Colorado retinopathy of prematurity model: a multi-institutional validation study

• Published in: Journal of AAPOS Vol. 20; no. 3; pp. 220 - 225
• Main Authors: Cao, Jennifer H., MD, Wagner, Brandie D., PhD, Cerda, Ashlee, MPH, McCourt, Emily A., MD, Palestine, Alan, MD, Enzenauer, Robert W., MD, MBA, MPH, Braverman, Rebecca S., MD, Wong, Ryan K., MD, Tsui, Irena, MD, Gore, Charlotte, MD, Robbins, Shira L., MD, Puente, Michael A., MD, MS, Kauffman, Levi, MD, Kong, Lingkun, MD, PhD, Morrison, David G., MD, Lynch, Anne M., MD, MSPH
• Format: Journal Article
• Language: English
• Published: United States Elsevier Inc 01.06.2016
• Subjects: Algorithms; Birth Weight; Cohort Studies; Colorado; Diagnostic Techniques, Ophthalmological; Female; Gestational Age; Humans; Infant; Infant, Extremely Low Birth Weight; Infant, Newborn; Infant, Very Low Birth Weight; Male; Models, Statistical; Neonatal Screening - methods; Ophthalmology; Retinopathy of Prematurity - diagnosis; Retrospective Studies; Risk Factors; Sensitivity and Specificity; Weight Gain
• ISSN: 1091-8531; 1528-3933
• DOI: 10.1016/j.jaapos.2016.01.017

Abstract Purpose The Colorado retinopathy of prematurity (ROP) prediction model (CO-ROP), developed using a cohort of infants from Colorado, calls for ROP examination of infants meeting all of the following criteria: gestational age of ≤30 weeks, birth weight of ≤1500 g, and a net weight gain of ≤650 g between birth and 4 weeks of age. The purpose of this study was to perform an external validation to assess the sensitivity and specificity of the CO-ROP model in a larger cohort of babies screened for ROP from four academic institutions in the United States. Methods The medical records of neonates screened for ROP according current national guidelines was conducted at 4 US academic centers were retrospectively reviewed. Sensitivity, specificity, and respective 95% confidence intervals in detecting ROP using CO-ROP were calculated for type 1, type 2, and any grade of ROP. Results A total of 858 cases were included. The CO-ROP algorithm had a sensitivity of 98.1% (95% CI, 93.3%-99.8%) for type 1 ROP, 95.6% (95% CI 78.0-99.9%) for type 2 ROP, and 95.0% (95% CI, 93.1-97.4%) for all grades of ROP. The CO-ROP model would have reduced the total number of infants screened by 23.9% compared to current 2013 screening guidelines. Conclusions CO-ROP demonstrated high sensitivity in predicting ROP and would have greatly reduced the number of infants needing examination.