Independent factors analysis and prediction model development of treatment-requiring retinopathy of prematurity
10.3760/cma.j.cn511434-20240108-00038
- VernacularTitle:需要治疗的早产儿视网膜病变独立预测因素分析与列线图预测模型构建
- Author:
Yuling XU
1
;
Wei SUN
;
Xiayin ZHANG
;
Jing LI
;
Honghua YU
;
Qiaowei WU
Author Information
1. 南方医科大学广东省眼科研究所 广东省人民医院（广东省医学科学院）眼科，广州 510080
- Keywords:
Premature infant;
Retinopathy of prematurity;
Prediction factors;
Prediction model;
Nomogram
- From:
Chinese Journal of Ocular Fundus Diseases
2024;40(10):750-757
- CountryChina
- Language:Chinese
-
Abstract:
Objective:To analyze independent factors for treatment-requiring retinopathy of prematurity (TR-ROP) and establish a predictive nomogram model for TR-ROP.Method:A retrospective cohort study. A total of 6 998 preterm infants who were born at Guangdong Women's and Children's Hospital between January 1, 2012 and March 31, 2022 and were screened for retinopathy of prematurity (ROP) were included in the study. TR-ROP was defined as type 1 ROP and aggressive ROP; 22 independent factors including general information, maternal perinatal conditions, interventions and neonatal diseases related to ROP were collected. The infants were divided at the level at an 8:2 ratio according to clinical experience, with 5 598 in the training cohort and 1 400 in the validation cohort. t test was used for comparison of quantitative data and χ 2 test was used for comparison of counting data between groups. Multivariate logistic regression analysis was carried out for the indicators with differences in the univariate analysis. The visualized regression analysis results of R software were used to obtain the histogram. The accuracy of the nomogram was verified by C-index and receiver operating characteristic curve (ROC curve). Results:Among the 6 998 children tested, 4 069 were males and 2 920 were females. Gestational age was (33.69±3.19) weeks; birth weight was (2 090±660) g. There were 376 cases of TR-ROP (5.4%, 376/6 998). The results of multivariate logistic regression analysis showed that gestational age [odds ratio ( OR) =0.63, 95% confidence interval ( CI) 0.47-0.85, P=0.002], intrauterine distress ( OR=0.30, 95% CI 0.10-0.99, P=0.048), bronchopulmonary dysplasia ( OR=0.23, 95% CI 0.09-0.60, P=0.003), hypoxic-ischemic encephalopathy ( OR=5.40, 95% CI 1.45-20.10, P=0.012), blood transfusion history ( OR=4.05, 95% CI 1.50-10.95, P=0.006) were the independent influencing factors of TR-ROP. Based on this and combined with birth weight, a nomogram prediction model was established. The C-index of the training set and validation set were 0.940 and 0.885, respectively, and the area under ROC curve were 0.945 (95% CI 0.930-0.961) and 0.931 (95% CI 0.876-0.986), respectively. The sensitivity and specificity were 86.2%, 94.0% and 83.2%, 93.3%, respectively. Conclusions:Gestational age, intrauterine distress, bronchopulmonary dysplasia, hypoxic-ischemic encephalopathy and blood transfusion history are the independent factors influencing the occurrence of TR-ROP. The TR-ROP nomogram prediction model based on independent influencing factors has high sensitivity and specificity.
