MMPro-HIP: Multimodal Model for Elderly Hip Fracture Risk Prediction

1,287 elderly clinical records from Beijing Jishuitan Hospital, including 643 patients with hip fractures and 644 controls, were analyzed to develop a progressive fusion model for risk prediction. The model, created by Songyuan Chen, achieved an accuracy of 90.94% and an AUC of 0.9423 on an independent test set. The dataset was last updated on 2026-04 28.

Use Cases

• Train hip fracture risk prediction models based on demographic and clinical variables mentioned in the description.
• Benchmark multimodal fusion techniques for handling missing clinical or imaging data.
• Study the predictive contribution of factors like bone mineral density, age, and sex in elderly fracture risk.
• Develop robust clinical decision-support tools for resource-constrained healthcare settings.

Strengths

• Includes 1,287 clinical records with a balanced case-control split of 643 fracture patients and 644 controls.
• Model performance is quantified with specific metrics: 90.94% accuracy and an AUC of 0.9423.
• Explicitly addresses the common problem of modular missing data in clinical settings.
• Released under a permissive CC-BY-4.0 license.

Limitations

• Row count and column-level documentation are unknown, limiting suitability assessment.
• Data is from a single-center retrospective cohort (Beijing Jishuitan Hospital), which may limit generalizability.
• The primary file format is a DOCX document (893.2 KB), which may not be a structured data file.

Provenance

Source

Beijing Jishuitan Hospital, Capital Medical University.

Collection Method

Retrospective analysis of clinical records.

Time Range

null

Freshness

Last updated 2026-04-28 04:24:46; freshness should be verified.

Geography

Beijing, China.