Personalized injury risk assessment for children with special needs

Objective: Personalized injury risk assessment would enhance injury prevention interventions targeting children. Our goal was to develop a basic model for assessing risk of injury in children (ages 5-17) with disabilities.

Methods: We used the 2000-2003 National Health Interview Survey (NHIS) data to build the model. Potential injury risk variables were first fitted into univariate logistic regression models to evaluate medically treated injuries during the three months prior to the interview. A multivariate logistic regression model was fitted to obtain odds ratios (ORs) for injuries at each level of the significant variables. These ORs were then multiplied by 10 and rounded to create individual weighting factors. An injury risk score was calculated for each child based on the significant variables and the associated weighting factors. To test the model, median injury scores were plotted against observed injury incidence for children who participated in the 2004-2005 NHIS.

Results: The injury risk assessment model was built using data from 79,378 children and was tested with data from 13,580 children. Nine variables were determined to be significant for predicting injury: sex, age, U.S. citizenship status, household family size, highest level of education in the household family, yearly household income, presence or absence of health insurance, presence or absence of depression, and disability status. Median injury scores were highly correlated with the observed injury rates with a correlation coefficient of 0.924.

Conclusions: Our preliminary results suggest that personalized injury risk assessments are a feasible tool for assessing risk of injury among children with disabilities.