Prevalence and drug resistance patterns of Gram-negative enteric bacterial pathogens from diarrheic patients in Ethiopia: A systematic review and meta-analysis

Background Diarrhoea is the leading cause of morbidity and mortality in the world particularly in developing countries and among vulnerable groups of the population. Gram-negative enteric bacterial pathogens (GNEBPs) are a group of organisms that reside mainly in the intestine and induce diarrhoea. Antimicrobial agents are usually the part of their treatment regimen. The therapeutic effect of antimicrobials is hindered by the emergence and spread of drug-resistant strains. The information regarding the prevalence and antimicrobial resistance patterns of GNEBPs in Ethiopia is limited and found in a scattered form. Objectives This study was designed to determine the pooled prevalence and drug resistance patterns of GNEBPs by meta-analysis of data from diarrhoeic patients in Ethiopia. Method A comprehensive literature search was conducted through internet searches using Google Scholar, PubMed, Science Direct, HINARI databases, and reference lists of previous studies. Published articles were included in the study based on priorly set inclusion and exclusion criteria. Results were presented in the forest plot, tables, and figures with a 95% confidence interval (CI). The inconsistency index (I2) test statistics was used to assess heterogeneity across studies. The pooled prevalence estimate of GNEBPs and their drug resistance patterns were computed by a random-effects model. Software for Statistics and Data Science (STATA) version 14 statistical software was used for the analysis. Result After removing those articles which did not fulfil the inclusion criteria, 43 studies were included in the analysis. Studies were conducted in 8 regions of the country and most of the published articles were from the Amhara region (30.23%) followed by Oromia (18.60%) and Southern Nations, Nationalities, and Peoples’ region (SNNP) (18.60%). The pooled prevalence of GNEBPs was 15.81% (CI = 13.33–18.29). The funnel plot indicated the presence of publication bias. The pooled prevalence of GNEBPs in Addis Ababa, Amhara, SNNP, and Oromia regions were 20.08, 16.67, 12.12, and 11.61%, respectively. The pooled prevalence was 14.91, 18.03, and 13.46% among studies conducted from 2006–2010, 2011–2015, and 2016–2021, respectively and it was the highest (20.35%) in children having age less than or equal to 15 years. The pooled prevalence of Escherichia coli, Campylobacter spp., Shigella spp., and Salmonella enterica were 19.79, 10.76, 6.24, and 5.06%, respectively. Large proportions (60–90%) of the isolates were resistant to ampicillin, amoxicillin, tetracycline, and trimethoprim-sulphamethoxazole. The pooled prevalence of multidrug resistance (MDR) was 70.56% (CI = 64.56–76.77%) and MDR in Campylobacter spp., Shigella spp., E. coli, and S. enterica. were 80.78, 79.08, 78.20, and 59.46%, respectively. Conclusion The pooled estimate showed a high burden of GNEBPs infections and a high proportion of drug resistance characters to commonly used antimicrobial agents in Ethiopia. Therefore, performing drug susceptibility tests, establishing an antimicrobial surveillance system and confirmation by molecular techniques are needed.


Introduction
World Health Organization (WHO) defines diarrhoea as the passage of three or more loose or liquid stools per day (24 hours). During diarrhoea, the water content and volume of stool and defecation frequency will usually increase. The syndrome may be accompanied by other illnesses like vomiting, fever, dysentery, nausea, and abdominal cramps. It is the leading cause of morbidity and mortality in the world and contributes about 4% of all deaths and 5% of health loss to disability [1][2][3]. Diarrhoea is the fifth leading cause of death and it contributes to one in nine deaths among children younger than 5 years [4,5]. The problem is severe among the vulnerable population such as children, people with HIV, the elderly, and other individuals having weak immunity. Many factors contribute to diarrhoea; however, childhood wasting (low weight-for-height score), unsafe water, and unsafe sanitation are the leading risk factors [5]. The incidence of diarrhoea is different among the regions or continents of the world. It is highly prevalent in Sub-Saharan Africa and South Asia. The report from WHO showed that these countries account for about 78% of all diarrheal deaths among children in the developing world [2,4]. Ethiopia is one of the top three countries with very high child mortality due to diarrhoea in Africa [5][6][7]. Diarrhoea can be induced by a variety of causes. However, infectious agents like viruses and bacteria are among the leading causes. Bacteria, particularly Gram-negative enteric bacterial pathogens (GNEBPs) are the common causes of the syndrome. The group includes bacteria that reside mainly in the intestine. Genera such as Escherichia, Shigella, Campylobacter, Salmonella, Enterobacter, Klebsiella, Yersinia, Serratia, Proteus, and others are included in the group. design, study region, study population, sample size, sample type, age, gender, isolated bacteria species, their prevalence, resistance patterns of the isolates, and prevalence of multidrug resistance. The extracted data were checked at least twice for their accuracy.

Quality control
The quality of eligible studies was checked using a set of criteria based on Joanna Briggs Institute critical appraisal tools including appropriateness of the research design to address the target population, adequate sample size, quality of paper, completeness of the information, and appropriateness of methods for isolation of the bacteria and appropriate statistical analysis [16]. The eligibility of selected articles was also assessed and approved by experts in the discipline.

Data analysis
The data were compiled in Excel 2010 (Microsoft, Redmond, WA, USA) spreadsheet and summarized by descriptive statistics. A random-effect model was used to determine the pooled prevalence and the 95% confidence interval (CI). All statistical analysis was achieved by using Software for Statistics and Data Science (STATA; https://www.stata.com/company/our-sites/) version 14. The data were described using forest plots, figures, and tables. The presence of publication bias was assessed by funnel plot. Sub-group analysis was performed based on the regions in the country, age group; (children, adults, and all age groups), and year of study (2006-2010, 2011-2015, and 2016-2021). Statistical heterogeneity was evaluated by the inconsistency index (I 2 ) test. The I 2 provides an estimate of the percentage of the variability in effect estimates that is due to heterogeneity rather than sampling error or chance differences. Hence, the I 2 test measures the level of statistical heterogeneity among studies [17,18]. Of 7,349 identified studies, 6,680 articles were excluded upon reviewing the titles and abstracts because they were irrelevant (were not focusing on GNEBPs, diarrhoea, and drug resistance or were outside Ethiopia or duplicates). The remaining 669 articles were assessed for eligibility; of these, 626 articles were excluded since they were review, KAP, or meta-analysis studies. Finally, 43 studies meeting the inclusion criteria were included in this study. Selected articles were focusing on one or more GNEBPs. Fig 1 shows a flow diagram of the selection of articles for the analysis. Table 1 shows the overall characteristics of articles included in the analysis, type and prevalence of Gram-negative isolates recovered from diarrheic patients. Studies were conducted in 8 regions of the country and most of the published articles were from the Amhara region (30.23%) followed by the Oromia region (18.60%) and South Nation and Nationalities Region (SNNP) (18.60%). Published articles were not found in other regions of the country (Afar, Somali, and Benishangul Gumuz regions) (Fig 2).

Characteristics of published articles
All studies were cross-sectional; conducted from 2006 to 2020 and published online from 2010 to August 2021. Almost all studies (97.67%) were institution-based (conducted on patients visiting health facilities). The stool samples were the specimen used to isolate the bacterial species. Isolates were characterized and their identities were confirmed by cultural and conventional biochemical tests, but molecular techniques were not used. Patients suffering from diarrhoea were used as the population of the studies and three studies were conducted on diarrhoeic patients with HIV. More than half of the studies (55.81%) were focusing on children less than equal to fifteen years of age. However, 39.53% of studies were considering all age groups (Fig 3 and Table 3).
The distribution of the studies using a funnel plot (Fig 5) showed the asymmetrical distribution of effect estimates; hence, there was a publication bias. To minimize the effect of the bias, subgroup analysis was used. Regionally, the pooled prevalence of GNEBPs from diarrheic patients in Addis Ababa, Amhara, SNNPs and Oromia were 20.08, 16.67, 12.12, and 11.61%, respectively. The pooled prevalence based on the study period was 14. 91, 18.03 and 13.46% among studies from 2006-2010, 2011-2015 and 2016-2021, respectively. The pooled prevalence was the highest (20.35%) in children having age less or equal to 15 years, followed by all age groups (10.83%) and adults greater than 15 years (8.51%) ( Table 2). Table 3 shows the types of bacterial isolates reported by published articles from diarrheic patients. Shigella spp. were the most frequent isolate (41.67%), followed by S. enterica (38.09%). The pooled estimate of E. coli was the highest (19.79%) among enteric bacterial isolates. The pooled prevalence of Campylobacter spp., Shigella spp. and Salmonella enterica. were 10.76, 6.24 and 5.06%, respectively. https://doi.org/10.1371/journal.pone.0265271.g001 Table 1. Characteristics, quality, and the number of Gram-negative isolates recovered from diarrheic patients. Escherichia coli was the most common (15.95%) isolate among children less than or equal to fifteen years of age whereas Salmonella enterica and Shigella spp. were common among studies that were focused on all age groups (Tables 1 and 4).

Drug resistance patterns of Gram-negative enteric bacterial pathogens
Drug resistance in Shigella spp. Twenty-six antimicrobial panels were used to assess the drug resistance pattern of Shigella spp. The highest percentage of resistance was reported among members of the penicillin group such as ampicillin (85.01%) and amoxicillin (82.07%). Resistance against the tetracycline group was also very common (67.01%). A considerable proportion of resistance was also reported among cephalosporin groups, particularly on the firstgeneration agents and erythromycin. Resistance was not reported against carbapenems (Table 5).  Drug resistance of Salmonella enterica. Twenty-five antimicrobial panels were used to assess the drug resistance patterns of S. enterica. The highest percentage of resistance was reported among members of the penicillin group such as ampicillin (64.98%) and amoxicillin (82.89%). Resistance against tetracycline and trimethoprim-sulphamethoxazole were also  common. A considerable proportion of resistance was also reported against erythromycin and cephalosporin groups. Resistance was not reported against doxycycline, meropenem and azithromycin (Table 6). Drug resistance of Escherichia coli. Sixteen antimicrobial panels were used to assess the drug resistance pattern of E. coli. The highest percentage of resistance was reported on agents like ampicillin (77.97%). Resistance against tetracycline (76.87%) and trimethoprim-sulphamethoxazole (66.97%) were also common. A considerable proportion of resistance was also reported among cephalosporin groups. Resistance was not reported against meropenem (Table 7).
Drug resistance of Campylobacter spp. Eighteen antimicrobial panels were used to assess the drug resistance patterns of Campylobacter spp. The highest percentage of resistance was  reported for Cephalothin (81.52%). Resistance against ampicillin (65.61%) and trimethoprimsulphamethoxazole (52.6%) were also common. Resistance was not reported against meropenem and azithromycin (Table 8).
Drug resistance of other enteric bacterial species. The number of published articles on other enteric bacterial species from diarrheic patients was very limited and impossible to summarize. However, one study conducted by Zenebe et al. [21] reported the presence of Klebsiella spp., Proteus spp., Enterobacter spp. in under-five children with diarrhoea. These bacteria were showing antimicrobial resistance character as indicated in Table 9.

Discussion
Diarrhoea is a common health problem, causing mortality and morbidity for thousands of people around the globe. Both infectious and non-infectious agents can induce the problem, among the infectious agents, enteric bacterial pathogens like diarrheagenic E. coli, S. enterica, Shigella spp., and Campylobacter spp. play important roles in the induction or severity of diarrhoea [61]. Determining their burden in a given population is very essential to design strategies Almost all studies (97.67%) were institutional, and samples were collected from patients visiting health facilities. Focusing only on health facilities may not reflect the overall prevalence of GNEBPs and their drug resistance patterns in the country. Since some GNEBP infection cases  may not arrive at health institutions and widespread use or misuse of antimicrobial drugs in the community may accelerate the occurrence of antimicrobial resistance [64,65]. Sub-grouping of the prevalence of GNEBPs based on the studies conducted in different regions of the country showed that the pooled prevalence was high in Addis Ababa (20.08%). In line with this, a spatial variation across the regions of Ethiopia was reported by Bogale et al. [66]. The difference based on the study period was not significant. However, a declining pattern of diarrhoea at the national level was reported by Bogale et al. [66]. Published articles were from 8 regions of Ethiopia, published articles were not found in Afar, Somali and Benishangul Gumuz regions of the country in this study. Hence, the pooled prevalence was calculated from 8 regions ignoring others. However, the scenario may not be equivalent to the region of the country having no reports.
Grouping of study participants based on their age showed that the pooled prevalence was the highest (20.35%) in children having age below or equal to 15 years which indicated that children are more exposed to the GNEBPs and express severe syndromes to visit health facilities. In line with this, Kotloff [4] and Havelaar et al. [67] reported that children contribute a huge proportion of diarrheal diseases in the world.
Diarrheagenic E. coli, Shigella spp., S. enterica and Campylobacter spp. were the most common isolates among GNEBPs. In line with this, Getie et al. [62] reported that Shigella spp. enterohemorrhagic E. coli (EHEC) and S. enterica were important isolates of GNEBPs among food handlers. In this study, the pooled estimate of E. coli was the highest (19.79%) among enteric bacterial pathogens. The result is very close to the report of Zenebe et al. [68] who reported that the pooled prevalence of E. coli was 25% in Ethiopia. A 33.8% pooled prevalence was reported by Oppong et al. [69] in Sub-Sharan Africa. Oppong et al. [69] also found that E. coli detection was the highest in the East African region and lowest in the middle part of Africa. The difference may be due to the number of studies in the summary and the targeted strain of E. coli. For example, in a single study in Niger, 11.1% of diarrhoeic children were positive for diarrheagenic E. coli [70]. Similarly, the Global enteric multicentre study on infants and children showed that diarrheagenic E. coli was among the four major pathogens responsible for diarrhoea in low-income and middle-income countries [71].
The pooled prevalence of Campylobacter spp. in this analysis was 10.76%. In line with this report, Kassie et al. [72] reported a prevalence of 10.5% among children in Denbia district, Ethiopia and 13.8% prevalence was also reported by Gedlu and Aseffa [73] among children in northwest Ethiopia. Oppong et al. [69] reported a 12.3% pooled prevalence of Campylobacter in East Africa. A high rate of Campylobacter infection among children was also reported in Kenya [63]. In contrary to the report of this study, Fletcher et al. [74] reported a pooled prevalence of 2.7% in Sub-Saharan countries. The difference may be related to the area coverage, disease prevention and control practices. The pooled prevalence of Shigella spp. in this study was 6.24% which is equivalent to the report of Hussen et al. [75]. According to their report, the pooled prevalence of Shigella spp. in Ethiopia was 6.6%. Similarly, Oppong et al. [69] reported a pooled prevalence of 5.6% from children under five and Fletcher et al. [74] 4.3% of children aged less than 12 years in Sub-Saharan countries.
The pooled prevalence of S. enterica was 5.06% which is almost comparable with the reported pooled prevalence by Abate and Assefa [76], which was 4.8% among human stools and animal origin foods in Ethiopia.
Resistance to antimicrobial agents is a natural evolutionary process for the bacteria, however, the process is accelerated by human activities in terms of antimicrobial usage patterns and infection control or prevention practices. The risks are very high in developing countries like Ethiopia where there is a widespread use or misuse of antimicrobial agents with a high burden of infectious diseases [76][77][78][79].
In this meta-analysis, Shigella isolates were more resistant to the penicillin group of antimicrobial agents like ampicillin (85.01%) and amoxicillin (82.07%). A high percentage of resistance against ampicillin (83.1%) and amoxicillin (84.1%) were also reported by Hussen et al. [75]. There were also reports of drug resistance among the first-line drugs like ciprofloxacin (11.86%) and ceftriaxone (29.53%) for the treatment of Shigellosis. Resistance development against such types of antimicrobial agents was also reported by Hussen et al. [75]. They reported 8.9 and 9.3% resistance against ciprofloxacin and ceftriaxone, respectively.
In this analysis, high proportions of Salmonella isolates were also resistant against the penicillin group of antimicrobial agents like ampicillin (64.98%) and amoxicillin (82.89%). In line with this Tadesse [80] was also reported a high percentage (86.01%) of resistance of Salmonella against ampicillin and/or amoxicillin. Resistance of Salmonella isolates to fluoroquinolone like ciprofloxacin (2.27%) and third-generation cephalosporin (ceftriaxone 16.68%) were also reported. Similarly, according to Tadesse's report, the pooled prevalence of ciprofloxacin resistance among Salmonella isolates was 3.61% [80] and a prevalence of 2.9% of resistance against ciprofloxacin was also reported in Iran [81].
Resistance was common among E. coli isolates in this analysis, particularly on antimicrobial agents like ampicillin (77.97%), tetracycline (76.87%) and trimethoprim-sulphamethoxazole (66.97%). Resistance of E. coli against a wide array of antimicrobials was also reported by Pormohammad et al. [82], Zenebe et al. [68] and Tuem et al. [83]. Resistance in non-pathogenic strains of E. coli may not have a direct effect on health, however, non-pathogenic resistant strains may acquire virulence genes and induce disease that may not be treated easily or nonpathogenic strains having resistant character may act as a reserve for the resistant character for other bacteria [82].
The MDR report among Campylobacter isolates in this analysis (80.78%) was higher than the report from Bangladesh (28.8%) [97] and Kenya 50% [98] but lower than the report from Ghana (97%) [84]. This may be due to differences in the use of antimicrobial agents, the area coverage, sample type and technique of detections. The MDR report of E. coli in this analysis (78.20) was not in line with other reports like 50% prevalence in Nigeria [99], 40% in Spain [95], 22% among human isolates in the world [82], 26% in China [100], 39.8% in Egypt isolates from animals [101], 28% in low and middle-income countries [102]. The MDR report of this meta-analysis among Shigella isolates (79.08) was almost similar to the report by Hussen et al. [75] (83.2%) but it was lower than other reports from Iran (89.4%) [103], and Bangladesh (94%) [104]. However, it was higher than other reports such as 53.8% among migrants in Europe [74], 60% in Kenya [98] and 19% in Somalia [105]. Factors like the magnitude and style of antimicrobial use and infection prevention practices may play roles in the differences.
The development of MDR character among S. enterica is also an important public health concern around the globe [106]. In this analysis, the prevalence of MDR character among Salmonella isolate was 59.46% which is comparable with the MDR report of Garedew et al. [107] (46.2%) among food handlers in Gondar. However, it was lower than the report from Dagnew et al. [108] (76%) and Admassu et al. [109] (100%). The difference may be due to the target population and study methods (single versus pooled meta-analysis reports).

Limitations of the study
Some of the studies included in the analysis were targeting the most common bacterial pathogens that did not rule out the absence of others. Therefore, for bacteria that are thought to be less frequent, the reported prevalence may not accurate. Publication bias and heterogenicity were observed in the analysis, but attempts were made to reduce their impact on the analysis by following the random effect model and subgrouping. However, these may not totally avoid their impact on the interpretation of the pooled results.

Conclusion
According to this analysis, the burden of Gram-negative enteric bacterial pathogens (GNEBPs) was high and may be considered a major cause of diarrhoea in Ethiopia. A significant proportion of the isolates exhibited resistance to the commonly used antibacterial agents which are expected to affect the treatment response and cost, morbidity and progression of the infection. Shigella spp., S. enterica, E. coli and Campylobacter spp. were the commonly isolated GNEPB from diarrheic patients in the country. The pooled estimate of Campylobacter spp. was the highest followed by E. coli, Shigella spp. and S. enterica. Resistance to antimicrobial agents was most common among the penicillin groups, followed by tetracycline, and trimethoprim-sulphamethoxazole. However, there were also resistant strains against very relevant drugs for the treatment of GNEBP such as fluoroquinolones and thirdgeneration cephalosporins. Almost all isolates were susceptible to meropenem.
Since the incidences of these bacterial diseases are related to hygiene, all activities that enhance hygienic practices (clean water and food, handwashing, proper use of latrine) must be advocated and implemented. Performing drug sensitivity tests for suspected diarrheagenic bacteria is extremely advantageous to select the appropriate antimicrobial drugs for the treatment. The antimicrobial resistance surveillance system must be established to understand the trend of resistance among pathogenic bacteria and to plan and implement mitigating strategies like proper control and prevention of infectious diseases and antimicrobial stewardship programs. Almost all studies were using conventional techniques for confirmation of the isolates, thus, adding molecular methods in the future will increase analysis precision.