303 - Can we optimize ROP Screening by combining risk scores?

Poster Number: 303
Publication Number: 303.2747

Background: Retinopathy of Prematurity (ROP) is a leading cause of childhood blindness. Retinal exams are costly and stressful for premature neonates, yet current screening for ROP lacks specificity for severe disease. Importantly, premature infants with significant lung disease share a higher burden of severe ROP. Here, we leverage early respiratory biomarkers to improve the specificity of ROP screening and reduce the number of infants receiving eye exams while maintaining 100% sensitivity for treatment-requiring ROP (TR-ROP).

Objective: To evaluate a composite metric incorporating the DigiROP-Birth (DRB) score and 2011 Neonatal Research Network 14-Day BPD Outcome Estimator (NRN) for predicting stage 2 ROP and TR-ROP.

Design/Methods: Among 1408 premature infants born at < 30 weeks completed gestation (CGA) or < 1500g birthweight (BW) admitted to the University of Utah NICU between Jan 2010 - Sept 2023, 990 received retinal exams. The primary outcome was TR-ROP and secondary outcome was stage 2 or greater ROP. Retrospectively, we quantified the DRB and NRN scores for each infant. Optimal cutoff thresholds for predicting TR-ROP were selected using a receiver operator characteristic (ROC) analysis. Sensitivity and Specificity were then assessed for all infants meeting either threshold.

Results: Of the 990 infants receiving retinal exams, 363 (36.7%) had Stage 2 or greater ROP, and 68 (6.9%) underwent ROP treatment. Demographics and scoring data are shown in Table 1. ROC Areas Under the Curve (AUC) and 95% CI were 0.867 (0.829-0.906) and 0.845 (0.809-0.881) for DRB and NRN respectively. Optimal cutoff scores were 1.7 and 40 for DRB and NRN with respective sensitivities of 97% and 96%. Composite screening for babies meeting either cutoff would maintain 100% sensitivity for predicting TR-ROP while decreasing the number of infants screened from 990 to 558, a 43% reduction.

Conclusion(s): A composite metric using DigiROP-Birth and NRN BPD Outcome Estimator scores may predict TR-ROP with 100% sensitivity and allow significant reduction in the number of infants requiring screening ROP exams, however validation studies are needed.