Systematic review and meta-analysis using computerized databases
Patients were not directly involved in the design of this study.
To identify potentially relevant articles, a comprehensive search was carried out on Pub Med/MEDLINE, Google Scholar, Scopus, and Science Direct databases. Our search extended by hand searches for grey literature and other relevant literature collections also done by retrieving reference lists of eligible articles. The search of the literature was conducted between January and April 2019 and considered all relevant publications from January 2016 to April 2019. The search protocol was formulated by using common key words ‘prevalence AND associated factor OR determinants AND birth asphyxia OR perinatal asphyxia OR neonatal asphyxia OR Neonates OR neonatal intensive care unit OR Hospitals AND Ethiopia. We presented this meta-analysis according to the Preferred Reporting Items for Systematic Reviews and Meta-Analysis (PRISMA) guide line(12)(Additional file 1).
Study area: All studies conducted in Ethiopia
Study design: All observational study designs reporting the prevalence of PNA
Population: Studies involving infants immediately before, during and after birth within seven days of postpartum
Language: Only articles reported in English language
Publication condition: published and unpublished articles
Exclusion criteria: case reports, national reports, clinical studies, reviews, and articles which have not defined criteria for PNA and unable to assess the quality of articles due to the absence of full text.
Primary outcome (PNA) was diagnosed by persistently low APGAR (Appearance, Pulse, Grimace, Activity, and Respiration score at 5th min less or equal to 6(less than 7) and it was calculated as the total number of PNA divided by the total number of live birth infants in the study multiplied by 100.
The second outcome is the associated factors which influence the primary outcome PNA in the form of the log odds ratio. For key associated factors, the odds ratio determined based on binary outcomes; prolonged labor (<24hours vs. >24hours), birth weight (<2500 grams vs. >2500 grams), and meconium stained (yes vs. no) from primary studies.
All relevant data were extracted by two reviewers (DB and SA) independently from included articles; discussions and mutual consensus were in place when divergent points of view were raised with third author (GT). The data extraction format included primary author, publication year, regions of the country, study area, sample size and prevalence with 95% confidence interval.
Risk of bias and quality appraisal
The risk of bias for included studies was assessed using the Joanna Briggs Institute’s(JBI) critical appraisal checklist for prevalence studies(13). In addition, included articles were evaluated for quality by two authors (BG and AM) independently and the reviewers compared their quality appraisal scores and resolved any discrepancy before calculating the final appraisal score by the third author (MA). To assess the quality of each study, the Newcastle-Ottawa Scale for cross-sectional and cohort studies was adopted(14). The tool has three sections; the first section focuses on the outcome and statistical analysis of each original study with a possibility of two stars to be gained. The second section graded from five stars considers the credibility on the methodological quality of each original study. The third section of the tool deals with the comparability of the study cases or cohorts, with a possibility of two stars to be given. Articles with a rating of ≥ 5 out of 10 stars were considered as high quality. In our case, all eligible studies met this standard.
Data processing and analysis
Data on study design, sample size, and others were extracted in Microsoft Excel format, and then analysis was carried out using STATA/se Version 14. Heterogeneity among reported prevalence was assessed by computing p-values of I2-statics. Higgins’s I2 statistic measures the percentage of variation between the sample estimates that is due to heterogeneity rather than to sampling error (15). The pooled prevalence of PNA was carried out with a random effects meta-analysis model, generating the pooled 95% confidence interval using the Der Simonian and Laird's methods. To minimize the random variations between the point estimates of the primary study; subgroup analysis was done for study settings/region, study design, and sample size. To identify the possible sources of heterogeneity, univariate meta-regression was undertaken for sample size and publication year, and regions. Publication bias was assessed by visually checking for asymmetry in funnel plots and objectively testing using Egger’s test at 5% significant level. Point prevalence as well as 95% confidence intervals were presented in the forest plot format. In this plot, the size of each box indicated the weight of the study, while each crossed line refers to 95% confidence interval.