To develop an algorithm that allows advanced identification of infants requiring treatment for retinopathy of prematurity (ROP).

A retrospective observational study was performed at 2 tertiary neonatal units serving a multiethnic population in the UK, using data on 929 infants eligible for ROP screening. The relationships between study variables and the risk of developing ROP requiring treatment were analyzed using multiple logistic regression.

After applying exclusion criteria, data from 589 infants were analyzed; of these, 57 required laser treatment. The proportion of treated infants was 5.9% of those born to black mothers, 9.39% of those born to white mothers, and 12.8% of those born to Asian mothers (P = .047). Multiple logistic regression showed that gestational age, birth weight, maternal ethnicity, and early weight gain were predictors for the development of ROP requiring treatment, with maternal ethnicity having greater predictive power compared with early weight gain. We developed an algorithm for predicting the development of ROP requiring treatment with sensitivity, specificity, and positive and negative predictive values of 100%, 65.7%, 23.8%, and 100%, respectively.

Gestational age, birth weight, early weight gain, and maternal ethnicity are important predictors for the development of ROP requiring treatment. In a multiethnic population, an algorithm to predict development of ROP requiring treatment should include maternal ethnicity. If confirmed through prospective studies, this algorithm could reduce the number of opthalmologic examinations performed for ROP screening.