower clinical impact. When taking all this in consideration, the clinical relevance of difference in speed between automated and manual systems may be lower than expected. Future studies comparing speed of growth of different systems, including comparisons between automated and manual systems, should take these considerations into account.

A study that has looked at the practical organization in the lab was performed by Youngs et al. in 1985 (261). They recognized the added value of early subculture (after 6–17 h of incubation) but mentioned that the actual timing of subculture also depended on arrival of the bottle in the lab. To be sure that a subculture was performed within 6–17 h, they implemented two timeslots for subculturing: one in the early morning for all bottles having arrived after office hours the previous day and another subculture at the end of the work day for all bottles that arrived during the day. This led to growth of pathogens on solid media the day after incubation for 10 out of 14 grown blood cultures, making an appreciable impact for patient management, as none of these bottles had shown visual signs of growth the day before (261).

A relatively simple way of improving the speed of detection is by ensuring that the bottles arrive in the laboratory in a timely manner. It has been shown that a delayed incubation significantly prolongs detection time and may lead to false-negative results in automated systems (23, 262, 263), however the exact critical time has not been formally established. Guidelines recommend maximum needle-to-incubator time of 2 h (88) to 4 h (70); however, many centers actually observed much longer average transport times (264). Long distances and irregular transport of laboratory specimens from the site of collection to the laboratory is more common in LMICs, resulting in even higher transport times.

Blood culture bottles must be transported to the lab at room temperature (88, 265) or in a thermostable carrier and should never be refrigerated or frozen, as many fastidious organisms are vulnerable to cold (88, 266). Pre-incubation of blood culture bottles at 35°C, for example in a small incubator in the emergency department, can speed up detection of growth in manual blood culture systems (267), but is not advised for automated systems as it may result in false-negative results (23, 268–270). However, high “room temperatures” are common in LMICs, warranting short transport times when using automated blood culture systems to avoid false-negative results.

Performing direct identification and/or antibiotic susceptibility testing on grown blood culture broth can also lead to faster diagnosis and decision-making. Molecular assays and matrix-assisted laser desorption/ionization—time-of-flight (MALDI-TOF) methods are currently out of reach for most LMICs, but lateral-flow immunoassays and other simple testing methods can provide a convenient solution. A number of simple direct tests have been described. Examples are detection of Staphylococcus aureus by direct tube coagulase test for detection of Staphylococcus aureus (271, 272), latex agglutination tests (273) and immunochromatographic tests (274), and the recent development of lateral flow assays for Burkholderia pseudomallei (275) and Salmonella species (276).

For antibiotic susceptibility testing, lateral flow assays have been developed to rapidly detect methicillin-resistant Staphylococcus aureus (277), extended-spectrum beta-lactamase (ESBL) (278) and carbapenemases (279), directly from blood culture broth. Moreover, the European Committee on Antibiotic Susceptibility Testing (EUCAST) recently provided guidelines on direct disk diffusion testing from blood culture broth, with adapted procedure and diameter cut-offs⁴. Performing antibiotic susceptibility testing directly from blood culture broth decreases the turnaround time with 1 day but requires rapid identification as the procedure often depends on the pathogen.

Access to quality-assured in-vitro diagnostics is a problem in LMICs; diagnostics are frequently more expensive when purchased in a LMIC and they are not always easily available (104, 280, 281). Although data are rare, it is estimated that the cost of blood cultures is twice as high in LMICs than HICs (41). Most healthcare systems in LMICs rely heavily on out-of-pocket payments of patients, including for diagnostics (153, 282, 283). According to a recent WHO report, out-of-pocket payments, although in decline, still account for 37% of the total healthcare cost in Africa and 44% in South-East Asia (282). Assuming a positivity rate of only 10% for blood cultures, the cost-benefit ratio for the individual patient of performing blood culture may be unacceptable in very poor populations. A Kenyan study of 2009 reports a blood culture cost of 18.36 US $ per individual test, much higher than other diagnostic tests in the same setting such as HIV test (3.68 US $) or malaria microscopy (3.50 US $) (284). Moreover, this price refers to a single blood culture bottle; as mentioned before, often two to four blood culture bottles are needed per patient, further increasing the total price. According to a panel of stakeholders, an acceptable cost for blood cultures would have to be <10 US $, and ideally <5 US $ (153). In the short term, donor involvement will therefore be necessary for blood cultures to be affordable.

Transport of commercial blood culture bottles from abroad is very costly and transport times are considerable, compromising shelf life. Low volumes of cultures per lab, skilled labor needs and short shelf lives further affect affordability of blood cultures in many settings (41). For this reason, many laboratories resort to in-house prepared blood culture bottles, typically containing brain-heart infusion or trypticase soy broth (41). This practice is not advised, because the origin and quality of the culture media are difficult to track down and useful additives, such as the ones described in Table 3, will not be present in the broth, compromising performance. An alternative to the commercially available bottles is to support local production of blood culture media by reference laboratories, from where they can be dispersed to smaller hospitals (2, 285). An exemplary initiative is the laboratory logistic and educational support provided by the Diagnostic Microbiology Development Program in Cambodia (1)⁵.

When choosing a blood culture medium to procure, it is good to remind that there is no perfect all-purpose blood culture medium. The ideal medium will depend on the expected pathogen distribution, which may vary according to the setting. Furthermore, the choice of broths will depend on whether anaerobic blood culture is deemed necessary. Most centers that analyzed the frequency of isolation of anaerobes from blood cultures have noted a declining incidence; coinciding with this, other (aerobic) organisms such as fungi, Pseudomonas species and fastidious aerobic bacteria are increasingly associated with disease (69). Some have suggested to replace routine inoculation of anaerobic media by a second aerobic bottle and reserve the anaerobic bottle only when clinically indicated or in patients at risk of anaerobic infections (286–289).

Not all hospitals, however, have seen this decrease in anaerobic bacteremia (290, 291). Moreover, Vena et al. found that in half of the cases of anaerobic infection, this could not have been suspected on clinical grounds (291). The choice of which bottles to use is therefore highly dependent on the specific setting, and every center is encouraged to perform its own evaluation on the need for anaerobic blood culture (69). These differences in regional incidence of anaerobic isolates from blood culture were also confirmed in a more recent review (292). In this review, it is stated that anaerobic cultures can safely be omitted in most cases, unless clinical indications for anaerobic infections are present. Moreover, identification and antibiotic susceptibility testing of anaerobic bacteria is often difficult in LMICs due to the need of specialized and generally more expensive techniques⁶ (293). CLSI however still recommends that routine blood cultures include paired aerobic/anaerobic blood culture bottles, as the data are conflicting and the recommendation of limiting anaerobic blood culture bottles has never been validated by controlled clinical studies (88, 290, 294, 295). Because anaerobic bacteria are rarely recovered in pediatric patients, and usually only one bottle is sampled, the use of only aerobic bottles has been recommended in this patient population (296, 297).

Equally important is the selection of the correct bottle type (Figure 8); for reasons of (bio)safety and shipping costs, plastic bottles may be preferred over glass bottles. The specific type of plastic may be important for subsequent waste management (see below). The bottle material also needs to be completely transparent to allow visual inspection for growth. Bottles have to be stored and shipped adequately to avoid scratches on the material. To increase ease of use, the septum of the bottle must be easy to perforate by a needle and the vacuum in the bottle must be sufficient to aspirate the required amount of blood (10 ml for adult bottles, 4 ml for pediatric bottles). A screw cap may be more practical than other cap types in terms of waste management; after autoclavation, the fluid can then be poured out safely and the empty bottle incinerated. A screw cap is also advantageous for sampling the blood-broth mixture (e.g., for subculture), as this can avoid the use of needles.

Lastly, blood culture bottles should be “tropicalized”, i.e., adapted to harsh environmental conditions (1, 153), in order to be useful for LMICs. In many LMICs, no manufacturers of in-vitro diagnostics are locally present, necessitating import and shipment from other countries. Therefore, shipment stability is an important factor, as is a sufficiently long shelf life. Many compounds in blood culture bottles, such as resins and SPS, may shorten shelf life, as exemplified by the much shorter shelf life of bottles designed for automates when compared to manual blood culture bottles. In line with requirements for malaria rapid diagnostic tests according to the Prequalification of in vitro Diagnostics (PQDx) Programme⁷ of the WHO, shelf life must be at least between 1.5 and 2 years. Storage temperature must be sufficiently flexible to allow storage or at least shipment at tropical temperatures (153).

A laboratory performing blood cultures usually functions as a Biosafety Level (BSL)-2 laboratory (305). When handling blood cultures, it is possible that laboratory staff comes into contact with unknown pathogens causing airborne infection. This raises the question whether laboratories in LMICs must also be equipped with a biosafety cabinet, keeping in mind the challenges in LMICs of dusty environments, lack of electricity and equipment maintenance (Table 7). A class II biosafety cabinet provides protection to laboratory staff against infectious materials as well as to biological samples against external contamination. The presence of a biosafety cabinet is generally recommended for a BSL-2 laboratory in case of likely aerosol formation or splashes (305, 306, 309). A biosafety cabinet most obviously provides protection from inhalation of infectious aerosols. For most other types of exposure, on the other hand, good personal protective equipment, such as safety goggles, face shield, gloves and masks, and the use of safe and appropriate techniques (no mouth pipetting, use of disposable loops) are sufficient. Therefore, the decision of the installation of a biosafety cabinet in a specific setting should be taken after a risk analysis, taking into account possible laboratory-acquired infections when processing blood cultures, and existing alternative methods to mitigate this risk (Table 7).

Because of gas production and possible aerosol formation, opening the blood culture bottle to perform subculture substitutes a risk to the laboratory technician. Apart from measures implemented to protect the laboratory technician from this aerosol, aspects of the bottle itself can play a role in safety. For example, biphasic bottles provide opportunities of subculturing without having to open the bottle. Bottle septa that are difficult to perforate may be a risk for needle stick injuries.

Another important biosafety aspect is the possibility to safely destroy (decontaminate) grown blood culture bottles, for instance by autoclaving and incineration. The bacterial load of grown blood culture bottles is very high, easily reaching 10⁶-10⁹ CFU/ml (310). In case of steam autoclavation of blood culture bottles, it must be kept in mind that the fluid inside the blood culture bottle must reach 121°C during at least 15 min to ensure effective decontamination. This requires longer cycles when a higher volume of liquids (e.g., more blood culture bottles) are autoclaved, a phenomenon that is often not taken into account. Monitoring the autoclave cycle by using chemical indicators is often not sufficient, as they will only show the presence of steam but do not give an indication of the temperature inside the liquids (311). The autoclave must be validated before taken into use with the help of temperature sensors (if available) and biological indicators, preferably placed inside (non-inoculated) blood culture bottles. Given the importance of the total volume of liquids in the autoclave on the time needed to sterilize, it is important not to overfill the autoclave with liquid waste and to validate the maximum liquid load the autoclave can still safely sterilize.

Moreover, not all plastic can be safely autoclaved after use. Even if re-use of the bottles is not necessary, it is important that the bottle material can be autoclaved without posing direct biological risks by tearing and implosion of the bottles or loosening of the stopper. Polycarbonate (PC) and polypropylene (PP) bottles withstand autoclavation, however polyethylene terephthalate (PET) bottles cannot be autoclaved without substantial damage to the bottle, potentially causing spill of infectious material inside the autoclave. In case of interrupted or only partially successful cycles, this may bring the healthcare worker into contact with infectious aerosols. Therefore, other types of plastic that withstand autoclavation are preferred (like PP or PC)⁸.