Back to search

Aristotle University Of Thessaloniki (AUTH)

Development and validation of prognostic models in retinopathy of prematurity: a cohort study

Abstract

dc:description

Retinopathy of prematurity (ROP) remains a leading cause of preventable childhood visual impairment worldwide. Current screening strategies are primarily based on gestational age (GA) and birth weight (BW) criteria and are designed to maximize sensitivity. However, because only a small proportion of screened infants develop treatment-warranted ROP (TW-ROP), many preterm infants undergo repeated ophthalmologic examinations unnecessarily. Optimising screening efficiency while maintaining patient safety is therefore a key challenge.This thesis aimed to evaluate current ROP screening practices in Greece, explore opportunities to optimise screening criteria and examination schedules, systematically review existing prognostic models for sight-threatening ROP, externally validate established prediction tools in a Greek population, and develop a novel individualised prediction model for TW-ROP.The study population comprised consecutive preterm infants undergoing ROP screening at the 2nd Neonatology Department of Aristotle University of Thessaloniki as well as a national cohort of treated infants. Multiple imputation was used when necessary to address missing data and bootstrap resampling was applied as internal validation to correct for optimism. Previously developed models (CO-ROP, G-ROP, G-ROP 180 g, ROPScore) were externally validated using the published cut-offs from the original development studies, without recalibration. TWO-ROP and FIRST-ROP algorithms were validated as originally described. Penalised ridge logistic regression was used to develop the THESS-ROP model incorporating GA, BW, sepsis and hydrocephalus within 4 weeks postnatal age (PNA). The systematic review was conducted according to PRISMA and TRIPOD-SRMA guidance, and methodological quality was assessed using PROBAST+AI. Current screening criteria demonstrated high sensitivity but resulted in examining many infants who never developed severe disease. Local analyses suggested that more restrictive GA and BW thresholds could substantially reduce screening burden, although national data indicated that additional clinical risk factors are required to avoid missed cases. External validation of the FIRST-ROP and TWO-ROP algorithms showed potential to reduce examinations in selected medium- and low-risk infants without missed cases of TW-ROP, although careful consideration of local epidemiology and clinical characteristics is warranted. The systematic review identified 35 prognostic models for sight-threatening ROP across more than 120 studies; however, most exhibited methodological limitations, incomplete reporting of performance measures, and high or unclear risk of bias. Among externally validated models, DIGIROP-v1.0, CO-ROP, G-ROP, modified G-ROP 180 g, and ROPScore showed high sensitivity, although specificity varied considerably across models. The THESS-ROP model, based on readily available clinical predictors and providing early risk estimation at 4 weeks PNA, demonstrated excellent discrimination for TW-ROP (optimism-corrected AUC 0.935; 95% CI 0.905–0.965), with 98% sensitivity, and 58% specificity, while correctly identifying all infants who developed type 1 ROP or ultimately received treatment.Overall, these findings support more individualised, risk-adapted approaches to ROP screening. Although no model can currently replace clinical assessment, rigorously developed and validated prediction tools may substantially reduce unnecessary examinations while maintaining patient safety. Further prospective multicentre validation is required before routine clinical implementation.

Degree

thesis:*
Grantor dc:publisher
Aristotle University Of Thessaloniki (AUTH)
Year dc:date
2026

Author and committee

dc:creator, dc:contributor.*
Authors dc:creator
  • Moutzouri, Stella
  • Μουτζούρη, Στέλλα

Subjects

dc:subject × 16

Rights

Language dc:language
eng

Identifiers

dc:identifier.*
Identifier
10.12681/eadd/62852

Chain of custody

source
Harvested from
Greek National Archive of PhD Theses
Base URL
phdtheses.ekt.gr/eadd_oai/request
Last updated
2026-11-20
Source record
OAI-PMH GetRecord
citation

Moutzouri, Stella; Μουτζούρη, Στέλλα. Development and validation of prognostic models in retinopathy of prematurity: a cohort study. Aristotle University Of Thessaloniki (AUTH), 2026. http://hdl.handle.net/10442/hedi/62852