High rates of multi-drug resistant gram-negative organisms associated with surgical site infections in a teaching hospital in Ghana

Background There is limited data to guide the prevention and management of surgical site infections (SSI) in low- and middle-income countries. We prospectively studied aetiological agents associated with SSI and their corresponding antibiotic susceptibility patterns in a tertiary hospital in Ghana. Methods As part of a cohort study carried out at the surgical department of the Korle Bu Teaching Hospital (KBTH) from July 2017 to April 2019, wound swabs were collected from patients diagnosed with SSI. Isolates cultured from the wound swabs were identified by MALDI TOF and susceptibility testing was conducted according to EUCAST 2020 guidelines. Clinical data were monitored prospectively. Results Of 4577 patients, 438 developed an SSI and 352 microbial isolates were cultured. Isolates were predominantly Gram negative (286, 81%), a pattern seen for all kinds of surgery and all wound classes. The most common species included Escherichia coli, Pseudomonas aeruginosa, Klebsiella pneumoniae, Staphylococcus aureus and Acinetobacter baumannii. The majority of organisms were multi-drug resistant including 86% of E. coli, 52% of A. baumannii and 86% of K. pneumoniae; and 65% (17/26) of the cefotaxime-resistant K. pneumoniae were extended spectrum β-lactamase producing. One of 139 E. coli, 15 of 49 P. aeruginosa, and 6 of 23 A. baumannii were meropenem resistant, but no clonal pattern was found. There was a 1% (5/428) prevalence of methicillin-resistant S. aureus. Conclusions The predominance of Gram-negative organisms and the high level of multi-drug resistance indicate a need to re-evaluate antibiotic prophylaxis and treatment protocols in surgical practice in low- and middle-income countries.


Background
Surgical site infections (SSI) is a major type of healthcare associated infections (HCAI), forming as much as 33% of all HCAI in sub-Saharan Africa [1,2]. A major teaching hospital in Ghana also described an SSI incidence risk of 10% with an associated 9 extra days of stay and $1519 excess cost to the institution per patient [3,4].
The global spread of antibiotic resistance further compounds the problem of SSI. Infections caused by resistant bacteria lead to poor treatment outcomes. Major drivers of antimicrobial resistance in low-and middleincome countries include inappropriate prescription practices and poor infection control measures [5]. Thus, the prevalence of antibiotic use among surgical patients in sub-Saharan Africa is 24-73% [6], and the majority of these antibiotics are prescribed for prophylaxis and usually beyond the recommended 24-h period [7]. Reasons for this prolonged antibiotic use include a fear of infections due to poor infection control [8].
In Ghana, as in most low-and middle-income countries, antimicrobial treatment of wound infections is mainly empirical due to limited laboratory services in most health facilities and the costs to patients of culture and susceptibility testing, when available [9]. The selection of prophylactic antibiotic therapy is based on the operation site and the normal resident flora. Thus, empirical choice of antibiotics relies on knowledge of the susceptibility patterns of common local pathogens. Absence of this data precludes rational use of antibiotics for treatment and prevention of SSI.
In this study, we prospectively identified the aetiological agents of SSI at the surgical department of a tertiary hospital in Ghana, with emphasis on their antibiotic susceptibility patterns, and related to the patient characteristics.

Study site, patient recruitment and collection of samples
Samples were collected as part of a cohort study at the surgical department of the Korle Bu Teaching Hospital from July 2017 to April 2019. The study design and results (for the period July 2017-December 2018) of the surveillance have been reported elsewhere [3]. All patients who underwent surgery in the unit were followed actively for the occurrence of SSI during admission and post-discharge for 30 days, and infection was defined according to Centers of Disease Control and Prevention (CDC) criteria [3,10]. Wound contamination and SSI type were classified based on the CDC classification [10]. Implant surgery and surgery with wounds that could not be closed immediately were excluded.
Patients on admission had their change of dressing carried out on their bed in the wards. Post-discharge change of dressing was carried out in dressing rooms situated on each ward. Wound swabs were taken from consecutive patients diagnosed with SSI, either on admission or post discharge for microbiology analysis. The project staff had been trained to carefully collect exudate from the infected surgical site using sterile cottontipped applicators (Sterilin, U.K), which were transferred to the microbiology laboratory within an hour of sampling. Preliminary culture and susceptibility results were reported immediately to the physicians for management of patients. Isolates were then frozen at − 80°C and transported to Copenhagen for confirmatory testing and additional analyses.
Enterobacterales resistant to third generation cephalosporins were screened for production of extended spectrum beta lactamases (ESBL) using the double disc diffusion method [13]. Isolates with phenotypic resistance to meropenem were screened depending on the species. Pseudomonas spp. were screened for carbapenemases using GeneXpert (Xpert® Carba-R, Cepheid, France), Acinetobacter spp. were screened for OXA-23 using a rapid diagnostic test (OXA-23 K-SeT, Coris, Belgium), and Acinetobacter spp. and enterobacterales were screened for carbapenemases using NG-CARBA 5® (Hardy Diagnostics, CA, USA). All Staphylococcus aureus isolates were screened for methicillin resistance using cefoxitin. Multidrug resistance (MDR) was defined as resistance to ≥1 antibiotic in ≥3 antibiotic groups [14].

Data analysis
For each bacterial agent, the percentage frequency of resistant isolates was determined. The resistance patterns of isolates of the same species were screened visually for possible clonal distribution. Distribution of bacterial types by clinical groups were compared by chi-square test and p < 0.05 was considered significant. We used Stata /MP version 15.1 (Stata Corp., College Station, Tx, USA) for the analysis.

Demographic and patient characteristics
A total of 4577 patients were included in the study. Of these, 438 (9.6%) patients, developed an SSI. The SSI risk was 9.3% (376/4054) in the general surgical department, 26.5% (31/117) in the department of urology and 7.6% (31/406) in paediatric surgery. The median age of the patients who developed an SSI was 45 years (interquartile range 31-60 years) and 239 (54%) patients were female.
From 382 (87%) of the 438 patients, a wound swab was taken for microbiological analysis. No wound swabs were taken in 13% (56/438) of the cases due to; surveillance team missing the opportunity to take a swab at change of wound dressing or at a relaparotomy for an SSI, missing the diagnosis of an SSI until the attention was drawn to the clinical signs and, wound swabs taken but accidentally sent to a private laboratory for analysis.
There was a difference between type of wound and the ratio between Gram-negative and Gram-positive organisms (Table 3), and S. aureus was only cultured from superficial SSI (Table 4). Conversely, all isolates from organ-space SSI were Gram negative. A much higher proportion of organ/space infections than other types of infections were not cultured, mostly due to the surveillance team missing the opportunity to take a wound swab during a relaparotomy to drain abscesses.

Antimicrobial susceptibility patterns
The majority of bacterial isolates were MDR, ranging from 23 to 86% for Proteus spp. and E. coli isolates, respectively (Table 1). Acquired resistance to commonly used antibiotics ranged from 1% to meropenem in E. coli to 95% to ampicillin, ciprofloxacin and trimethoprimsulphamethoxazole in E. coli (Fig. 1a).
Fifteen percent (5/33) of S. aureus isolates were methicillin resistant (MRSA). Overall, the MRSA prevalence was 1% (5/438) in the cultured wounds. Resistance of S. aureus against other antibiotics ranged from 6% to gentamycin and clindamycin to 42% to tetracycline. Among coagulase-negative staphylococci, the percentage resistance ranged from 23% to clindamycin to 81% to tetracycline. Neither S. aureus nor coagulase negative staphylococci showed resistance to linezolid (Fig. 1c).
The sensitivity pattern of the MDR organisms, including those for which specific resistance mechanisms such as ESBL or carbapenemases were detected, did not give any indication of an outbreak of one or more clones (data not shown).

Discussion
In this study, from a surgical facility in Africa, Gramnegative rods were the primary aetiology of SSI. Despite differences in ratios between Gram-negative and Grampositive organisms by types of surgery and SSI types, Gram-negatives dominated in all categories. There was a high level of MDR among isolated organisms, including carbapenem-resistant P. aeruginosa and A. baumannii and ESBL-producing E. coli and K. pneumoniae. There was no phenotypic indication of clonality of the MDR organisms, indicating that the findings could not be explained by an outbreak. These findings challenge the standard recommendations for empiric prevention and treatment of SSI in low-and middle-income countries.
Unlike earlier studies in Ghana, which recorded low rates of microbiology testing [9], this study recorded a high rate of testing. This can be explained by the fact that under the active study conditions, wound swab samples were taken by the surveillance team when an SSI was diagnosed, irrespective of the clinicians' diagnosis, and the cost of the testing was also borne by the study. In low-and middle-income countries, low rates of microbiology testing are often reported, presumably due to the limited facilities for microbiology testing, the costs of testing, limited numbers of trained personnel, and the tendency for clinicians to underutilize existing microbiology facilities [9,[15][16][17][18] As reported previously, the active surveillance and access to microbiological testing in this study lead to immediate improvements in SSI rates in the course of the study [3].
E. coli was the commonest isolated organism, whereas, S. aureus and coagulase negative staphylococci constituted less than a fifth of the isolates. Some studies have also predominantly isolated E. coli from infected surgical sites post abdominal surgeries [19,20] but in most studies in Africa, S. aureus was the most common pathogen isolated from SSI [21]. The large numbers of gastrointestinal surgeries performed in this study cannot fully explain the predominance of E. coli since Gram-negative rods were the predominant aetiology of SSI in all forms of surgery. A possible explanation of this could be a high rate of skin carriage of Gram-negative organisms as    shown in a study from Tanzania [22]. In addition, the routine use of antibiotic prophylaxis with an effect on S. aureus in our department coupled with a relatively low rate of methicillin resistance may also have skewed the distribution of microorganisms toward Gram negatives. Finally we recently found high levels of antimicrobial air contamination in our surgical facility and demonstrated a causal relationship with SSI [23]. This study indicated that air contamination may have contributed to infections with environmental bacteria such as P. aeruginosa and A. baumannii, whereas enterobacterales were uncommon in the air samples.
Skin carriage of Gram-negative organisms may be associated with previous antibiotic use. In line with this, it is likely our patients had received antibiotic therapy at referring facilities before coming to our tertiary hospital, where the treatment may have been continued. We have recently documented long periods of administration of antibiotics in surgical units in Ghana, at all levels of health facilities [9]. Long periods of administration of antibiotics have also been reported in other health facilities in low-and middle-income countries [24,25].
The high usage of antibiotics coupled with the low rate of microbiology testing to inform choice of antibiotic    therapy, may explain the high levels of MDR. Hospitalbased antimicrobial stewardship programs are said to decrease antibiotic use, though data on this is limited in low-and middle-income countries [26]. There is a need to develop and document the effect of functioning antibiotic stewardship programs based on longitudinal monitoring of microbiological test results from the surgical department.
The choice of antibiotics in the department is essentially based on the Ghana standard treatment guidelines, usually reviewed at intervals of 5 years or more [27,28].
Antibiotics like ciprofloxacin are used routinely by surgeons as therapeutic and prophylactic treatment for gastrointestinal surgery, and in combination with clindamycin for limb amputation for dry and wet lower limb gangrene. The high prevalence of ciprofloxacin resistance in this study, ranging between 95 and 33% among E. coli and S. aureus respectively, shows the evolving pattern of resistance. In comparison, a 10% resistance to ciprofloxacin in fecal E. coli has been reported in the past [29]. Conversely, the low resistance to vancomycin, piperacillin/tazobactam and linezolid may reflect their low usage or unavailability in most Ghanaian facilities, based on the essential medicines list of the ministry of health [27].
The low prevalence of MRSA confirms previous findings from inpatients in our institution [30]. A high use of amoxicillin/clavulanic acid may explain the low frequency of S. aureus in our study, but the high prevalence of amoxicillin/clavulanic acid-resistant Gram-negative organisms suggests the need to re-evaluate the protocols.
The resistance to 3rd generation cephalosporins by K. pneumoniae and E. coli in our study was mainly caused by ESBL. This cannot be explained by a high usage of third generation cephalosporins as the Ghana treatment guidelines do not recommend this drug class. We have recently reported clonal outbreaks of MDR K. pneumoniae in neonatal intensive care units in Ghana [31], but in the present study we did not find indication of ongoing outbreaks, based on phenotypic analysis. The high ESBL rate may thus mimic antibiotic resistance in patients' own flora indicating widespread carriage of resistant organisms in the society [32].
This study involves data from only one hospital, limiting the generalization of the results, though the hospital serves as a major referral center for a population of over 30 million.