Statistics from Altmetric.com
Despite data from the UK that are consistent with a gradual decline in teenage conception rates, the UK continues to have the highest rate of teenage pregnancies in Western Europe.1 ,2 Teenage pregnancies are a target within the England Teenage Pregnancy Strategy and their equivalents within the devolved governments of the UK.3–6 Teenage pregnancies have considerable impacts on the individual well-being of teenage parents and their children. Inherent within the national strategy responses, and a range of other national policy documents addressing this issue, is the recognition that babies of teenage mothers have increased mortality in their first year and a significantly increased risk of living in poverty, achieving less at school and being unemployed later in life7 with substantial costs to society.
Repeat pregnancies represent a considerable proportion of the overall rate; one-fifth of births among under-18s are repeat pregnancies8 and are thus a crucial potential focus for intervention. Around three quarters of teenage pregnancies are unplanned with up to half resulting in abortion.3 Within the UK, teenage pregnancy is strongly associated with social disadvantage. The social predictors of repeat adolescent pregnancy are varied and have previously been usefully grouped into predictors operating at individual, couple, family, community and social levels.9 These predictors share much common ground with those of first teenage pregnancy.
Aims of the review
The overall aims are to identify and evaluate the effectiveness of interventions for preventing repeat unintended pregnancies among adolescents, and to investigate the barriers and facilitators for their implementation and uptake. While these overall aims are broad, the focus will be on the implementation of interventions; specific research objectives are to determine:
What factors characterise subgroups which are at greater risk of repeat unintended pregnancies (ie, what are the predictors of a repeat unintended pregnancy)?
Which (elements of) interventions appear to be effective, how do they work, in what setting and for whom (conversely, why are they ineffective, why don't they work)?
What are the barriers and facilitators to the acceptability, uptake and implementation of interventions?
What is the relative cost-effectiveness of interventions?
Overall plan of research
Initial scoping searches informed a tailored, four-phase approach to the review (figure 1). For the overall framework of the mixed method review, we will draw on the structured, phased Evidence for Policy and Practice Information (EPPI)-Centre approach10 and use their reviews of young people, pregnancy and social exclusion, and the barriers to and facilitators of children's healthy eating as methodological exemplars.10 ,11 After conducting extensive literature searches, screening the evidence against explicit inclusion and exclusion criteria and appraising study quality, a mapping exercise is undertaken to organise and describe the evidence so as to give a clear picture of the body of research. From this, the scope of the review can be refined in order to focus in depth on the most important areas of the topic in question. Our research will be guided by an expert panel whom we will engage with via a regular agenda item on the quarterly meeting of the Public Health Wales's Addressing Teenage Conceptions in Wales; Task and Finish group, and we will present our findings for feedback and discussion to a group of young mothers who have experience of teenage pregnancy and early parenthood and are in contact with the Flying Start programme. The Task and Finish group draws its membership from public health practitioners, policymakers, general practitioners and specialist doctors, midwives, academics and third sector representatives from across Wales with a professional interest in reducing teen conception rates.
Phase 1: Identifying the literature, quality appraisal and mapping the evidence
We will search the following electronic databases for published literature using strategies that combine thesaurus terms and keywords relating to pregnancy, termination of pregnancy or parenthood with adolescence and text word synonyms for repeat or subsequent:
MEDLINE and MEDLINE in Process, PsycInfo, CINAHL (Cumulative Index to Nursing & Allied Health Literature), the Cochrane Library (Cochrane Database of Systematic Reviews, Database of Abstracts of Reviews of Effects, the Health Technology Assessment (HTA) Database, National Health Service (NHS) Economic Evaluation Database), EMBASE (The Excerpta Medica database), BNI (British Nursing Index), ERIC (Educational Resource Index and Abstracts), SocAbs (Sociological Abstracts), ASSIA (Applied Social Sciences Abstract & Indexes), BiblioMap (The EPPI-Centre register of health promotion and public health research) and the Social Sciences Citation Index. We will use a search strategy developed and piloted in MEDLINE (see online supplementary appendix 1) and subsequently modified for use in the remaining databases (full details available from the authors).
Further modified versions of the same strategy will be used to search the following databases for ‘grey’ literature: OpenGrey, Scopus, Scirus, Social Care Online, National Research Register, NIHR portfolio database, and Index to theses. Similarly, to capture economics studies, we will search RePEc (Research Papers in Economics) and EconLit. We will apply an alternative search strategy specifically designed to capture the type of descriptive titles that are common in qualitative studies to selected databases: SocAbs, ASSIA, BNI and the Social Sciences Citation Index. This strategy will combine additional synonyms for ‘pregnancy’, ‘adolescence’ and ‘repeat’ with a brief qualitative filter comprising three broad free-text terms, ‘qualitative’, ‘findings’ and ‘interviews’, which has been shown to be as effective as a more complex one.12 ,13
Since the issues surrounding implementation of interventions are a primary focus of this review, qualitative and process evaluation evidence associated with trials is highly relevant—particularly to better understand the facilitators and barriers to implementation.14 Therefore, a key feature of our search strategy will be to identify ‘evidence clusters’ related to key randomised controlled trials; we will use search engines such as Google and Google Scholar, conduct citation searches and contact Principal Investigators to ascertain whether a qualitative/views study or process evaluation was conducted, as these data are sometimes published separately or not published at all.
We will be further guided by our expert panel members, some of whom have frequent contact with young women who have experience of teenage pregnancy and early parenthood; they will inform the direction of our searches by identifying interventions which their clients may have experienced and giving us feedback. This will help identify areas where evidence is lacking and supplementary searches may be necessary. Additional sources will include the bibliographies of included papers. If necessary, we will also manually search key journals such as the Journal of Epidemiology and Community Health, the Journal of Adolescent Health, Contraception, the Journal of Adolescence and Health Care, the Journal of Paediatric and Adolescent Gynaecology, Adolescence, Maternal and Child Health Journal and the Journal of Reproductive Medicine and make personal contact with professional networks such as the Family Planning Association, reproductive health nursing networks, General Practitioners’ fora, etc to identify Department of Health and third sector policy documents and evaluations. Finally, we will circulate a list of included studies to key stakeholders and researchers in the field and ask them if they are aware of any important omissions.
A number of systematic reviews related to teenage pregnancy have been published in the past two decades. The earliest we found in our scoping search was dated 1997.15 Therefore, we will limit our searches to 1995 onwards, but we will conduct a separate search excluding the terms for second or subsequent pregnancies but including a filter for systematic reviews as a means of capturing relevant data from earlier studies.
References will be managed by using bibliographic reference management software (Endnote). Two reviewers will independently screen titles and abstracts to identify potentially relevant documents, which will be retrieved and assessed according to the inclusion criteria below. Disagreements will be resolved by discussion or, if necessary, by a third reviewer.
Our search strategy will capture published studies of any design including trials of interventions, effectiveness studies, interrupted time series studies (ITS), cost-effectiveness studies, process evaluations, surveys and qualitative studies of participants’ views and experiences of interventions. We will also consider relevant grey literature of any type, such as unpublished reports, service evaluations and theses.
The population of interest is young women, who are aged up to and including 19 years and have had at least one unintended pregnancy, whether the outcome was termination, miscarriage or delivery. Where study populations are mixed, we will include all studies whose reported population comprised at least 75% young women in our target age group.16
We will include studies of any intervention designed to reduce repeat unintended pregnancies (also referred to as ‘birth-spacing’ or ‘pregnancy-spacing’) in these young women, delivered in any educational, healthcare or community setting. Interventions may have single or multiple components, and could be delivered to individuals or communities. We will also include studies designed to identify risk factors or subgroups at increased risk of repeat unintended pregnancy, where there may be no actual intervention.
We will include studies that identify barriers and facilitators to the implementation and uptake of interventions, and explore the views of intervention recipients or providers or health professionals, particularly with regard to whether the intervention was implemented and worked in the way it was intended. We will look for studies that help us identify programme theories and logic—and we will look for, and develop, candidate theories as to why some young women have more than one unplanned pregnancy, which could begin to explain the relative success or failure of particular interventions.
Comparators could be no intervention, standard practice or another intervention. Comparators are likely to be location-specific and standard care in one setting may be seen as an intervention in another setting.
Outcomes and other phenomena of interest
We will report on the primary and secondary outcomes and other phenomena of interest below. Outcomes will be addressed by a range of evidence types and analytical techniques (see figure 1).
Primary outcomes and other phenomena of interest
Identification of at-risk groups
Identification of barriers and facilitators of interventions relating to:
Feasibility of implementation
Views and experiences of young women, families and professionals
Conceptualisation of repeat teenage pregnancy rates as an issue
Effectiveness of interventions
Uptake of interventions
Change in repeat pregnancy rates
Other outcomes and other phenomena of interest
Identification of, access to and uptake of reproductive services and social care
Feasibility of widespread adoption of interventions in health and social care
Cost of interventions
Effectiveness of interventions:
Change in validated quality-of-life indices
Change in rate of abortion
Cost-effectiveness of interventions
Study exclusion criteria
We will include published studies from any country as, although there may be different cultural and social attitudes towards teenage pregnancy, they may have a bearing on minority populations within the UK. If any major studies (eg, randomised controlled trials or national cohort studies) are found in non-English language publications, we will attempt translation. Smaller qualitative studies or process evaluations published in other languages may be missed, but exclusion of those studies is unlikely to bias any synthesis of evidence. We will limit our searches for unpublished material to the UK to enhance the direct applicability of the results to the NHS and UK public health bodies. We will not exclude studies on the basis of quality, but will incorporate judgements about study quality when interpreting the evidence.
RCTs and quasi-RCTs will be assessed using the Cochrane Collaboration's Risk of Bias (RoB)12; we will categorise and report the overall risk of bias of each of the included trials:
Low risk of bias (plausible bias unlikely to seriously alter the results) if all criteria were met;
Unclear risk of bias (plausible bias that raises some doubt about the results) if one or more criteria were assessed as unclear;
High risk of bias (plausible bias that seriously weakens confidence in the results) if one or more criteria were not met.
The Cochrane RoB tool will be supplemented by the Mixed Methods Appraisal Tool (MMAT) for mixed studies reviews.17 This tool has the advantage of incorporating the appraisal of several different study designs (qualitative, RCT, non-RCT, observational, mixed methods) using a single tool with a coherent range of quality criteria. Piloting suggests that the MMAT is an efficient and reliable tool; however, it does not cover economic evaluations, so we will also use the Drummond checklist18 and the Phillips checklist19 recommended for economic studies and economic models, respectively.20 Where the evidence comprises, for example, policy documents or reports of locally implemented initiatives, it is unlikely that we will be able to formally assess quality. We will therefore take a broad view of the evidence as being most relevant for the context and be guided by the advisory group as to its merit.
Certainty of evidence
We will use GRADE as the accepted approach to assessing the certainty of findings of reviews of effectiveness,21 but since GRADE is not suitable for appraising the certainty of qualitative evidence, we will also use a recently developed method, CerQual (certainty of the quality of evidence).22 The CerQual approach is based on the methodological limitations of individual studies contributing to the review finding, in this case indicated by the MMAT, combined with the coherence of each review finding. Coherence is assessed by the extent to which a clear pattern is identifiable across the individual study data, and is strengthened where findings are consistent across multiple contexts or where variation across individual studies is explained, especially where studies contributing to the finding are drawn from a wide range of settings. CerQual is similar to GRADE in that both approaches aim to assess the certainty of (or confidence in) the evidence and both rate this certainty for each finding across studies rather than for each individual study. GRADE also bases its assessment on a combination of the quality of the evidence and other factors, including consistency across studies.
Mapping the evidence
We will undertake a mapping exercise following the EPPI-Centre methods.10 We will assign codes to records using lists that we will develop of predefined keywords to aid the sorting and grouping of documents for mapping and synthesis. Keywords will be related to, for example, the study design or type of record (intervention study, process evaluation, qualitative study, report, etc), the country and health, educational or community setting in which the study took place, the topic or focus of the study/report, the population focus of the study/report and study design. From the coding exercise, we will develop a descriptive map of the literature and identify gaps in the research. The map will also provide a context for interpreting the results of the synthesis and a basis for refining the scope of the review, that is, it will aid the advisory group in identifying areas for in-depth focus and, possibly, areas where additional searches need to be conducted. The advisory group will meet with the review group late in phase 1 to discuss the results of the mapping exercise and to ratify or recommend refinements to that work which would be needed prior to the completion of phase 2 where the evidence will be selected and prioritised for an in-depth review.
Phase 2: Selecting and prioritising the evidence for in-depth review and data extraction
The review team, with advice from the advisory group obtained during the mapping exercise, will prioritise the evidence for in-depth review. Beginning with evidence relating to primary outcomes, and within each grouped set of data, we will prioritise the best quality evidence of most relevance to address the research questions. We will be particularly interested in ‘evidence clusters’, that is, trials of interventions accompanied by qualitative studies, process evaluations, reports etc; ‘sibling’ studies, that is, trials of interventions paired with qualitative studies using the same participants, and ‘orphan’ studies where effectiveness studies and views studies relate to similar interventions and populations (and so can inform each other), although the actual participants are different. These studies have the potential to indicate the effectiveness of an intervention as well as its acceptability to users and barriers to implementation and uptake. We will apply the CART framework, where evidence is judged on the criteria of Completeness, Accuracy, Relevance and Timeliness.23 Thus, we will prioritise the evidence according to relevance and stop when data-saturation is reached.
Study selection for the Cochrane Review
We have registered the title with the Cochrane Fertility Regulation review group, and will develop a separate Cochrane protocol to be published in the Cochrane library. The main difference between the Cochrane protocol and this HTA protocol is that we will adapt the inclusion and exclusion criteria to exclude unpublished evidence and PhD theses from the Cochrane review. The intention is to follow the same design and methods across the two reviews. This variation in inclusion criteria between the Cochrane and HTA reviews offers an opportunity to undertake a sensitivity analysis to see what contribution unpublished evidence makes to overall outcomes and interpretation of evidence. The Cochrane Collaboration is also calling for exemplar reviews of complex interventions where complexity is considered important. The Cochrane version of the HTA review has the potential to be classified as an ‘innovative’ review in the Cochrane library and would be one of the first to demonstrate the value of an additional realist synthesis in understanding complexity at multiple levels of context, intervention and implementation.
We will create a bespoke set of data extraction forms to collect data from each study: study characteristics (design, sample type, sample size, etc); description of intervention or risk factors; contextual factors in the study setting; outcomes including costs of implementing intervention; programme theories or mechanisms described by the authors in the rationale behind the intervention or postulated in the explanation of the results. We will draw on the components of Greenhalgh et al24 framework to map facilitators and barriers to intervention implementation, including characteristics of the intervention (eg, risk, complexity, acceptability), the contextual elements that influence whether it might be used or not (eg, resources, organisation of services, leadership) and what implementation processes are required to implement the intervention (eg, change agents, project management support and communication). Data extraction will be undertaken by two reviewers independently; disagreements will be resolved by discussion.
We will present the findings of the data extraction exercise in a table of study characteristics. This will include: the study details, setting, population, quality score, methods, etc. We will present sociodemographic characteristics known to be important from an equity perspective. For this process, the PROGRESS (Place, Race, Occupation, Gender, Religion, Education, Socioeconomic status (SES), Social capital) framework will be utilised.25
Phase 3: Evidence synthesis by type of evidence
Data to be considered in this review will be extremely diverse, and data synthesis complex. The choice of synthesis method depends on the questions addressed and the type of data included; in this review, there are many questions and diverse data. Figure 1 illustrates the method of synthesis proposed for each evidence type.
Measures of intervention effect
We will present quantitative continuous outcomes on the original scale as reported in each individual study. These may be standardised, if they use different scales, by dividing the estimated mean difference by its SD. Dichotomous outcome data will be fitted with a random effects model using the Mantel-Haenzel test and presented as risk ratio (RR). ITS studies will be analysed by using the rate change (α) and slope (β) parameters as the intervention effect. Time to event data will be analysed using HRs as the intervention effect. All outcomes data will be reported as their effect size with associated 95% CIs.
Unit of assessment
We anticipate that many of the studies will be clustered by community, service or geographical area. Where possible, cluster RCTs (or quasi-RCTs) that do not account for the correlated nature of the data will be reanalysed by inflating the study standard. If this is not possible, we will report only the point estimate, and adjust the ‘other risk of bias’ to account for the likely impact of not accounting for the clustering.
Multiple time-points per outcome in non-ITS study designs
If RCT, quasi-RCT or cohort studies are found reporting repeated incidence of repeat pregnancy over time, these outcomes will be defined and summarised at both short-term and long-term time. We will follow the Cochrane Effective Practice and Organisation of Care Group (EPOC) and non-randomised studies methods group guidance on incorporating these diverse types of evidence in the review.
Assessment of heterogeneity
We will assess both heterogeneity of the populations, context and interventions (clinical) and statistical heterogeneity. Clinical heterogeneity will be assessed by examining the characteristics of the studies, the similarity between the types of participants and the interventions, while statistical heterogeneity will be assessed using the I2 statistic. We will report heterogeneity as important if it is at least moderate to substantial by I2 >30% and will not pool data if the statistical heterogeneity is severe I2 >90%. If there is statistical heterogeneity between 60% and 90% and this can be explained by clinical reasoning and a coherent argument can be made for combining the studies, they will be entered into a meta-analysis. After exploring the heterogeneity where a coherent scientific argument is not found, the included study causing the heterogeneity will be excluded and the analysis repeated as a sensitivity analysis. Where the heterogeneity cannot be adequately explained, the data will not be pooled in a meta-analysis. In this case, or when only single outcomes are reported, we will present study findings in tables and explore the relationships within and between studies in a narrative summary.26
Assessment of reporting biases
Where there are an adequate number of studies (nominally at least 10), an assessment of reporting bias will be carried out by testing for funnel plot asymmetry.27 Funnel plots will only be presented where there is some evidence of asymmetry in the plots. Possible sources of asymmetry will then be explored with an additional sensitivity analysis and the studies at greatest risk of bias will be removed and the most likely unbiased intervention effect will be summarised in the meta-analyses.
Dealing with missing data
Where we have missing or unclear data or information, we will contact the investigators of the primary research. Following this, correspondence data may be reanalysed according to a treatment by allocation principle whenever possible.28 If loss to follow-up data is not fully reported and authors have conducted a perprotocol analysis, we will inspect the degree of imbalance in the dropout between the groups to determine the potential impact of bias. In the absence of a treatment by allocation population, we will use an available case population. Where possible, an intention to treat analysis will be carried out. In cases where non-ITT analyses are reported, we may report the missing data using a best–worst and worst–best scenario analysis if an adequate number of studies were included in a meta-analysis (nominally three).28 Where there are missing variances, these will be imputed using methods that will be highlighted explicitly in the methods section and the assumptions fully described.
Subgroups included and highlighted within the primary research as important confounders will be used to identify risk factors. If there are adequate studies, we will investigate the following subgroup of studies, or subgroups within studies to try to identify differences in intervention effectiveness: age of mother, deprivation index of area, length of follow-up, history of substance misuse, looked after children (or care leavers).
For qualitative studies or qualitative elements in mixed-method studies, we will develop an a priori coding framework specifically designed to address questions and issues of interest and conduct thematic syntheses using the framework method developed by Ritchie and Spencer.29 We will use Greehalgh et al's24 framework to support the synthesis of findings about the facilitators and barriers to intervention implementation.
Surveys, process evaluations and other types of data
Data from surveys, process evaluations and other sources, for example, reports may be either semiqualitative or quantitative or both. Data may be extracted to present evidence of acceptability or uptake of interventions and synthesised in a narrative summary or aggregated using thematic analysis.
We will select subsets of evidence and apply the principles of realist synthesis.30–32 We will identify explicit or implicit theories by which it is postulated how an intervention has an underlying causal mechanism that works in a defined social context to result in a particular outcome. Such theories may also be used to explain the failure of an intervention to work. Additional theories will be identified from the wider literature (eg, policy documents), the advisory group members or personal contact with other experts in the field. Data synthesis will involve individual reflection and team discussion and will question the integrity of each theory, adjudicate between competing theories, consider the same theory in different settings and compare the stated theory with actual practice.33 Coded data from the studies will then be used to confirm, refute or refine the candidate theories. Thus, we will attempt to explain what interventions work, for whom and in what circumstances.
Investigation of risk factors using meta-regression
A key aim of the review will be to search, identify and summarise the population of young girls who are at greatest risk of repeat pregnancy (eg, considering income, social deprivation, ethnicity, degree of rurality, substance misuse, currently in care or care leavers, those from a vulnerable or in at-risk communities). We anticipate that these factors will be considered and summarised as important confounding variables within RCTs, quasi-RCTs and controlled before and after studies (CBAs) for studies investigating an intervention to reduce repeat pregnancies. We will also identify non-interventional studies that present epidemiological data (eg, cohort studies, cross-sectional studies and policy documents) of effect modifiers associated with an increased risk of repeat pregnancy. Data will be extracted and summarised and, if possible, presented in a meta-regression of summarising the standardised effect sizes, with their associated 95% CIs.
We will provide a narrative review of economic evaluations of interventions specifically designed to address the issue in question. We will stratify any economic studies found by the public health lever mechanism used, for example, government, statutory or legal, public information, school-based group or targeted intervention, NHS initiated, charity initiated. We will be particularly interested in perspective of analysis, type of economic evaluation (eg, cost analysis, cost–benefit analysis, cost-effectiveness analysis, cost-utility analysis). We will document the way that these studies have attempted to overcome the particular methodological challenges of this type of complex, preventative, behaviour change-based intervention.34–37 Where possible, we will conduct a meta-analysis of the economic evidence (if there is sufficient homogeneity to allow it).
Phase 4: Overarching syntheses
Reporting the studies
Results will be presented with reference to the PRISMA reporting guidelines for systematic reviews,38 the MOOSE guidelines for Meta-analyses of observational studies39 and the RAMESES publication standards for realist syntheses.40 We will summarise the evidence for each outcome in a summary of findings table.
Overarching narrative synthesis and interpretation of findings
We will conduct an overarching narrative synthesis, juxtaposing the qualitative and quantitative data on service user views and implementation issues and comparing programme theories with effectiveness.41 To aid this process, we will develop a methodological and conceptual matrix or logic model to integrate our findings.42 ,43 The processes involved in this review are essentially iterative; thus, we anticipate that the mapping exercise and the work on relevant programme theory will contribute towards a framework for the meta-synthesis.
Discussing the evidence
Within the discussion, we will consider the following areas:
Summary of findings, strength and certainty of evidence;
Strengths and limitations of the review;
Implications for UK policy and practice, including intervention implementation;
Implications for further research.
During the draft stage of the final report, a consultation process, defined and facilitated by the frontline organisations engaged with our team, will be undertaken with service users via their representatives in the voluntary sector. This aim of this process is to share our emerging findings and to seek comments and feedback, which will also feed into the discussion section of the report.
Systematic reviews of complex interventions are challenging44; interventions in the field of teenage pregnancy prevention are diverse in terms of content, context and implementation factors. To encompass this diversity, the methods for this review are ambitious and innovative, drawing on a number of methodological sources and with a degree of flexibility built into the design. To ensure the validity and reliability of findings based on novel research methods, our review includes:
An overall framework to provide structure to the review;
A high degree of methodological expertise in the review team;
An expert advisory group including access to stakeholders in the field;
Feedback from clients/service users to authenticate findings.
These factors will ensure that we deliver a high quality review with a high degree of relevance in its findings that will be of interest to policymakers and those who design and commission services, as well as making a valuable contribution to the debate on methods for conducting systematic reviews of complex interventions, where understanding complexity is considered important.
The authors wish to thank Dr Huw Roberts for his invaluable contribution to the study management team and Dr Richa Sinha for her contribution to the development of the search strategies.
Review history and Supplementary material
This web only file has been produced by the BMJ Publishing Group from an electronic file supplied by the author(s) and has not been edited for content.
Files in this Data Supplement:
- Data supplement 1 - Online appendix
Contributors RW and NC initiated the study. MH devised the study structure and all authors contributed to its design. RW, MH and BC drafted the manuscript. AB, MH, DP and JC contributed to the search strategies. BC contributed to the quantitative analysis and JN to the qualitative analysis and mixed methods synthesis. JRM was involved in the realist synthesis, ML, NC and NW in the health care perspective and RTE and JC in the health economics perspective. All the authors read and approved the final manuscript.
Funding This work is funded by the National Institute for Health Research Health Technology Assessment project number 11/141/05. Professor Rhiannon Tudor Edwards and Dr Joanna Charles are funded by Public Health Wales.
Competing interests None.
Provenance and peer review Not commissioned; externally peer reviewed.
If you wish to reuse any or all of this article please use the link below which will take you to the Copyright Clearance Center’s RightsLink service. You will be able to get a quick price and instant permission to reuse the content in many different ways.