Validity and reliability of ROPScore scoring method to predict the severity of retinopathy of prematurity in premature infants

Abstract Purpose: To assess the accuracy and efficacy of ROPScore scoring system an ancillary method to predict the severity of retinopathy of prematurity (ROP) in very low birth weight (VLBW) premature infants. Methods: The medical records of 131 premature babies having a birth weight  1500 gram and gestational age (GA) ≤ 30 weeks were included in this study. The ROPScore was calculated for each baby at six weeks of life using an Excel spreadsheet (Microsoft®). Area under curve (AUC) analysis was used in both any stage of ROP and type-1 (severe) ROP to ascertain the cut-off points for the scoring model. Sensitivity, specificity, positive predictive values (PPV) and negative predictive values (NPV) of the scoring system with the calibrated cut-off points were analyzed. Results: The sensitivity of the ROPScore scoring system was 88.5% ( 95% CI 79-94) and 100% (95% CI 82-100) was for predicting any stage and type-1 retinopathy of prematurity, respectively. The PPV and NPV of the models were 62% and 74.1% for any stage of ROP and those of were 50% and 100% for type-1 ROP, respectively. In ROC analysis, the mean AUCs of ROPScore model was statistically significant compared than BW and GA for predicting type -1 ROP (p < 0.001). Conclusion: This study indicated that ROPScore scoring model with customized cutoff levels might be a useful method for early prediction of premature retinopathy, particularly in type-1 (severe) ROP. In addition, this model may also reduce the number of eye examinations which are essential for detecting the retinopathy of prematurity