When conducting the systematic review, we followed the guidelines of the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) criteria (Liberati et al., 2009; Supplementary Table S1).

The search was conducted in various databases, including PubMed, Web of Science, Science Direct, Google Scholar, African Journals Online, and the Cochrane Library. The scope of the literature search was limited to English-language studies examining knowledge, attitudes, and practices related to antimicrobial resistance in Africa. A systematic approach was followed to conduct an advanced search on PubMed. Initially, search terms were formulated for four key areas: “antimicrobial resistance,” “knowledge,” “attitudes,” and “practices.” These keywords were retrieved from Google Scholar and then individually searched in PubMed to identify relevant MeSH terms within the MeSH hierarchy tree. Searching strategies were established by using Boolean operators (“OR” or “AND”) to bring the concepts together effectively. The search was also done by combining the above search terms with the names of all countries included in Africa. The search period was from July 1/2023 to August 10/2023.

Condition (Co): This comprehensive review and meta-analysis included studies that assessed knowledge, attitudes, and practices related to antimicrobial resistance. Context/settings (Co): All observational epidemiological studies conducted in Africa. Population/participants (P): patients, the general population, healthcare workers, pet owners, and students.

This study included various research articles, including full-text, English-language, published and unpublished, cross-sectional, case–control, and cohort studies. Excluded from the analysis were duplicate sources, intervention studies, case reports, systematic reviews, qualitative articles, case series, conference abstracts, letters to the editor, and any article that remained inaccessible after at least two email exchanges with the primary author. The COCOPOP (Condition, Context, and Population) paradigm was utilized to determine the suitability of the included studies for this investigation. The study population (POP) consisted of diverse participants, including patients, the general population, healthcare workers, ranchers, and students. The setting of the study was Africa.

The two authors, YAA and KAG, conducted separate assessments of study quality using the Joanna Briggs Institute (JBI) standardized quality assessment checklist (Aromataris et al., 2017). Our colleague, Natnael Atnafu Gebeyehu, moderated a discussion to resolve discrepancies during the evaluation process. This joint effort ultimately resolved the dispute, resulting in a mutual agreement. The critical analysis checklist consists of eight parameters and offers the selection options Yes, No, Unclear, and Not applicable. Studies were evaluated using quality assessment indicators to determine the risk level. The studies that achieved 50% and above were considered low risk, as indicated in an attached additional file (Supplementary Table S2).

We used the methods described by Hoy et al. proposed methods (Hoy et al., 2012) To assess the internal and external validity of the study using 10 specific criteria. These criteria were used to assess the potential risk of bias. Each criterion was assigned either a low or high risk of bias. Articles that lacked precise assessment tools for data collection were classified as high-risk of bias articles. Ultimately, the overall risk of bias was classified as low (total score ≤ 2), moderate (total score between 3 and 4), or high (total score > 5) based on the likelihood of bias (Supplementary Table S3).

YA and KA, two independent authors, used a standardized Microsoft Excel spreadsheet to extract relevant data from the studies. To maintain consistency, they used the JBI prevalence studies tool as a reference point for data extraction from the selected articles (Aromataris et al., 2017). The data extraction tool includes several details, including the author’s name and year of the study, the title of the study, the year of publication, the geographical area and country of the study, the sub-region, the design and type of the study, the study population, the sample size, the response rate, the measured outcome, and the prevalence rate. In addition, information on publication status was also collected.

Data were extracted using Microsoft Excel and analyzed using STATA version 14 statistical software. Significant heterogeneity between studies was assessed using Cochrane Q and I2 statistics. A forest plot was used to represent the heterogeneity visually. A random effects model was used to determine the pooled effect employed as a significant degree of heterogeneity was found. Subgroup analysis was performed based on the subregion, study population, and design. A sensitivity analysis was conducted to examine the influence of a single study on the overall prevalence estimate of the meta-analysis. Publication bias was assessed using funnel plots and Eggers regression tests. If the p-value was less than 0.05, it indicated a statistically significant presence of publication bias. The overall effect of knowledge, attitudes, and practices regarding AMR was assessed using a random effects model, and prevalence rates and odds ratios with a 95% confidence interval were calculated.

The present study is a comprehensive review and meta-analysis focused on examining the dissemination of knowledge, attitudes, and practices related to antimicrobial resistance in Africa. The researchers conducted a thorough search of electronic databases and found a total of 1,850 records. After removing 986 duplicate records, the remaining 864 articles were screened based on their titles and abstracts, excluding 762. Subsequently, 102 full-text articles were carefully assessed for eligibility. Of these, 63 articles were excluded and did not meet the specified criteria. Finally, 39 primary articles were selected for quantitative analysis, representing the most relevant and reliable sources of information for this study (Figure 1).

This systematic review and meta-analysis included 39 (Thriemer et al., 2013; Abera et al., 2014; Pereko et al., 2015; Simba et al., 2016; Asante et al., 2017; Farley, 2017; Genga et al., 2017; Tafa et al., 2017; El Sherbiny et al., 2018; Seid and Hussein, 2018; Mate et al., 2019; Sunusi et al., 2019; Fetensa et al., 2020; Gemeda et al., 2020; Kabba et al., 2020; Mengesha et al., 2020; Nisabwe et al., 2020; Sindato et al., 2020; Wesangula et al., 2020; Zulu et al., 2020; Aworh et al., 2021; Babatola et al., 2021; Chukwu et al., 2021; Gebeyehu et al., 2021; Geta and Kibret, 2021; Mostafa et al., 2021; Ogoina et al., 2021; Russom et al., 2021; Adegbite et al., 2022; Belachew et al., 2022; Dejene et al., 2022; Kanyike et al., 2022; Mudenda et al., 2022; Paye and McClain, 2022; Tembo et al., 2022; Abubakar and Sárváry, 2023; Davwar et al., 2023; Maarouf et al., 2023; Otieku et al., 2023) published studies involving 18,769 study subjects. The review included primary studies with a remarkably varied sample size, ranging from just 47 (Adegbite et al., 2022) to 2,477 (Russom et al., 2021). Of the included studies, 33 were cross-sectional (Thriemer et al., 2013; Abera et al., 2014; Pereko et al., 2015; Simba et al., 2016; Asante et al., 2017; Farley, 2017; Genga et al., 2017; Tafa et al., 2017; El Sherbiny et al., 2018; Seid and Hussein, 2018; Mate et al., 2019; Fetensa et al., 2020; Gemeda et al., 2020; Mengesha et al., 2020; Nisabwe et al., 2020; Sindato et al., 2020; Zulu et al., 2020; Aworh et al., 2021; Babatola et al., 2021; Chukwu et al., 2021; Gebeyehu et al., 2021; Geta and Kibret, 2021; Mostafa et al., 2021; Ogoina et al., 2021; Russom et al., 2021; Adegbite et al., 2022; Dejene et al., 2022; Kanyike et al., 2022; Mudenda et al., 2022; Paye and McClain, 2022; Abubakar and Sárváry, 2023; Davwar et al., 2023; Maarouf et al., 2023), while six studies followed a cohort study design (Sunusi et al., 2019; Kabba et al., 2020; Wesangula et al., 2020; Belachew et al., 2022; Tembo et al., 2022; Otieku et al., 2023). The study participants for the primary studies were selected by the author from various groups including health workers (Abera et al., 2014; Asante et al., 2017; Genga et al., 2017; Tafa et al., 2017; Kabba et al., 2020; Babatola et al., 2021; Chukwu et al., 2021; Ogoina et al., 2021; Adegbite et al., 2022; Kanyike et al., 2022; Mudenda et al., 2022; Paye and McClain, 2022; Tembo et al., 2022; Abubakar and Sárváry, 2023; Davwar et al., 2023), patients (Farley, 2017; El Sherbiny et al., 2018; Geta and Kibret, 2021; Otieku et al., 2023), Tertiary school students (Thriemer et al., 2013; Seid and Hussein, 2018; Sunusi et al., 2019; Fetensa et al., 2020; Nisabwe et al., 2020; Zulu et al., 2020; Mostafa et al., 2021; Maarouf et al., 2023), community members or the general public (Pereko et al., 2015; Simba et al., 2016; Mate et al., 2019; Mengesha et al., 2020; Sindato et al., 2020; Wesangula et al., 2020; Russom et al., 2021; Belachew et al., 2022; Dejene et al., 2022), and livestock farmers (Farley, 2017; Gebeyehu et al., 2021; Geta and Kibret, 2021; Otieku et al., 2023). Furthermore, The majority of the primary studies included in this review were research reports from East Africa (Abera et al., 2014; Simba et al., 2016; Genga et al., 2017; Tafa et al., 2017; Seid and Hussein, 2018; Fetensa et al., 2020; Gemeda et al., 2020; Mengesha et al., 2020; Nisabwe et al., 2020; Sindato et al., 2020; Wesangula et al., 2020; Zulu et al., 2020; Gebeyehu et al., 2021; Geta and Kibret, 2021; Russom et al., 2021; Belachew et al., 2022; Dejene et al., 2022; Kanyike et al., 2022; Mudenda et al., 2022; Tembo et al., 2022; Table 1).

A comprehensive analysis was conducted on 17,969 participants to determine their knowledge of antimicrobial resistance. The study included 20 research studies in East Africa, nine in West Africa, four in North Africa, three in Southern Africa, and two in Central African countries. The overall pooled prevalence level of knowledge towards antimicrobial resistance in Africa was determined using the random-effect model with Mantel-Hanenszel heterogeneity at 55.33% (95% CI: 47.48, 63.18). This estimate was statistically significant, with a p-value of less than 0.001. Furthermore, heterogeneity between studies was found to be high, with an I² value of 99.3% (Figure 2).

A Pooled analysis was conducted on a sample of 12,039 participants to determine the level of positive attitudes regarding antibiotic drug resistances. This included 14 studies in East Africa, six in West Africa, four in North Africa, three in Southern Africa, and one in Central Africa. The overall pooled prevalence of positive attitudes toward antimicrobial resistance in Africa was 46.93%(95%CI: 35.10, 58.76), p < 0.001, I2 = 99.6% (Figure 3).

A total of 12,627 study participants were examined to estimate the level of good practices of antimicrobial drug resistances, which included 12 studies in East Africa, seven studies in West Africa, three studies in Northern Africa, two studies in Southern Africa, and one study in Central Africa country. The overall pooled prevalence of appropriate practice toward antimicrobial resistance in Africa was 51.05% (95%CI: 45.24, 56.87), p < 0.001, I² = 97.8% (Figure 4).

The subgroup analysis was performed in this meta-analysis by the study population. Among students, it was found that the level of good antimicrobial-resistant knowledge was 66.20% (95% CI: 51.77, 80.64), with the I² statistic from Higgins showing high heterogeneity (I²: 99.0%, p < 0.001). However, among healthcare workers, the level of good knowledge about antimicrobial resistance was 59.30% (95% CI: 45.19, 73.42), which was also associated with high heterogeneity (I²: 99.3%, p < 0.001). Likewise, Tertiary school students showed the highest positive attitudes towards antimicrobial resistance, with 55.81% (95% CI: 14.73, 96.90). Higgins and Thompson’s I² statistic showed a significant level of heterogeneity with a p-value of less than 0.001. Furthermore, the highest pooled prevalence of good practice was observed among students, with 57.94% (95% CI: 45.78, 70.10). Higgins and Thompson’s I² statistic also showed a significant level of heterogeneity (I²:99.5%) with a p-value of less than 0.001.

Similarly, Sub-group analysis was conducted to examine the effects of sub-region and study design type. It was found that the study conducted in North Africa had the highest pooled prevalence of good knowledge of antimicrobial resistance with 60.34% (95% CI: 46.76, 73.92); On the other hand, looking at the sub-region, West Africa had the lowest pooled prevalence at 48.62% (95% CI: 30.00, 67.23), and Higgins I² statistic showed a high degree of heterogeneity (I²: 97.7%, p < 0.001) between studies. Likewise, the highest pooled prevalence of positive attitudes of AMR was observed in the study conducted in East Africa at 60.20% (95% CI: 43.56, 76.85) with high heterogeneity (I²:99.6%, p < 0.001). Furthermore, a sub-group analysis was based on the study design type. The prevalence among the studies that used cross-sectional study design was relatively higher than those that used cohort study design for each on the level of good knowledge, positive attitudes, and good practices of AMR (Table 2).

A leave-one-out sensitivity analysis was performed to examine the possible influence of individual studies on the overall pooled prevalence effect. The results showed that no specific study significantly impacted the overall prevalence of good knowledge, positive attitudes, and good antimicrobial resistance (AMR) practices in Africa. The results showed that the combined effect remained significantly unchanged after excluding a specific study.

A meta-regression analysis was performed to check the presence of statistically significant heterogeneity, as indicated by an I-square test statistic of less than 0.05. The primary objective of this analysis was to determine the underlying source of heterogeneity to ensure an accurate interpretation of the research findings. However, the results of the meta-regression analysis revealed no significant variables that could account for the observed heterogeneity. Specifically, no statistically substantial study-level covariates, such as sample size or publication year, could explain the variations in good knowledge levels, positive attitudes, and good practices towards AMR. Consequently, the heterogeneity observed in this review is likely attributable to other factors not considered in the analysis (Table 3).

Assessment of publication bias was performed using subjective and objective approaches. Visual assessment included analysis of a funnel plot, which revealed a slightly asymmetrical distribution of studies. In addition, objective evaluation was performed using Egger’s and Begg’s tests, with the significance level set at p < 0.05. The results of the formal Egger linear regression test showed that there was no statistically significant good antimicrobial-resistant knowledge level (p = 0.596), positive attitude of antimicrobial resistance (p = 0.90), and good practices of AMR (p = 0.109). In addition, the presence of publication bias in the included studies was assessed using the Begg regression test. Nevertheless, the results of the Begg regression test showed no evidence of publication bias for the variables good knowledge about antimicrobial resistance (p = 0.669), positive attitude towards antimicrobial resistance (p = 0.161), and practices (p = 0.168) (Figures 5A–C), respectively.