Background Infants and toddlers with developmental difficulties represent a heterogeneous group who often receives early intervention (EI). Notable population heterogeneity exists and complicates unmet need and effectiveness research. However, a mix of relatively homogeneous clinically policy relevant ‘subgroups’ may create the apparent heterogeneity. To date, methodological challenges have impeded identifying these potential groups and their policy‐relevance. Methods From the 2005–2006 National Survey of Children with Special Health Care Needs, we derived a sample (n = 965) of infants and toddlers with parent‐reported developmental difficulties. We used latent class analysis (LCA) to identify subgroups of developmental vulnerability based upon functional, social and biological characteristics that would make children eligible for EI. Mixture modelling estimated the likelihood of each subgroup receiving parent‐reported EI, controlling for race/ethnicity, child's age, and state of residence. Results LCA identified four distinct subgroups of developmental vulnerability: developmental disability (Group 1), mild developmental delay (Group 2), socially at risk with behaviour problems (Group 3), and socially at risk with functional vision difficulties (Group 4). Black, non‐Hispanic children are significantly more likely than their white counterparts to be in Group 3 (β = 1.52, P = 0.001) or group 4 (β = 1.83, P < 0.001). Compared with children with a mild developmental delay (Group 2), children in group 1 (β = −0.61, P < 0.001), group 3 (β = −0.47, P = 0.001) and group 4 (β = −0.38, P = 0.009) are significantly less likely to receive EI. Conclusions Racial and ethnic differences exist with regard to membership in developmental vulnerability subgroups. Observed inconsistencies in access to EI suggest the need for improved surveillance, referral and outreach.