Abstract
BACKGROUND/AIMS: We assessed the associations between retinopathy of prematurity (ROP) and continuous measurements of oxygen saturation (SpO(2)), and developed a risk prediction model for severe ROP using birth data and SpO(2) data. METHODS: This retrospective study included infants who were born before 30 weeks of gestation between August 2009 and January 2019 and who were screened for ROP at a single hospital in Japan. We extracted data on birth weight (BW), birth length, gestational age (GA) and minute-by-minute SpO(2) during the first 20 days from the medical records. We defined four SpO(2) variables using sequential measurements. Multivariate logistic regression was used to develop a model that combined birth data and SpO(2) data to predict treatment-requiring ROP (TR-ROP). The model's performance was evaluated using the area under the receiver operating characteristic curve (AUC). RESULTS: Among 350 infants, 83 (23.7%) required ROP treatment. The SpO(2) variables in infants with TR-ROP differed significantly from those with non-TR-ROP. The average SpO(2) and high SpO(2) showed strong associations with GA (r=0.73 and r=0.70, respectively). The model incorporating birth data and the four SpO(2) variables demonstrated good discriminative ability (AUC=0.83), but it did not outperform the model incorporating BW and GA (AUC=0.82). CONCLUSION: Data obtained by continuous SpO(2) monitoring demonstrated valuable associations with severe ROP, as well as with GA. Differences in the distribution of average SpO(2) and high SpO(2) between infants with TR-ROP and non-TR-ROP could be used to establish efficient cut-off values for risk determination.