Predictors for the development of referral-warranted retinopathy of prematurity in the telemedicine approaches to evaluating acute-phase retinopathy of prematurity (e-ROP) study

Abstract

Importance: Detection of treatment-requiring retinopathy of prematurity (ROP) involves serial eye examinations. An ROP prediction model using predictive factors could identify high-risk infants and reduce required eye examinations.

Objective: To determine predictive factors for the development of referral-warranted (RW) ROP.

Design, setting, and participants: This multicenter observational cohort study included secondary analysis of data from the Telemedicine Approaches to Evaluating Acute-Phase Retinopathy of Prematurity Study. Infants included in the study had a birth weight (BW) of less than 1251 g.

Exposures: Serial ROP examinations of premature infants who had 2 or more ROP examinations.

Main outcomes and measures: Incidence of RW-ROP (defined as the presence of plus disease, zone I ROP, or ROP stage 3 or greater in either eye) and associations with predictive factors.

Results: Among 979 infants without RW-ROP at first study-related eye examination (median postmenstrual age, 33 weeks; range, 29-40 weeks) who underwent at least 2 eye examinations, 149 (15.2%) developed RW-ROP. In a multivariate model, significant predictors for RW-ROP were male sex (odds ratio [OR], 1.80; 95% CI, 1.13-2.86 vs female), nonblack race (OR, 2.76; 95% CI, 1.50-5.08 for white vs black race and OR, 4.81; 95% CI, 2.19-10.6 for other vs black race), low BW (OR, 5.16; 95% CI, 1.12-7.20 for ≤500 g vs >1100 g), younger gestational age (OR, 9.79; 95% CI, 3.49-27.5 for ≤24 weeks vs ≥28 weeks), number of quadrants with preplus disease (OR, 7.12; 95% CI, 2.53-20.1 for 1-2 quadrants and OR, 18.4; 95% CI, 4.28-79.4 for 3-4 quadrants vs no preplus disease), stage 2 ROP (OR, 4.13; 95% CI, 2.13-8.00 vs no ROP), the presence of retinal hemorrhage (OR, 4.36; 95% CI, 1.57-12.1 vs absence), the need for respiratory support (OR, 4.99; 95% CI, 1.89-13.2 for the need for controlled mechanical ventilator; OR, 11.0; 95% CI, 2.26-53.8 for the need for high-frequency oscillatory ventilation vs no respiratory support), and slow weight gain (OR, 2.44; 95% CI, 1.22-4.89 for weight gain ≤12 g/d vs >18 g/d). These characteristics predicted the development of RW-ROP significantly better than BW and gestational age (area under receiver operating characteristic curve, 0.88 vs 0.78; P < .001).

Conclusions and relevance: When controlling for very low BW and prematurity, the presence of preplus disease, stage 2 ROP, retinal hemorrhage, and the need for ventilation at time of first study-related eye examination were strong independent predictors for RW-ROP. These predictors may help identify infants in need of timely eye examinations.

Figure 1. Receiver Operating Characteristic Curves From the Prediction Models Using Various Combinations of Predictors for Referral-Warranted Retinopathy of Prematurity

The receiver operating characteristic curves from referral-warranted retinopathy of prematurity prediction models using various combinations of predictors including demographic predictors (birth weight [BW], gestational age [GA], sex, and race/ethnicity), ocular predictors (preplus disease, retinopathy of prematurity stage, and retinal hemorrhage), respiratory status, and postnatal weight gain. AUC indicates area under receiver operating characteristic curve.

Figure 2. Predicted Probability of Referral-Warranted (RW) Retinopathy of Prematurity (ROP) From the Final Multivariate Prediction Model

The predicted probability of RW-ROP for each infant with RW-ROP (orange dots) and for infants without RW-ROP (blue dots) was shown. Using a cut point of 0.05 (dashed line), the sensitivity is 96.0% and specificity is 52.7% for predicting RW-ROP.