Related Experiment Video
Updated: Jan 9, 2026
Fetal Echocardiography and Pulsed-wave Doppler Ultrasound in a Rabbit Model of Intrauterine Growth Restriction
Published on: June 29, 2013
Assessing adherence to TRIPOD+AI guidelines in machine learning models for predicting small for gestational age and
Giulia Zamagni1, Camilla Fregona2, Moira Barbieri3
1University of Trieste, Trieste, Italy (Zamagni); Clinical Epidemiology and Public Health Research Unit, Institute for Maternal and Child Health - IRCCS "Burlo Garofolo", Trieste, Italy (Zamagni and Monasta).
Insights
Machine learning models show promise for predicting fetal growth restriction (FGR) and small for gestational age (SGA). However, inconsistent definitions, small sample sizes, and poor reporting limit their reliability and clinical use.
Area of Science:
- Medical Informatics
- Perinatal Medicine
- Machine Learning in Healthcare
Background:
- Fetal growth restriction (FGR) and small for gestational age (SGA) are critical indicators of perinatal health, impacting long-term outcomes.
- Accurate prediction of FGR/SGA is essential for timely intervention and improved neonatal care.
- Machine learning (ML) offers potential for enhancing FGR/SGA prediction using clinical data.
Purpose of the Study:
- To systematically review machine learning (ML) applications for predicting FGR/SGA.
- To evaluate the methodological rigor and reporting quality of existing ML models in this field.
- To identify limitations and areas for improvement in ML-based FGR/SGA prediction.
Main Methods:
- Systematic literature search conducted in MEDLINE and Scopus following PRISMA 2020 guidelines.
- Inclusion of studies using ML for FGR/SGA prediction with clinical variables and reporting AUROC or accuracy.
- Assessment of risk of bias (PROBAST) and adherence to TRIPOD+AI guidelines, including sample size adequacy.
Main Results:
- 20 studies met inclusion criteria from 272 identified; definitions of FGR/SGA were inconsistent.
- Variable adherence to TRIPOD+AI guidelines; no models reported fairness or heterogeneity; only 15% reported calibration.
- Only 30% of studies had adequate sample sizes, suggesting potential overfitting and limited generalizability.
Conclusions:
- ML models hold potential for FGR/SGA prediction but face significant limitations.
- Inconsistent definitions, underpowered studies, and poor reporting of calibration hinder clinical translation.
- Future research requires standardized definitions, robust sample sizes, and comprehensive reporting for reliable ML models.
Objectives:
Fetal growth restriction (FGR) significantly contribute to perinatal morbidity, mortality, and long-term adverse health outcomes. While small for gestational age (SGA) is often used as a proxy for FGR, it does not necessarily indicate pathological growth restriction. Given the increasing interest in machine learning (ML) for predicting FGR/SGA, this study systematically reviews ML applications in this domain, evaluating their methodological rigor and reporting quality, following standardized guidelines.
Data Sources:
The systematic search was conducted in MEDLINE and Scopus on June 21, 2024, following PRISMA 2020 guidelines.
Study Eligibility Criteria:
Eligible studies implemented ML models for FGR/SGA prediction using routinely available clinical variables and reported at least one area under the receiver operating characteristic (AUROC) and/or accuracy. Exclusions included preprints, conference abstracts, systematic reviews, animal studies, and models relying exclusively on biomarkers or genomics, as not part of the clinical practice.
Study Appraisal And Synthesis Methods:
Two independent reviewers screened articles with the help of the Rayyan software. Risk of bias was assessed using the PROBAST checklist. Adherence to the guidelines on the transparent reporting of a multivariable prediction model for individual prognosis or diagnosis+artificial intelligence (TRIPOD+AI) was evaluated across methods, results, and discussion sections using a 4-point Likert scale. Sample size adequacy was assessed for each study, accounting for outcome type, predictors, and outcome prevalence.
Results:
The search identified 272 studies, with 20 meeting the inclusion criteria. Definitions of FGR/SGA were inconsistent, particularly in technical journals. Adherence to TRIPOD+AI guidelines was variable, as no model reported on fairness or heterogeneity across relevant subgroups, and only 15% reported on calibration. Only 30% of studies met the minimum sample size required for ML models, indicating potential overfitting and limited generalizability.
Conclusion:
Despite the potential of ML models in predicting FGR/SGA, key limitations persist, including inconsistent outcome definitions, underpowered models, and suboptimal reporting of calibration and clinical applicability. Future studies should emphasize standardized definitions, robust sample sizes, and comprehensive reporting to enhance model reliability and clinical translation. VIDEO ABSTRACT.
