BACKGROUND: Pregnancy is a time of optimal motivation for many women to make positive behavioural changes. We aim to describe pregnant women with similar patterns of self-reported health behaviours and examine associations with birth outcomes. METHODS: We examined the clustering of multiple health behaviours during pregnancy in the Born in Bradford cohort, including smoking physical inactivity, vitamin d supplementation and exposure to second-hand smoke. Latent class analysis was used to identify groups of individuals with similar patterns of health behaviours separately for White British (WB) and Pakistani mothers. Multinomial regression was then used to examine the association between group membership and birth outcomes, which included preterm birth and mean birthweight. RESULTS: For WB mothers, offspring of those in the 'Unhealthiest' group had lower mean birthweight than those in the 'Mostly healthy but inactive' class, although no association was observed for preterm birth. For Pakistani mothers, group membership was not associated with birthweight differences, although the odds of preterm birth was higher in 'Inactive smokers' compared to the 'Mostly healthy but inactive' group. CONCLUSIONS: The use of latent class methods provides important information about the clustering of health behaviours which can be used to target population segments requiring behaviour change interventions considering multiple risk factors. Given the dominant negative association of smoking with the birth outcomes investigated, latent class groupings of other health behaviours may not confer additional risk information for these outcomes.