Development review protocol and registration
The development of the review methodology has followed the procedural guideline that was endorsed by the preferred reporting items for systematic review and meta-analysis protocol (PRISMA-P) statement [16], and all of the items in the checklist were completed (see Additional file 1). The review protocol has been registered in international prospective register of systematic reviews (PROSPERO) with trial registration number (CRD42017081437).
Data source and searching strategies
The search of studies will be carried out by (MG and GT). Published and unpublished studies written in English will be retrieved and included into the review process. Databases such as MEDLINE (via PubMed), EMBASE, CINAHL, and POPLINE will be searched for studies that had been conducted since January 1, 1994. Relevant sources such as Google search engine, Google scholar, and WHO websites will be searched. In addition, experts on the field will be consulted to retrieve unpublished studies. The year 1994 was chosen because the international community recognized the pressing need to address unsafe abortion at the International Conference on Population and Development (ICPD) in the year 1994 [17] and many African countries endorsed semi-restricted abortion law since 1994 [18]. The search strings will emerge from the following keywords (unsafe abortion, induced abortion, abortion, Sub-Saharan Africa, or African South of Sahara). Depending on the specific requirement of the database, the search string will be modified, and relevant studies using search strings will be identified. The combinations of free keywords and MeSH (medical sub-headings) will be extensively used in the search process. The reference lists of relevant studies will also be reviewed for sources that may have been missed in the database search. The search strategy developed for selected database is attached (see Additional file 2).
Eligibility criteria
All observational studies (cross-sectional, case-control, and cohort) and survey reports will be included in the systematic review. However, case reports, case series, commentaries, and editorials will be excluded from the review. All studies with primary objective to determine the prevalence of unsafe abortion and/or its associated factors among reproductive aged women in Sub-Saharan Africa will be considered [8].
We will consider studies that defined unsafe abortion based on WHO definition [19]; WHO defines unsafe abortion as a procedure of pregnancy termination either by persons lacking the necessary skills or in an environment that does not conform to minimal medical standards or both. We will also include community or facility-based studies that used either primary or secondary data. We will include studies that had mainly reported prevalence of unsafe abortion and its associated factors. However, as far as our primary aim is to determine the prevalence of unsafe abortion, studies that reported only prevalence of unsafe abortion but not associated factors will be included. In addition, studies that at least had test statistics that measured association between predictor variables with unsafe abortion will be considered to identify the associated factors. The studies should have a crosstab showing difference in prevalence of unsafe abortion in the categories of the exposure variables. We will exclude studies that only investigated unsafe abortion with qualitative approach. If we come across studies that have both quantitative and qualitative study finding, we will only consider the quantitative findings.
Selection of studies
We will export all retrieved studies into the Endnote citation management software [20]. Initially, duplicated studies will be removed from the citation manger. The two authors (MG and AS) will independently screen the studies based on information contained in the titles and abstract based on the inclusion criteria. Studies that clearly mentioned unsafe abortion among reproductive aged women will be selected for the next step of evaluation. Consequently, studies that have been eligible based on their title and abstract will be further screened by GT and TU. Based on title and abstract assessment, the studies will be classified as included, excluded, and undecided studies. For studies that will be categorized as included and undecided, we will further examine and evaluate full texts of the studies for eligibility. The full-text screening will be carried out by GT and AS. Studies that will not be eligible based on the full-text assessment will be excluded and reasons will be described for their exclusion. Studies that will pass through this selection process will be included in qualitative and quantitative synthesis. During screening of the studies, any disagreement among reviewers will be resolved by discussion and reach common understanding. The study selection process flow diagram is adapted from PRISMA guideline [16] (see Additional file 3).
Quality assessment
Studies will be critically evaluated for their validity of the findings. To determine the methodological robustness and validity of the findings of the studies, we will use the JBI (Joanna Briggs Institute) tool for assessing the quality of evidence. Particular attention will be given to clear statement of the objective of the study, sampling techniques, precision of measurement of outcomes of interest and exposure variables, as well as documentation of sources of bias or confounding. The two review authors (GT and AS) will check the scientific quality of the studies independently using quality assessment tool mentioned above. In case of uncertainties, it will be resolved by joint discussion between them.
Data extraction
Data extraction template will be constructed on Microsoft Excel (2013). The two authors (MG and TU) will extract data systematically and stored using data extraction form. Piloting of the data extraction form will be carried out before the beginning of the actual data extraction. Study description tables will be used to record the type of study design, aim, sample size, primary outcomes of interest (prevalence of unsafe abortion), and secondary outcome (associated factors). Numerical data (frequency) will also be extracted and recorded in Microsoft Excel sheet. The systematic review and meta-analysis working group will contact authors of the studies to request for details through email in case of missing data, incomplete report, or any uncertainties.
Data synthesis and statistical analysis
The data will first be presented using narrative synthesis of the included studies. A summary table will be prepared to describe characteristics (author-date, country, design, aim, sampling method, sample size, response rate, and key findings) of the included studies. The presence of statistical heterogeneity will be checked by using the Cochran Q test. The level of heterogeneity among the studies will be quantified using the I2 statistics where substantial heterogeneity will be assumed if the I2 value is ≥ 60%. We will also check the presence of publication bias using funnel plot if more than ten studies are included. We will also do Egger’s and Beggar’s test to check publication bias [21]. To pool prevalence of unsafe abortion, we will conduct meta-analysis using Comprehensive Meta-analysis software [22]. We will use the random effects model and the raw numerical data (number of unsafe abortions (n) and total sample size (N)) from each study. We hypothesize that the legal and illegal status of abortion influences the magnitude of unsafe abortion. Therefore, we will conduct sub-group analysis of the prevalence of unsafe abortion based on countries abortion legal status. Moreover, we will use adjusted, and if none available unadjusted, odds ratios to assess the association between risk factors and unsafe abortion.