OBJECTIVE. We sought to identify sociodemographic and familial correlates of injury in children aged 2 to 3 years.
METHODS. The Healthy Steps data set describes 5565 infants who were enrolled at birth in 15 US cities in 1996–1997 and had follow-up until they were 30 to 33 months of age. Data were linked to medical claims reporting children's medically attended office visits by age 30 to 33 months. Each claim was accompanied by a reason for the visit. An analytical sample of 3449 was derived from the children who could be effectively followed up and linked to medical charts. Missing data were imputed by using multiple imputation with chained equations. The analytical sample showed no systematic evidence of sample selection bias. Multivariate logistic regression was used to determine the odds ratios of injury events.
RESULTS. Odds of medically attended injuries were decreased for children who received care from grandparents. Odds were increased for children who lived where fathers did not co-reside or in households where the parents never married. Statistical results were robust to the addition of a variety of covariates such as income, education, age, gender, and race.
CONCLUSIONS. Children are at higher risk for medically attended injury when their parents are unmarried. Having grandparents as caregivers seems to be protective. Household composition seems to play a key role in placing children at risk for medically attended injuries.