Identified studies and risk of bias
Search results and study selection are summarised in a PRISMA flow diagram (see Fig. 3).
Seventy-seven studies were eligible to be included in the in-depth review. These comprised 12 RCTs [9,38,39,40,41,42,43,44,45,, 19, 37–46] (one with an economic evaluation) [10], two quasi-RCTs [19, 47] and 10 qualitative studies [20,49,50,51,52,53,54,55,56,, 48–57]. The study characteristics have been summarised in Additional file 1: Section 3 and Section 4. There were also 53 other quantitative studies, but since they tended to be of poor quality and their findings were relatively inconclusive, they made no useful contribution to the review. We analysed 10 individually randomised trials in the primary meta-analysis [9,40,41,42,43,44,45,, 10, 19, 20, 37, 39–46]. Of the remaining trials, one was a cluster randomised trial [47], one was an analysis of smaller unpublished randomised and quasi-randomised trials and had an uncertain risk of bias [58] and two studies [59, 60] were non-randomised trials. These four studies were included in the sensitivity analysis. The quality assessments of the included quantitative and qualitative studies using these tools are presented in Additional file 1: Section 5, Section 6 and Section 7.
Quality of evidence
Of the trials, five studies had a high risk of bias [9, 11, 38,39,40, 46], three studies had a low risk of bias [19, 20, 37, 43] and four studies [41, 42, 44, 45] had an unknown risk of bias. We additionally applied the GRADE approach [32] to judge the quality of the overall evidence for each outcome (Additional file 1: Section 6), and rated the outcomes for each intervention as moderate quality.
Using the MMAT tool [25] and considering the extent to which findings were supported by extracts from the original data (i.e. “thickness” and “richness”), we judged the qualitative evidence to be of moderate to high quality. Applying the CERQual approach (Additional file 1: Section 7), based on the methodological limitations of the individual studies and the coherence of each finding [61], our confidence in the certainty of findings from the qualitative synthesis was high (18 findings) to moderate (4 findings) with three findings of low certainty because they were only found in one study and either lacked supporting data or the finding itself was equivocal.
The results are described in the following sections, using the two main types of interventions from the trials, and juxtaposing the contextual detail for these interventions from the qualitative studies, with an explanation of why these interventions work (or don’t work), using realist principles.
Interventions
The trial interventions fell into two broad categories: multi-element psychosocial interventions and a contraceptive programme.
Psychosocial interventions
The psychosocial programmes offered diverse services, such as case management and referral; education about pregnancy, labour and delivery, contraception and infant health; child developmental training; contact facilitation with the health-care system; and individual counselling. Most of these programmes involved home visits [19, 20], two were community based [40, 46] and one involved telephone counselling [42]. Follow-up periods also ranged from 12 months [42] to 24 months [19].
Home-based interventions
The interventions based on home visits had counsellors [19, 20], mentors [38, 46], midwives [44], nurses [43] or trained home visitors [19, 20] delivering the interventions to young mothers at their homes. These professionals [43] and paraprofessionals [45] could be state-sponsored, recruited from the community [19, 20] or from the same ethnic group [42]. All six trials [19, 20, 38, 39, 43, 45, 46] of home-based psychosocial interventions reported on the effectiveness of the intervention in reducing the proportion of repeat pregnancies. The combined event rate was 132 of 308 for the intervention arm versus 140 of 289 for the control arm, giving a non-significant risk ratio (RR) of 0.92 [95% confidence interval (CI) 0.78–1.08]. None of the individual studies showed a significant effect (Fig. 4). However, when four larger, but lower quality studies [47, 58, 59] were included in the sensitivity analysis of the primary outcome (unintended repeat pregnancy), the estimate approached but did not reach statistical significance: event rates of 288 of 1077 (27%) in the intervention arm and 297 of 1004 (30%) in the control arm, giving an RR of 0.88 (95% CI 0.78–1.00). (See Additional file 1: Section 10.)
We advanced the programme theory that repeat home visits facilitate access to services, address gaps in social support networks and sustain behaviour change by repeated contact with young mothers, thereby directing them away from repeat pregnancy.
In a feedback session, young mothers stated a preference for home visits, since this approach allowed them to express their individual needs. Nonetheless, to increase the likelihood of this intervention working, realist theory suggests that the staff conducting home visits should have specialist training. Two of the biggest concerns of the health-care professionals in our consultation group were (1) the inconsistent knowledge base of the health-care professionals who provide advice on contraception for young adolescents and (2) the absence of life-skills training, making young mothers more susceptible to repeat pregnancies. The qualitative studies could shed no further light here since none was undertaken in the context of an intervention.
Community-based interventions
Two trials of interventions based in the community, one of which involved a scheduled peer-centred prenatal care programme [40], and the other monetary incentives promoting mentor-led peer-support group participation [46], reported on their effectiveness in reducing repeat pregnancies. The combined event rate was 42 of 153 for the intervention arm versus 26 of 114 for the control arm, giving a non-significant risk ratio (RR) of 1.00 [0.65, 1.52) in favour of the intervention (Fig. 5).
These interventions did not reduce repeat conceptions in young women (GRADE rating of moderate), and there were no qualitative studies to support their approach. However, in feedback sessions, health professionals stated that transport to and from the location and the availability of food, refreshments and crèche facilities could all increase engagement and improve attendance rates. They also stated that using a ‘buddy system’ or peer support group could offer choices that empower young women and give them confidence, as well as giving them the opportunity to state what they want and need. The adolescent mothers in the service user group appreciated being part of a peer group.
The primary outcome, acceptability of the intervention (inferred as a proxy measure), from one study [40] showed significant differences between arms, but on combining the results with the other trial [46], there were no overall significant differences between arms.
Telephone-based interventions
One study reported a telephone-based mentoring intervention delivered by young female counsellors of similar ethnic backgrounds as the young women involved [42]. The event rate for the effectiveness was 39 of 167 for the intervention arm versus 17 out of 65 for the control arm. This gives a non-significant RR of 0.89 (0.55, 1.46) in favour of the intervention.
Contraceptive programme
The contraceptive programme offered education and advance provision of emergency contraception by a licensed health professional. The contraception intervention study showed a reduction in the number of repeat pregnancies in the intervention group (10 of 48) compared with the control group (14 of 43), giving an RR of 0.69 (95% CI 0.34–1.14); however, this was not statistically significant. Supplying emergency contraception is aimed at reducing repeat pregnancies by addressing frequent discontinuation or switching of contraceptive methods. Although not related to a specific intervention, the qualitative studies revealed some reasons why young women failed to maintain effective contraceptive use. Many women experienced side-effects with the more reliable methods. Women commonly stopped using one method before obtaining another, which rendered them vulnerable to unwanted pregnancy in the interim [49, 50, 56]. These women lacked basic knowledge about contraceptive methods [48, 55, 57]. There were common misconceptions, particularly about fertility soon after birth or when breastfeeding [49, 57] and about the side-effects of some types of contraception. Women also encountered significant barriers to accessing contraception, including restrictive clinic hours, patchy service provision and other system failures, such as lack of provider training [49, 50, 52, 56, 57].
Health-care professionals in our stakeholder group emphasised the disadvantages of using emergency contraceptives as the sole method of contraception. They also acknowledged that since long acting reversible contraceptives (LARCs) are not easily accessible through general practitioners, repeated appointments have to be made, which increases the susceptibility to repeat pregnancy. The service user group stated a preference for LARCs as they provided cover for a long period of time. They also highlighted the challenge presented by the 72-h requirement for emergency contraception, with bank holiday weekends or the prolonged Christmas break being a cause for concern. The user group told us women were also hesitant in asking for emergency contraception for fear of being judged.
The realist synthesis allowed the review team to identify contextual features of the included interventions and their underlying mechanisms such as connectedness and tailoring (Additional file 1: Section 9). The mechanism of connectedness through peer or mentor support may trigger self-determination and active control. Feeling connected and supported can help an adolescent feel that her life choices are being encouraged and that she is being heard. The mechanism of tailoring is evident through situating the intervention within a broad context, taking account of the adolescent’s life experiences, developmental stage, culture and experiences (including pregnancy). The review highlighted contraceptive methods and preferences, barriers and facilitators as ways to implement tailoring. An individual, holistic approach to care may be more successful than adopting a purely medical model of providing information and then encouraging the use of hormonal or long acting reversible methods. It is important to assess an adolescent’s knowledge of contraceptive methods and her individual preferences and needs. Furthermore, adolescents’ circumstances, including transport challenges and difficulties accessing services, needed consideration and a tailored approach. Facilitators such as home visits and school-based services could minimise travel and promote access. Incentives such as crèche facilities or transport could increase engagement with an intervention and attendance rates. The mechanisms uncovered could increase the likelihood of an intervention being effective in preventing rapid repeat pregnancy in adolescents, which has been further explored in a recently published paper [36].
Cost-effectiveness
Only one economic evaluation, a cost-effectiveness analysis [10], was found. However, as the intervention associated with this cost-effectiveness analysis showed no effect, we cannot make definitive conclusions about the economic evidence relating to interventions designed to prevent repeat adolescent pregnancy.
Risk factors for repeat pregnancy
Of the 53 quantitative studies which examined risk factors, most tended to be of poor quality, and their findings were relatively inconclusive. They examined and demonstrated no empirical evidence for the association between repeat unintended pregnancy and factors, such as age, education, history of abuse, smoking, living with the father of the children or the use of oral contraceptives or LARCs, beyond the risk factors present for first conception. However, we deduced from the qualitative evidence that risk factors and reasons for repeat unintended pregnancy appeared diverse and included:
Contextual factors, such as lack of family or peer support, educational or vocational opportunities, and chaotic lifestyles [48, 51, 53],
“…we just went and did it” … “it was a spur of the moment thing… we were partying.” (Herrman 2006, USA, teen mothers recruited from social service agencies)
Emotional factors, particularly to fill an emotional void after an abortion or adoption [49, 52],
“I was just devastated carrying a baby for nine months and feeling it move, going through labour and everything and seeing him for the first time and him just going. It was horrible. He went to foster carers within days after birth.” (Clarke 2010, UK and Caribbean, adolescents with two or more pregnancies in London)
Practical factors, such as the desire to complete one’s family whilst still young [48, 49, 53, 62],
For example, one teen said, “My baby needs a brother or sister—it is too sad to see him growing up without someone to play with.” Another mother stated, “Now that I've had one, I should just finish it, you know, before going back to school and dropping out all over again.” (Bull 1998, USA, teen mothers who received state food aid, and their mothers/guardians)
Motivational factors, such as personal goals and aspirations prompted young women to attempt to avoid a repeat pregnancy, but they were often not given the appropriate support to achieve their goals [49, 52, 63],
“Creche facilities to allow you to go and finish your education and go out and get a job, then you are off the social. Why don't they do things like that?” (Clarke 2010, UK and Caribbean, adolescents with two or more pregnancies in London)