Before 1999, UK government policy on teenage pregnancy has been described as being based more on moral judgements than on evidence.¹ However, in June 1999 the government presented its report on teenage pregnancy² based on the most comprehensive review to date of evidence on risk factors and recommending the development of a national strategy. The report has been regarded nationally and internationally as constituting a turning point in a move towards a more evidence-based approach to the prevention of unintended teenage pregnancy.³

However, much of the evidence available at the time of the report was from cross-sectional studies, which do not allow as clear an examination of causality as longitudinal designs. Although the cross-sectional papers cited acknowledged this fact, this was not indicated in the report itself. In addition, the report relied on evidence that was in some cases quite old and as such, because of the inevitably contextual nature of social research, may not have provided as clear a guide to current social trends as more recent research would have done.

Table 1^{4,5,6,7,8,9,10,11} summarises in what aspects the 1999 report relied on cross-sectional studies and on data that were more than a decade old at the time of publication of the report. Although any cut-off is of course arbitrary, we regard a decade as a reasonable guide, given likely intergenerational discontinuities. Our paper explores whether the report would have drawn attention to the same risk factors had it been able to incorporate evidence from a secondary analysis of more recent, longitudinal data from the Randomised Intervention of Pupil Peer-Led sex Education (RIPPLE) study, an ongoing cluster-randomised trial of sex education in England.¹² As well as providing longitudinal and more recent information on risk factors for teenage pregnancy among young women, the data allow an examination of the risk factors for pregnancy among the female partners of young men, a matter not examined in the 1999 report. Our analysis focuses on conceptions by age 16 years, whereas the 1999 report examined those by age 18 years. However, given health concerns are greatest regarding the 16-year-olds,¹³ this could be regarded as a further advantage of our analysis.

In undertaking this analysis, we are not assuming that teenage pregnancy is a negative outcome. As suggested earlier, although research indicates that pregnancies in the early teenage years may be associated with greater perinatal risk, this does not seem to be so with regard to births among older teenagers.¹³ We would also contend that the social harms associated with teenage births are not inevitable but are determined by societal reaction, in particular the support offered to young mothers and their children.¹⁴

METHODS

Data collection

In all, 27 coeducational secondary schools in central and southern England were recruited to the RIPPLE trial in 1997. The initial data were collected from two successive cohorts of students in 1997 and 1998 when pupils were aged 13 or 14 years. The first follow-up occurred when the participants were aged 14 or 15 years and the second, when most participants were aged 15 or 16 years, in 2001. Intervention occurred at age 13 or 14 years (in 1998 and 1999). Of the 9508 eligible pupils, 8766 (92.2%) completed a questionnaire at baseline, 7770 (81.7%) did so at first follow-up, and 6656 (70%) at second follow-up. Consent was obtained from all parents or carers for children’s participation in the study. The study was approved by the committee on the ethics of human research at University College London, London, UK. The research is ongoing; findings from phase one of the trial (outcomes at age 16 years) have been published.¹⁵

Data analysis

The putative risk factors for teenage pregnancy identified in the UK government’s 1999 report on teenage pregnancy were grouped into five categories relating to: socioeconomic status; family structure or relationships; expectations of the future and perceptions of peers; other risk behaviours; and sexual health knowledge (table 1). We identified comparable variables in the RIPPLE study (tables 1 and 2).

Our analysis considered the outcome “ever pregnant” (n = 83) and “reported partner’s pregnancy” (n = 69) by second follow-up (age 15/16 years). All conceptions occurred between baseline and second follow-up, the question being asked only at second follow-up. The age of the female partners of young men was not asked. Binary measures were created for most of the risk factors, however where creation of binary measures from categorical ones would have led to loss of coherence of categories, categorical measures were retained (table 2). We used simple logistic models with all young people included in the analysis to derive crude and adjusted odds ratios (ORs) for associations between our risk factors and pregnancy at or before age 16 years. Adjusted ORs were calculated by including all variables from the other four categories of risk factor in the regression model; these being pre-hypothesised as potential confounders, risk factors from the same category were not included. This was done to minimise collinearity. As it transpired, standard errors in the model were not large and therefore collinearity was unlikely to be a problem. Within-school correlation was accounted for by using the generalised estimating methodology of Liang and Zeger¹⁶ with an exchangeable correlation structure and robust standard errors.

Where risk factors were measured both at baseline and at first follow-up, both measures were included in the analysis. In such cases, calculation of explicit p values was restricted to overall tests of the association between a risk factor (at baseline and the first follow-up combined) and outcome. In addition, interpretation of the effects of the risk factors will be restricted to the overall effect of both the measure at baseline and that at follow-up 1. This is because different ORs for the same measure at the different time points simply serve as a guide as to which measure is dominant when the question has been answered at both time points.

Results are reported separately for boys and girls. Data from both trial arms were analysed together. Interactions between risk factors and trial arm were examined.

RESULTS

Tables 3 and 4 show the prevalence of risk factors and their associations with pregnancy at or before age 16 years. Some ORs differ from earlier publications owing to differences in exclusions. The following factors were significantly associated with increased risk of pregnancy at or before age 16 years in young women: non-privately owned housing; lack of expectation of being in education at age 20 years; expectation of being a parent by age 20 years; belief that over half one’s peers are sexually active; intention to skip school; being drunk monthly or more; and ability to identify a sexual health service. Being born to a teenage mother was associated with pregnancy at borderline significance. Lack of expectation of being in education at age 20 years, belief that over half one’s peers are sexually active and being drunk monthly or more at age 13 years remained significant on adjustment for factors from other risk categories. Young women who could communicate easily with their mother or female guardian were considerably less likely to report pregnancy at or before age 16 years. This remained significant on adjustment for factors in other categories.

The following were associated with increased risk of partner’s pregnancy among young men: expectation of being a parent by age 20 years; belief that over half of one’s peers are sexually active; intention to skip school; being drunk monthly or more. Being born to a teenage mother was associated with initiating pregnancy at borderline significance. Of these, only expectation of being a parent by age 20 years did not remain significant on adjustment for factors in other categories. On adjustment for factors in other categories, inability to identify a sexual health service was significantly associated with subsequent increased risk of a partner’s pregnancy for young men.

Both young women and young men who cited school as a source of information on sex were significantly less likely to become pregnant or have a pregnant partner at or before age 16 years. However, this association did not remain significant on adjustment. No significant differences between trial arms were found.

DISCUSSION

Our results provide support for some but not all the conclusions of the UK government’s 1999 report on teenage pregnancy regarding risk factors for teenage pregnancy among girls and provides new insights into risk factors for pregnancies among the partners of teenage boys. Factors found to be significantly associated among girls and boys with teenage pregnancy were peer norms, intended truancy, expectation of teenage parenthood and being drunk monthly or more, as well as being born to a teenage mother at borderline significance. Among girls, socioeconomic disadvantage and low educational expectations were also significantly associated with pregnancy. Although our results suggest that low knowledge of sexual health, as measured in our questionnaire, may not be a risk factor, they do support the conclusions of the 1999 report that gaining information about sex from school is associated with reduced risk of teenage pregnancy among girls and the sexual partners of boys.

We also found evidence that girls’ ease of communication with mothers or female guardians about private and personal matters in their early teens may be protective against pregnancy by age 16 years, but that no such protection is afforded by similar communication among boys or regarding girls’ or boys’ communication with fathers. This contrasts slightly with the findings of previous studies.^6,7 Although our question focused on personal or private matters in general rather than on sex in particular, we believe our question was a valid way of exploring communication about sex among this age group.

The strength of some of the above associations is striking. This is the case, for example, for associations of pregnancies among young women and partners of young men with drunkenness and with intention to skip school. This is in line with existing evidence on the clustering of risky behaviours.¹⁷ Such associations were suggested, but not fully evidenced, in the UK government’s 1999 report on teenage pregnancy. In addition, our finding that the belief that half or more of one’s peers are sexually active was significantly associated with increased risk of pregnancy for young women and the partners of young men provides support for the component of the teenage pregnancy strategy aiming to dispel myths about sex and give realistic information about numbers of teenagers having sex.

Another interesting finding is that participants’ ability to identify a sexual health service is associated with increased risk of pregnancy in young women, but with decreased risk among partners of young men on adjustment for other factors. Reverse causality among young women may explain these differing associations, whereby young women are more likely than young men to learn of the location of a service because of earlier risk-taking. However, it is unlikely that reverse causality would have involved girls’ conceptions leading them to have higher knowledge about services.

Some risk factors identified as significantly associated with pregnancy in unadjusted analysis, particularly among girls, do not remain significant after adjustment for other categories of risk. This is unsurprising given that these factors are likely to intercorrelate so that adjustment for each will diminish the associations each has with pregnancy.

Our findings are subject to several limitations. Our study did not explore the effects of low expectations regarding training and occupation, which were identified in the UK government’s 1999 report on teenage pregnancy as increasing the risk of teenage pregnancy. Our data are observational and so strongly suggest but cannot prove causality. In a very small number of cases, pregnancy may have occurred after baseline but before measurement of risk factors at the first follow-up. Although our data allowed an analysis of pregnancies among the partners of teenage boys, this outcome may have been reported less accurately than female participants’ pregnancies, and these female partners may not themselves have been teenagers.

Attrition among study participants, although low overall, may have been somewhat higher among those conceiving. This is unlikely to have been differential with regard to our risk factors, and so, although it may have produced a slight underestimation of some associations, it will not have biased our findings. As mentioned earlier, our analysis focuses on conceptions by age 16 years, whereas the 1999 report examined those by age 18 years. This is, however, unlikely to explain our differing findings regarding the effects of knowledge on risk of teenage pregnancy, and current evidence suggests that our outcome is a better guide to the health of mother and baby.¹³ As in studies referenced in the UK government review, our analysis only examined those children whose mothers were teenagers at the time of their birth, as opposed to including children whose mothers were teenagers at the time of the birth of an older sibling but not at the time of that child’s birth. Our estimate of the effect of the former may be smaller than an estimate derived from examination of the latter. Interestingly, our results are supported by a recent analysis of UK girls born in 1970, which found an association between being born to a teenage mother and girls’ own teenage pregnancies, after adjustment for socioeconomic status.¹⁸ Finally, our finding that young people’s expectations concerning parenthood do not correspond to their actual experiences is interesting but does not indicate that the former are invalid measures; rather they are valid indicators of perspectives at a point in time.

In terms of policy, our findings provide broad support for the multi-faceted approach to teenage pregnancy recommended in the 1999 report. We would suggest that socioeconomic disadvantage, low educational expectations, truancy and alcohol use are important influences on teenage pregnancy as well as other negative outcomes.¹⁷ It is noteworthy that these factors are now being dealt with in the UK by government initiatives focused on young people’s personal development and education.¹⁹ Our findings provide evidence to support the implementation of policies aiming to improve communication between young women and their parents or guardians about sex, such as the “Time to Talk” initiative. Our findings that school is a key source of sexual health information but that none of our five indicators of sexual health knowledge is a risk factor for pregnancy support the view that sex education should focus on skills and emotional literacy as well as knowledge, as recommended in the UK government guidance.²⁰ It might also suggest that delivery of sex education should not, in the absence of other action, be regarded as sufficient for reducing teenage pregnancies, again supporting the multifaceted approach taken in the UK.

Finally, we would like to emphasise that the aim of this paper has been to examine the implications of our own data for policy and practice concerning teenage pregnancy. Although beyond the remit of this study, we strongly recommend that policy in this area would benefit from a broader review of recent longitudinal evidence on risk factors using systematic methods of search and study appraisal.