Nomogram to predict severe retinopathy of prematurity in Southeast China

AIM: To define the predictive factors of severe retinopathy of prematurity (ROP) and develop a nomogram for predicting severe ROP in southeast China. METHODS: Totally 554 infants diagnosed with ROP hospitalized in the Second Affiliated Hospital and Yuying Children’s Hospital of Wenzhou Medical University and hospitalized in Taizhou Women and Children’s Hospital were included. Clinical data and 43 candidate predictive factors of ROP infants were collected retrospectively. Logistic regression model was used to identify predictive factors of severe ROP and to propose a nomogram for individual risk prediction, which was compared with WINROP model and Digirop-Birth model. RESULTS: Infants from the Second Affiliated Hospital and Yuying Children’s Hospital of Wenzhou Medical University (n=478) were randomly allocated into training (n=402) and internal validation group (n=76). Infants from Taizhou Women and Children’s Hospital were set as external validation group (n=76). Severe ROP were found in 52 of 402 infants, 12 of 76 infants, and 7 of 76 infants in training group, internal validation group, and external validation group, respectively. Birth weight [odds ratio (OR), 0.997; 95% confidence interval (CI), 0.996-0.999; P<0.001], multiple births (OR, 1.885; 95%CI, 1.013-3.506; P=0.045), and non-invasive ventilation (OR, 0.288; 95%CI, 0.146-0.570; P<0.001) were identified as predictive factors for the prediction of severe ROP, by univariate analysis and multivariate analysis. For predicting severe ROP based on the internal validation group, the areas under receiver operating characteristic curve (AUC) was 78.1 (95%CI, 64.2-92.0) for the nomogram, 32.9 (95%CI, 15.3-50.5) for WINROP model, 70.2 (95%CI, 55.8-84.6) for Digirop-Birth model. In external validation group, AUC of the nomogram was also higher than that of WINROP model and Digirop-Birth model (80.2 versus 51.1 and 63.4). The decision curve analysis of the nomogram demonstrated better clinical efficacy than that of WINROP model and Digirop-Birth model. The calibration curves demonstrated a good consistency between the actual severe ROP incidence and the predicted probability. CONCLUSION: Birth weight, multiple births, and non-invasive ventilation are independent predictors of severe ROP. The nomogram has a good ability to predict severe ROP and performed well on internal validation and external validation in southeast China.