Prognostic Models Predicting Mortality in Preterm Infants: Systematic Review and Meta-analysis

Area of Science:

• Neonatal research
• Medical informatics
• Biostatistics

Background:

• Prognostic models are crucial for assessing mortality risk in preterm infants.
• Existing models require rigorous evaluation for clinical utility.

Purpose of the Study:

• To summarize prognostic models for predicting mortality in very preterm infants.
• To assess the quality and applicability of these models.

Main Methods:

• A comprehensive literature search of Medline up to June 2020.
• Inclusion of developed or externally validated prognostic models for mortality in infants <32 weeks gestation or <1500g.
• Data extraction and risk of bias assessment using the Prediction model Risk of Bias Assessment Tool by independent authors.

Main Results:

• 142 models from 35 studies (development) and 112 models from 33 studies (external validation) were included.
• The majority of models exhibited high risk of bias, primarily due to inadequate analysis reporting.
• Internal and external validation were lacking in 41% and 96% of models, respectively. Notable C-statistics were observed for specific scores.

Conclusions:

• A significant proportion of prognostic models for preterm infant mortality suffer from high risk of bias and insufficient validation.
• There is a critical need to shift focus from developing new models to externally validating and adapting existing ones.
• Improving the quality and reliability of prognostic models is essential for better clinical decision-making in neonatal intensive care units.