Purpose In order to consider the potential contribution of universal versus targeted prevention interventions, the authors examined what is the distribution of established risk variables for teenage motherhood? from where in these distributions do births arise? and how does this distribution/determination of risk vary between studies?

Methods Secondary data analysis of three British longitudinal studies.

Results For all cohorts and variables, the ‘risk’ category was the least frequent. Continuous risk factors were normally distributed. A high rate of teenage motherhood within a risk category often translated into low ‘contribution’ to the overall rate (eg, expectation to leave school at the minimum age among the 1989/1990-born cohort) and vice versa. Most young women had a low probability of teenage motherhood. For any targeting strategy, combining risk factors and a low threshold of predicted probability would be necessary to achieve adequate sensitivity. Assessing between-cohort applicability of findings, the authors find that the numbers of teenage parents is poorly estimated and estimates of the variability and direction of risk may also be inadequate.

Conclusions With reference to a number of established risk factors, there is not a core of easily identifiable multiply disadvantaged girls who go on to constitute the majority of teenage mothers in these studies. While individual risk factors are unlikely to enable targeting, a composite may have some limited potential, albeit with a low threshold for ‘risk’ and with the caveat that evidence from one population may not inform good targeting in another. It is likely that universal approaches will have more impact.