This systematic review and meta-analysis followed the Preferred Reporting Items for Systematic Reviews and Meta-Analyses (PRISMA) guidelines (24) (S1 File). This protocol has been registered in PROSPERO with ID CRD42022329641.

Comprehensive searches were conducted using the databases of PubMed, CINAHL, Google Scholar, and African Journals Online to find potentially relevant articles. In addition, accessible Ethiopian higher education institutional repositories and the cross-reference lists of already identified articles were also systematically searched to obtain similar articles. Electronic database searches were initially conducted on April 10–18, 2023, and then updated on February 27–29, 2024. The key MeSH terms used for the PubMed database searches were “surgical wound infection,” “caesarean section,” and “Ethiopia.” For PubMed’s entry terms, “all fields” is used. For CINHA, Google Scholar, and African Journals Online, the same key words were used with “all fields.” For institutional repositories, we used the phrase “Surgical site infection among cesarean section women in Ethiopia.” See more comprehensive search strategies in the S2 File.

The search strategy was implemented by two authors (TGW and JAM). Studies identified through different database searches were combined and exported using Endnote X8.1 software. Duplicated studies were removed with Endnote X8.1 software. TGW and JAM independently assessed the selected articles for their applicability to the review objective using their titles and abstracts. After the initial screening, the full texts of all studies considered relevant were obtained. Two reviewers (TGW and JAM) separately examined the eligibility of the full texts. Disagreement between the two reviewers (TGW and JAM) reached by scientific consensus.

Incidence is a measure of risk that is the probability that a subject within a population will develop a given disease or other health outcome over a specified follow-up period. It can be calculated by dividing the number of subjects developing the disease over a certain period by the total number of subjects followed over that period (25). In this study, incidence is a measure of the risk of the probability that women delivered via cesarean section developed a surgical site infection over a specific follow-up period. Two parameters were required to estimate the incidence of surgical site infection: the number of cases who develop surgical site infection over a specific follow-up period and the total number of women delivered via cesarean section in the same period of time. The incidence was calculated by dividing the number of patients who developed surgical site infections over a certain period by the total number of women who undergo cesarean sections (sample size) over that period.

Using a standard data extraction format, the two authors independently extracted the necessary information from each study. The disagreement between two data extractors was resolved by scientific consensus. The primary author, publication year, region, study design, sample size, number of women with surgical site infection, and percentages of surgical site were retrieved for each included study (S3 File).

The methodological quality of the included studies was assessed by two independent authors (TGW and JAM). Any disagreement between the two authors was settled by scientific consensus. The quality of each included cross-sectional and cohort study was assessed by the Joanna Briggs Institute’s quality appraisal checklist (26). Studies that scored at least 50% were considered low-risk and included in the final meta-analysis.

All the necessary data was retrieved from the articles using a Microsoft Excel spreadsheet and imported to STATA software version 15 for further analysis (27). A meta-analysis using the random-effects DerSimonian and Laird models was used due to expected heterogeneity (28). The random-effects model was employed when heterogeneity was greater than 50% (based on the I² statistic) (29). Subgroup and sensitivity analyses were performed to find potential heterogeneity moderators when significant heterogeneity existed. Publication bias was checked by objectively computing the Egger test (30) and subjectivity inspecting funnel plots (31). The trim-and-fill method was applied when publication bias existed (32). The estimated pooled incidence of surgical site infection was presented by forest plots with a 95% CI. The pooled odds ratio was computed to determine factors associated with surgical site infection.

In the first search, 230 articles related to surgical site infections were retrieved via electronic databases. One hundred eleven were removed due to duplication, while 80 articles were excluded after reading their titles and abstracts due to their irrelevance. The remaining 39 articles were subjected to a full text review, 16 of which were removed because 15 studies failed to include women who had undergone cesarean sections as study participants, and one did not report an outcome of interest. Twenty three eligible studies were included in the final analysis. The specific screening procedures are depicted in a PRISMA flow chart (Figure 1).

The final analysis included 23 studies with 10,201 participants; of them, 1,281 developed surgical site infections. Seventeen studies were cross-sectional, and six were cohort studies. The sample sizes of the studies ranged from 166 (33) to 1,069 (34). In the southern nation nationality people (SNNP) (35) and Amhara region (36), the highest (79.2%) and the lowest (2.4%) surgical site infection were reported, respectively. There were nine studies conducted in the Amhara region and four in the SNNP region. Addis Ababa and Oromia each had three studies, while two were from Tigray region and the other two from Harari region. All of the included studies had a low risk of bias (Table 1).

The quality of each cross-sectional and cohort study was independently evaluated by the two authors (TGW and JAM). The quality score of each included primary study, based on the JBI quality appraisal criteria, showed low risk for all studies. For cross-sectional and cohort studies, the quality score ranges from 6 to 8 and 8 to 11, respectively. The quality assessment of the generally agreed-upon conclusions from the primary studies that were included is displayed in S4 File.

The pooled incidence of surgical site infection among women delivered via cesarean section was 12.32% (95% CI: 8.96–16.11). There was a significant amount of heterogeneity between studies (I² = 96.73%, p < 0.001; Figure 2).

Due to the significant heterogeneity across the included studies, a sub-group analysis based on study region, study design, publication year, and sample size was taken into consideration to identify a possible source of heterogeneity. The incidence of surgical site infection was highest in the southern nation nationality of people (SNNP) at 26.2% (95%CI: 5.4–55%), followed by Addis Ababa at 11.14% (95%CI: 8.1–14.6%), and Harari at 11.1% (95%CI: 9–13%). The highest heterogeneities (I²), which were 99.3% (p < 0.001) and 94.14% (p < 0.001), were observed in the southern nation nationality of people (SNNP) and Amhara region, respectively. The remaining regions lacked heterogeneity. There were comparable incidences of surgical site infection in the cohort and cross-sectional studies, with significant heterogeneity. In addition, in the sample size of less than or equal to 420 and in the sample size of greater than 422, comparable magnitudes of surgical site infection were noted. In the publication years 2020 and after, surgical site infection was 16% (95%CI: 9.2–24.28%) with significant heterogeneity (I² = 98.48%, p < 0.001); in the publication year prior to 2020, it was 9.4% (95%CI: 8.4–10.5%), with moderate heterogeneity (Table 2).

In this systematic review and meta-analysis, the funnel plot indicated the asymmetric distribution of the studies (Figure 3), but the Egger’s test revealed no statistical significance [bias =1.61 (95% CI: −0.2.84–6.1), p = 0.46].

A sensitivity analysis was carried out to see the effect of an individual study on the pooled effect size. According to the leave-one-out sensitivity analysis (Figure 4), there were no visible differences. The results of the meta-analysis showed it was stable.

Place of residence, general anesthesia, rupture of the membrane, chorioammionitis, and a post-operative Hgb level less than 11 g/dL were risk factors for surgical site infection, but emergency cesarean section, hypertension, interrupted skin closure, and midline incision did not have a significant association with surgical site infection.

Three studies (35, 42, 52) were identified to check the association between rural residence and surgical site infections. All three studies found that rural residence is a risk factor for surgical site infection compared to urban residence. The findings of this meta-analysis revealed that rural residents were 2.5 times more likely to develop surgical site infections compared to urban residents (AOR = 2.51, 95% CI: 1.15–3.87). No evidence of heterogeneity was reported (I² = 0.0%, p = 0.46; Figure 5).

Five studies (38, 40, 42, 43, 45) were used to examine the relationship between hypertension and surgical site infection. Three of these primary studies (40, 42, 45) revealed a significant association, while the other two studies (38, 43) reported an insignificant association. The result of the random effect model meta-analysis showed that there is no significant association between hypertension and surgical site infection (AOR = 1.94, 95% CI: 0.15–3.72). Moderately insignificant heterogeneity was found (I² = 46.8%, p = 0.111; Figure 6).

Six studies (35, 38, 45, 50, 52, 55) were found to determine the association between chorioammionitis and surgical site infection, and all of the studies’ findings reported a statistically significant association. In the current meta-analysis, women with chorioammionitis during cesarean section were 4 times more likely to develop surgical site infection than women without chorioammionitis (AOR = 4.13, 95% CI: 1.45–6.8), and there was no evidence of heterogeneity (I² = 0, p = 0.964; Figure 7).

Six studies (35, 37, 38, 41, 49, 52, 54) were included to determine the relationship between emergency cesarean sections and surgical site infections. Except for the one study (35) that found a significant association, the remaining five studies (37, 38, 41, 49, 52, 54) did not find a significant association. This meta-analysis revealed that no relationship existed between emergency cesarean section and surgical site infection (AOR = 1.09.95% CI: 0.95–1.24), with no heterogeneity (I² = 0, p = 0.721; Figure 8).

The relationship between membrane rupture and surgical site infection was examined in five studies (34, 37, 41, 46, 52, 53), while the relationship with membrane rupture greater than or equal to 12 h was examined in five studies (38, 44, 45, 47, 55). One study for mothers with membrane rupture (37) and the other study for those with membrane rupture greater than or equal to 12 h (45) did not show a significant association, while the remaining study had a significant association. Mothers with membrane ruptures (AOR = 2.04, 95% CI: 1.24–2.85), as well as those ruptured of membrane greater or equal to 12 h (AOR = 3.93, 95%CI: 1.93–5.92), were more likely to develop surgical site infection than mothers with intact membrane and ruptured membrane less than 12 h, respectively (Figure 9).

To examine the effect of general anesthesia on surgical site infection, four studies were identified (34, 40, 45, 46). Half of the studies (34, 40) found a significant association between general anesthesia and surgical site infections; the other half (45, 46) did not detect such a significant association. The results of this meta-analysis showed that women who underwent general anesthesia during cesarean section were more likely to develop a surgical site infection than those who underwent spinal anesthesia (AOR = 1.99, 95% CI: 1.22–2.75), and there was no evidence of heterogeneity (I² = 0, p = 0.893; Figure 10).

Three studies (41, 45, 49) were used to examine the relationship between longitudinal incision and surgical site infection, while four studies (37, 40, 46, 50) were used to examine the relationship between midline incision and an infection at the surgical site. In the longitudinal incision, only one study (41) had a significant association, whereas in the midline incision, only one study (40) did not have a significant association. The remaining had significant associations in both midline and longitudinal. In this meta-analysis, neither a midline (AOR = 2.77, 95% CI: 0.68–4.87) nor a longitudinal (AOR = 1.08, 95% CI: 0.6–2.76) skin incision showed a significant association with the surgical site in comparison to a low transverse skin incision (Figure 11).

Three studies’ findings were reviewed to assess the relationship between skin closure and surgical site infection (40, 41, 45). One study’s findings indicated a significant relationship (41), whereas two studies’ findings showed no significant relationship (40, 45). The findings of the meta-analysis reported that no significant relationship was revealed between interrupted skin closure and surgical site infection as compared to subcuticular skin closure (AOR = 0.66, 95%: −0.19–1.52), with no heterogeneity (I² = 0, p = 0.421; Figure 12).

The association between post-operative Hgb level and surgical site infection was determined by four studies (37, 40, 50, 54). All of the studies found a significant association. In this systematic review study, cesarean-delivered women who had post-operative Hgb levels less than 11 mg/dL during cesarean-section had a nearly 3 times higher risk of developing surgical site infection than women who had post-operative Hgb levels higher than 11 mg/dL (AOR = 3.25, 95% CI: 1.54–4.96), with the absence of heterogeneity (Figure 13).