Artificial intelligence for predicting postoperative outcomes and complications in pediatric urology: a scoping review

Background Postoperative outcomes in pediatric urology vary substantially despite generally high procedural success, creating a need for more individualized risk prediction. Artificial intelligence (AI) offers new opportunities to integrate clinical, anatomical, laboratory, imaging, and perioperative data, but the breadth, methodological strength, and clinical readiness of current prediction models remain unclear. Objective To map AI and machine-learning models predicting postoperative complications and outcomes in pediatric urology by characterizing procedures, prediction targets, predictor domains, model types, validation strategies, explainability, and clinical readiness, and to identify priorities for future validation and clinical implementation. Methods Six databases were searched through 20 July 2026. Eligible studies developed, validated, compared, or evaluated an individual-level AI/ML model predicting an outcome after a pediatric urologic procedure or kidney transplantation. Evidence was synthesized narratively by clinical domain, prediction target, predictor domain, model family, and validation level; no meta-analysis was performed. Results Of 1,478 records identified through the database searches, 25 studies (40 prediction tasks, six clinical domains) met eligibility. All were retrospective. Kidney transplantation contributed the most studies (n=8); stone disease contributed the most tasks (n=13). Machine-learning models based on tree ensembles and boosting (e.g., random forest, gradient boosting) were selected in 24/40 tasks (60.0%). Among 24 tasks reporting AUC, values ranged 0.559–0.970 (14/24, 58.3% ≥0.80), but several favorable estimates rested on very few outcome events. Only two studies (8%) performed independent external validation; calibration was reported in four studies (16%) and decision-curve analysis in seven (28%). No study prospectively evaluated a model, embedded it into routine care, or assessed patient benefit or fairness. Conclusion AI-based postoperative prediction in pediatric urology is expanding and targets clinically important outcomes, with several models showing promising discrimination for individualized risk stratification. However, current evidence remains at an early stage, with predominantly retrospective, single-center studies. Multicenter external validation, standardized outcome definitions, calibration reporting, dynamic multimodal modeling, and prospective impact evaluation are needed before these tools can be considered ready for routine pediatric urologic practice.

Authors

Institutions

Publication Details

Journal
Frontiers in Urology
Published
2026-11-20
DOI
https://doi.org/10.3389/fruro.2026.2020104
Primary Topic
Artificial Intelligence in Healthcare and Education
Type
article
Field-Weighted Citation Impact
0.00
Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Artificial intelligence for predicting postoperative outcomes and complications in pediatric urology: a scoping review

Tariq O. Abbas, Rawan AlSaad, Rajat Thomas, Muthana Ghazi Al-Salihi et al.
7 links
Frontiers in Urology
Artificial Intelligence in Healthcare and Education
article

Artificial intelligence for predicting postoperative outcomes and complications in pediatric urology: a scoping review

Tariq O. Abbas, Rawan AlSaad, Rajat Thomas, Muthana Ghazi Al-Salihi, Santiago Vallasciani
article en

Abstract

Background Postoperative outcomes in pediatric urology vary substantially despite generally high procedural success, creating a need for more individualized risk prediction. Artificial intelligence (AI) offers new opportunities to integrate clinical, anatomical, laboratory, imaging, and perioperative data, but the breadth, methodological strength, and clinical readiness of current prediction models remain unclear. Objective To map AI and machine-learning models predicting postoperative complications and outcomes in pediatric urology by characterizing procedures, prediction targets, predictor domains, model types, validation strategies, explainability, and clinical readiness, and to identify priorities for future validation and clinical implementation. Methods Six databases were searched through 20 July 2026. Eligible studies developed, validated, compared, or evaluated an individual-level AI/ML model predicting an outcome after a pediatric urologic procedure or kidney transplantation. Evidence was synthesized narratively by clinical domain, prediction target, predictor domain, model family, and validation level; no meta-analysis was performed. Results Of 1,478 records identified through the database searches, 25 studies (40 prediction tasks, six clinical domains) met eligibility. All were retrospective. Kidney transplantation contributed the most studies (n=8); stone disease contributed the most tasks (n=13). Machine-learning models based on tree ensembles and boosting (e.g., random forest, gradient boosting) were selected in 24/40 tasks (60.0%). Among 24 tasks reporting AUC, values ranged 0.559–0.970 (14/24, 58.3% ≥0.80), but several favorable estimates rested on very few outcome events. Only two studies (8%) performed independent external validation; calibration was reported in four studies (16%) and decision-curve analysis in seven (28%). No study prospectively evaluated a model, embedded it into routine care, or assessed patient benefit or fairness. Conclusion AI-based postoperative prediction in pediatric urology is expanding and targets clinically important outcomes, with several models showing promising discrimination for individualized risk stratification. However, current evidence remains at an early stage, with predominantly retrospective, single-center studies. Multicenter external validation, standardized outcome definitions, calibration reporting, dynamic multimodal modeling, and prospective impact evaluation are needed before these tools can be considered ready for routine pediatric urologic practice.

Frontiers in UrologyVol. 6
Sidra Medical and Research Center (QA), Weill Cornell Medical College in Qatar (QA), Weill Cornell Medicine (US), Qatar University (QA)
Good health and well-being
Artificial Intelligence in Healthcare and Education
AI Navigator

Ask Laika to Summarize, Analyze, and Connect papers live on the map.

Summarize Papers & Methodologies

Extract key findings, datasets, and comparative methods across publications.

Benchmark Rankings & Visual Analytics

Rank top research institutions, authors, funders, topics, and journals by Field-Weighted Citation Impact (FWCI) and paper volume with instant charts.

Connect Distant Disciplines

Bridge topological clusters on the map to find hidden collaborative intersections.

Laika access coming soon