Retinopathy of prematurity (ROP) is a fibrovascular proliferative disease of the immature retina and an important cause of visual impairment and blindness in premature infants. The results of treating severe (threshold) ROP are not always successful. There is growing interest in preventing severe ROP, as is apparent from proposals to study earlier interventions in infants deemed to be at risk for threshold ROP. Accurate earlier identification of infants at greatest risk for severe ROP is essential to permit aggressive early intervention trials, as well as to avoid exposing lower-risk infants unnecessarily to experimental or clinical intervention. A series of ROP risk models were developed based on data from the Multicenter Trial of Cryotherapy for Retinopathy of Prematurity (CRYO ROP), but these reflect infants and care practices of 1986-1987. Development of ROP risk models on contemporary premature infants which include new predictive factors should improve earlier identification of infants at risk for severe ROP. We propose to test the hypothesis that a contemporary ROP risk model which incorporates measures of illness severity, sequential scoring of ophthalmic findings, gestational age, and predictive demographic variables will improve estimation of an individual's risk of severe ROP compared to the CRYO ROP risk model. This will be accomplished through secondary analyses of a newly merged database (SNAP ROP database) from two existing research data sets which are the result of support by the NEI (ROP database) and AHCPR (illness severity SNAP database). The integration of these two data sets into a singularly rich, research quality database (SNAP ROP) presents an extraordinary opportunity to identify new physiologic-based, illness severity determinants of ROP and to develop contemporary ROP risk models for predicting severe (threshold) ROP. We will validate the illness severity ROP risk models (SNAP ROP risk models) on an independent neonatal population similar to the derivation cohort, and compare the performance of the SNAP ROP risk model with the CRYO ROP risk model for predicting threshold ROP (using CRYO ROP threshold definition) on a contemporary, neonatal population.
