Do Severity Prediction Models for Paediatric Appendicitis Work in Another Hospital? A Bidirectional External Validation and Local Recalibration Study

Background/Objectives: Models predicting complicated appendicitis in children could support selection for non-operative management, but are rarely validated in a second hospital. We examined how severity models transport between hospitals, why they fail and what local recalibration can achieve. Methods: Using two open paediatric cohorts, we developed models in Regensburg (164 operated children with histology-verified appendicitis, 50 complicated) and validated them in Düsseldorf (229 operated children, 49 complicated), and vice versa. Laboratory, clinical and ultrasound models, random forest, XGBoost and two clinical scores were compared. We examined heterogeneity in case mix, predictor measurement, documentation and outcome definition, and local updating with 25–150 patients. Results: Internal AUCs of 0.81–0.82 fell to 0.71–0.74 externally; calibration slopes were 0.52–0.70 and risks were overestimated (observed/expected 0.61–0.85), whereas the reverse direction underestimated risk (1.40–1.62). The hospitals differed in age range, referral and management pathways, documentation practices and the histological definition of complicated appendicitis. A larger same-hospital development sample did not improve calibration. Recalibration with 50 local patients (11 events) brought the median calibration slope close to 1, but estimates were imprecise; precise estimation would require several hundred events. Including non-operatively managed children without histology inflated internal AUC from 0.82 to 0.88. Conclusions: In operated children, discrimination was partly preserved between hospitals, but calibration was not. Because outcome definitions, patient selection and documentation differed, poor transportability cannot be attributed to the models alone; the comparability of clinical data and outcomes should be established before models are transferred or recalibrated.

Authors

Institutions

Publication Details

Journal
Diagnostics
Published
2026-11-17
DOI
https://doi.org/10.3390/diagnostics16223827
Primary Topic
Appendicitis Diagnosis and Management
Type
article
Field-Weighted Citation Impact
0.00

Funders

Controls
|||
ALL TIME
JAN
FEB
MAR
APR
MAY
JUN
JUL
AUG
SEP
OCT
NOV
article

Do Severity Prediction Models for Paediatric Appendicitis Work in Another Hospital? A Bidirectional External Validation and Local Recalibration Study

Amal Khalifa Alkhalifa, Sarah A. Alzakari, Şeyma Yaşar, Abedelmalek Kalefh Tabnjh et al.
5 links
Diagnostics
Appendicitis Diagnosis and Management
article

Do Severity Prediction Models for Paediatric Appendicitis Work in Another Hospital? A Bidirectional External Validation and Local Recalibration Study

Amal Khalifa Alkhalifa, Sarah A. Alzakari, Şeyma Yaşar, Abedelmalek Kalefh Tabnjh, Fatma Hilal Yağın, Fahaid H. Al-Hashem
article en

Abstract

Background/Objectives: Models predicting complicated appendicitis in children could support selection for non-operative management, but are rarely validated in a second hospital. We examined how severity models transport between hospitals, why they fail and what local recalibration can achieve. Methods: Using two open paediatric cohorts, we developed models in Regensburg (164 operated children with histology-verified appendicitis, 50 complicated) and validated them in Düsseldorf (229 operated children, 49 complicated), and vice versa. Laboratory, clinical and ultrasound models, random forest, XGBoost and two clinical scores were compared. We examined heterogeneity in case mix, predictor measurement, documentation and outcome definition, and local updating with 25–150 patients. Results: Internal AUCs of 0.81–0.82 fell to 0.71–0.74 externally; calibration slopes were 0.52–0.70 and risks were overestimated (observed/expected 0.61–0.85), whereas the reverse direction underestimated risk (1.40–1.62). The hospitals differed in age range, referral and management pathways, documentation practices and the histological definition of complicated appendicitis. A larger same-hospital development sample did not improve calibration. Recalibration with 50 local patients (11 events) brought the median calibration slope close to 1, but estimates were imprecise; precise estimation would require several hundred events. Including non-operatively managed children without histology inflated internal AUC from 0.82 to 0.88. Conclusions: In operated children, discrimination was partly preserved between hospitals, but calibration was not. Because outcome definitions, patient selection and documentation differed, poor transportability cannot be attributed to the models alone; the comparability of clinical data and outcomes should be established before models are transferred or recalibrated.

DiagnosticsVol. 16(22)
Princess Nourah bint Abdulrahman University (SA), Balıkesir University (TR), Jordan University of Science and Technology (JO), Inonu University (TR), King Khalid University (SA), Saveetha University (IN), University of Gothenburg (SE)
Princess Nourah Bint Abdulrahman University
Good health and well-being
Appendicitis Diagnosis and Management
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