Risk factor ranking of retinopathy of prematurity using classification and regression tree analysis

Area of Science:

• Neonatal Ophthalmology
• Pediatric Ophthalmology
• Computational Medicine

Background:

• Retinopathy of prematurity (ROP) is a significant cause of vision impairment in premature infants.
• Established risk factors like gestational age and birth weight may have evolving thresholds due to advancements in neonatal care.

Purpose of the Study:

• To apply classification and regression tree (CART) analysis to a contemporary dataset.
• To predict the likelihood of ROP development, treatment need, and plus disease.
• To identify updated critical thresholds for ROP in a modern U.S. preterm population.

Main Methods:

• A cross-sectional study of 1296 premature infants undergoing ROP screening.
• Utilized CART analysis with recursive partitioning for predictive modeling.
• Data randomized for training (80%) and testing (20%), with 10-fold cross-validation for generalizability.

Main Results:

• Identified critical thresholds for ROP development: gestational age <27.2 weeks and birth weight <775 g.
• Predictive model for ROP incidence achieved a mean absolute error of 0.25.
• ROP progression models indicated age <25 weeks/weight <569 g for treatment need, and age <24 weeks/non-Black maternal race for plus disease.

Conclusions:

• The CART analysis provides current clinical thresholds for ROP development and progression.
• These updated thresholds are relevant for a modern U.S. preterm infant population.
• Prospective validation in diverse populations is recommended to enhance model utility.