The review protocol has been reported using Preferred Reporting Items for Systematic review and Meta-Analysis Protocols (PRISMA-P) guidelines(41) (Additional file 1). It was registered in the International Prospective Register of Systematic Reviews (PROSPERO) database CRD42019126927.
Eligibility criteria
Types of studies
This review will consider quasi-experimental studies which aim to estimate the causal effect of a change in a specific law or reform and an outcome, but in which participants (in this case jurisdictions, whether countries, states/provinces, or smaller units) are not randomly assigned to treatment conditions(42). Eligible designs include the following:
- Pretest-posttest designs where the outcome is compared before and after the reform, as well as nonequivalent groups designs, such as pretest-posttest design that includes a comparison group, also known as a controlled before and after (CBA) designs.
- Interrupted time series (ITS) designs where the trend of an outcome after an abortion law reform is compared to a counterfactual (i.e., trends in the outcome in the post-intervention period had the jurisdiction not enacted the reform) based on the pre-intervention trends and/or a control group(43, 44).
- Differences-in-differences (DD) designs, which compare the before vs. after change in an outcome in jurisdictions that experienced an abortion law reform to the corresponding change in the places that did not experience such a change, under the assumption of parallel trends(45, 46).
- Synthetic controls (SC) approaches, which use a weighted combination of control units that did not experience the intervention, selected to match the treated unit in its pre-intervention outcome trend, to proxy the counterfactual scenario(47, 48).
- Regression discontinuity (RD) designs, which in the case of eligibility for abortion services being determined by the value of a continuous random variable, such as age or income, would compare the distributions of post-intervention outcomes for those just above and below the threshold(49).
There is heterogeneity in the terminology and definitions used to describe quasi-experimental designs, but we will do our best to categorize studies into the above groups based on their designs, identification strategies, and assumptions. Our focus is on quasi-experimental research because we are interested in evidence from studies with a design that permits inference regarding the causal effect of abortion legislation, which is not possible from other types of observational designs such as cross-sectional studies, cohort studies or case-control studies that lack an identification strategy for addressing sources of unmeasured confounding (e.g., secular trends in outcomes). We are not excluding randomized studies such as randomized controlled trials, cluster randomized trials, or stepped-wedge cluster-randomized trials; however, we do not expect to identify any relevant randomized studies given that abortion policy is unlikely to be randomly assigned. Since our objective is to provide a summary of empirical studies reporting primary research, reviews/meta-analyses, qualitative studies, editorials, letters, book reviews, correspondence, and case reports/studies will also be excluded.
Population
Our population of interest includes women of reproductive age (15–49 years) residing in LMICs, as the policy exposure of interest applies primarily to women who have a demand for sexual and reproductive health services including abortion.
Intervention
The intervention in this study refers to a change in abortion law or policy, either from a restrictive policy to a non-restrictive or less restrictive one, or vice versa. This can, for example, include a change from abortion prohibition in all circumstances to abortion permissible in other circumstances, such as to save the woman’s life, to preserve the woman’s health, in cases of rape, incest, fetal impairment, for economic or social reasons, or on request with no requirement for justification. It can also include the abolition of existing abortion policies or the introduction of new policies including those occurring outside the penal code, which also have legal standing, such as:
- national constitutions;
- supreme court decisions, as well as higher court decisions;
- customary or religious law, such as interpretations of Muslim law;
- medical ethical codes; and
- regulatory standards and guidelines governing the provision of abortion.
We will also consider national and sub-national reforms, although we anticipate that most reforms will operate at the national level.
Comparator
The comparison group represents the counterfactual scenario, specifically the level and/or trend of a particular post-intervention outcome in the treated jurisdiction that experienced an abortion law reform had it, counter to the fact, not experienced this specific intervention. Comparison groups will vary depending on the type of quasi-experimental design. These may include outcome trends after abortion reform in the same country, as in the case of an interrupted time series design without a control group, or corresponding trends in countries that did not experience a change in abortion law, as in the case of the difference-in-differences design.
Outcome measures
Primary outcomes
- Access to abortion services : There is no consensus on how to measure access but we will use the following indicators, based on the relevant literature(50): (1) the availability of trained staff to provide care, (2) facilities are geographically accessible such as distance to providers, (3) essential equipment, supplies and medications, (4) services provided regardless of woman’s ability to pay, (5) all aspects of abortion care are explained to women, (6) whether staff offer respectful care, (7) if staff work to ensure privacy, (8) if high-quality, supportive counselling is provided, (9) if services are offered in a timely manner, and (10) if women have the opportunity to express concerns, ask questions, and receive answers.
- Use of abortion services refers to induced pregnancy termination, including medication abortion and number of women treated for abortion-related complications.
Secondary outcomes
- Current use of any method of contraception refers to women of reproductive age currently using any method contraceptive method.
- Future use of contraception refers to women of reproductive age who are not currently using contraception but intend to do so in the future.
- Demand for family planning refers to women of reproductive age who are currently using, or whose sexual partner is currently using, at least one contraceptive method.
- Unmet need for family planning refers to women of reproductive age who want to stop or delay childbearing but are not using any method of contraception.
- Fertility rate refers to the average number of children born to women of childbearing age.
- Neonatal morbidity and mortality refer to disability or death of newborn babies within the first 28 days of life.
- Maternal morbidity and mortality refer to disability or death due to complications from pregnancy or childbirth.
Timing
There will be no language, date or year restrictions on studies included in this systematic review.
Setting
Studies have to be conducted in a low- and middle-income country. We will use the country classification specified in the World Bank Data Catalogue to identify LMICs (Additional file 2).
Search methods
We will perform searches for eligible peer-reviewed studies in the following electronic databases.
- Ovid MEDLINE(R) (from 1946 to present)
- Embase Classic+Embase on OvidSP (from 1947 to present)
- CINAHL (1973 to present); and
- Web of Science (1900 to present)
The reference list of included studies will be hand searched for additional potentially relevant citations. Additionally, a gray literature search will be done with the help of Google and Social Science Research Network (SSRN).
Search strategy
A search strategy, based on the eligibility criteria and combining subject indexing terms (i.e. MeSH) and free-text search terms in the title and abstract fields, will be developed for each electronic database. The search strategy will combine terms related to the interventions of interest (i.e., abortion law/policy), etiology (i.e., impact/effect), and context (i.e., LMICs), and will be developed with the help of a subject matter librarian. We opted not to specify outcomes in the search strategy in order to maximize the sensitivity of our search. See Additional file 3 for a draft of our search strategy.
Data collection and analysis
Data management
Search results from all databases will be imported into Endnote reference manager software (Version X9, Clarivate Analytics) where duplicate records will be identified and excluded using a systematic, rigorous and reproducible method that utilizes a sequential combination of fields including author, year, title, journal and pages. Rayyan systematic review software will be used to manage records throughout the review(51)
Selection process
Two review authors will screen titles and abstracts and apply the eligibility criteria to select studies for full-text review. Reference lists of any relevant articles identified will be screened to ensure no primary research studies are missed. Studies in a language different from English will be translated by collaborators who are fluent in the particular language. If no such expertise is identified, we will use Google Translate (52). Full text versions of potentially relevant articles will be retrieved and assessed for inclusion based on study eligibility criteria. Discrepancies will be resolved by consensus or will involve a third reviewer as an arbitrator. The selection of studies, as well as reasons for exclusions of potentially eligible studies, will be described using a PRISMA flow chart.
Data extraction
Data extraction will be independently undertaken by two authors. At the conclusion of data extraction, these two authors will meet with the third author to resolve any discrepancies. A piloted standardized extraction form will be used to extract the following information: authors, date of publication, country of study, aim of study, policy reform year, type of policy reform, data source (surveys, medical records), years compared (before and after the reform), comparators (over time or between groups), participant characteristics (age, socioeconomic status), primary and secondary outcomes, evaluation design, methods used for statistical analysis (regression), estimates reported (means, rates, proportion), information to assess risk of bias (sensitivity analyses), sources of funding and any potential conflicts of interest.
Risk of Bias and Quality assessment
Two independent reviewers with content and methodological expertise in methods for policy evaluation will assess the methodological quality of included studies using the quasi-experimental study designs series risk of bias checklist(53). This checklist provides a list of criteria for grading the quality of quasi-experimental studies that relate directly to the intrinsic strength of the studies in inferring causality. These include: 1) relevant comparison; 2) number of times outcome assessments were available; 3) intervention effect estimated by changes over time for the same or different groups; 4) control of confounding; 5) how groups of individuals or clusters were formed (time or location differences); and 6)assessment of outcome variables. Each of the following domains will be assigned a ‘yes’, ‘no’ or ‘possibly’ bias classification. Any discrepancies will be resolved by consensus or a third reviewer with expertise in review methodology if required.
Confidence in cumulative evidence
The strength of the body of evidence will be assessed using the Grades of Recommendation, Assessment, Development and Evaluation (GRADE) system(54).
Data Synthesis
We anticipate that risk of bias and heterogeneity in the studies included may preclude the use of meta-analyses to describe pooled effects. This may necessitate the presentation of our main findings through a narrative description. We will synthesize the findings from the included articles according to the following key headings:
- Information on the differential aspects of the abortion policy reforms.
- Information on the types of study design used to assess the impact of policy reforms.
- Information on main effects of abortion law reforms on primary and secondary outcomes of interest.
- Information on heterogeneity in the results that might be due to differences in study designs, individual-level characteristics, and contextual factors.
Potential Meta-analysis
If outcomes are reported consistently across studies, we will construct forest plots and synthesize effect estimates using meta-analysis. Statistical heterogeneity will be assessed using the I2 test where I2 values over 50% indicate moderate to high heterogeneity(55). If studies are sufficiently homogenous, we will use fixed effects. However, if there is evidence of heterogeneity, a random effects model will be adopted. Summary measures, including risk ratios or differences or prevalence ratios or differences will be calculated, along with 95% confidence intervals (CI).
Analysis of subgroups
If there are sufficient number of included studies, we will perform sub-group analyses according to type of policy reform, geographical location and type of participant characteristics such as age groups, socioeconomic status, urban/rural status, education or marital status to examine the evidence for heterogeneous effects of abortion laws.
Sensitivity analysis
Sensitivity analyses will be conducted if there are major differences in quality of the included articles to explore the influence of risk of bias on effect estimates.
Meta-biases
If available, studies will be compared to protocols and registers to identify potential reporting bias within studies. If appropriate and there are a sufficient number of studies included, funnel plots will be generated to determine potential publication bias.