Advanced Prediction of Recurrent Fragility Fractures Using Large Language Models

Recurrent fragility fractures are a big challenge in managing bone health, especially in older adults and people with osteoporosis or related disorders. The prediction of such fractures depends on the overall understanding of various clinical and lifestyle factors contributing to bone fragility. The...

Full description

Saved in:
Bibliographic Details
Main Authors: Mohammad Alshraideh, Arafat Al-Dhaqm, Ahmad Alshammari, Abedalrahman Alshraideh, Bahaaldeen Alshraideh, Bayan Mohamed Al-Fayoumi, Maged Nasser
Format: Article
Language:English
Published: IEEE 2025-01-01
Series:IEEE Access
Subjects:
Online Access:https://ieeexplore.ieee.org/document/10969766/
Tags: Add Tag
No Tags, Be the first to tag this record!
_version_ 1850141172642611200
author Mohammad Alshraideh
Arafat Al-Dhaqm
Ahmad Alshammari
Abedalrahman Alshraideh
Bahaaldeen Alshraideh
Bayan Mohamed Al-Fayoumi
Maged Nasser
author_facet Mohammad Alshraideh
Arafat Al-Dhaqm
Ahmad Alshammari
Abedalrahman Alshraideh
Bahaaldeen Alshraideh
Bayan Mohamed Al-Fayoumi
Maged Nasser
author_sort Mohammad Alshraideh
collection DOAJ
description Recurrent fragility fractures are a big challenge in managing bone health, especially in older adults and people with osteoporosis or related disorders. The prediction of such fractures depends on the overall understanding of various clinical and lifestyle factors contributing to bone fragility. The goal of this analysis is to carry out advanced predictive analytics related to the problem of recurrent fragility fractures concerning several key features about each patient: age, sex, body mass index, physical activities, smoking status, and several others: T-score, along with biomarkers such as Vitamin D3, calcium levels, and the rest. Indeed, the results show that the features most indicative of recurrent fractures include age, T-score, physical activity, and glucocorticoid usage; greater emphasis has been laid on bone mineral density and lifestyle factors. The detailed feature importance analysis showed that age and T-score have the highest important values for fracture recurrence prediction, followed by physical activity and glucocorticoid treatment. The study finds that the proposed models achieved an accuracy of 91.3% (LLaMA 7GB) and 90.2% (Gamma 7GB), with ROC AUC scores of 93% and 92%, respectively. The study highlights that age, T-score, and glucocorticoid usage were the most significant predictive factors. These findings enable improved clinical decision-making for fracture prevention. This model can facilitate the identification of high-risk patients with recurrent fragility fractures by clinicians, thus allowing for targeted intervention and personalized treatment to reduce the risk of future fractures. The findings contribute to the development of fracture prediction, incorporating a wide array of clinical data that, in turn, will improve patient outcomes and their quality of life.
format Article
id doaj-art-8839b7bb2d6f4ad6a7ecfc4f1a3b2894
institution OA Journals
issn 2169-3536
language English
publishDate 2025-01-01
publisher IEEE
record_format Article
series IEEE Access
spelling doaj-art-8839b7bb2d6f4ad6a7ecfc4f1a3b28942025-08-20T02:29:31ZengIEEEIEEE Access2169-35362025-01-0113715037152010.1109/ACCESS.2025.356225910969766Advanced Prediction of Recurrent Fragility Fractures Using Large Language ModelsMohammad Alshraideh0https://orcid.org/0000-0002-2724-9290Arafat Al-Dhaqm1https://orcid.org/0000-0002-0729-2654Ahmad Alshammari2https://orcid.org/0009-0000-2051-2757Abedalrahman Alshraideh3https://orcid.org/0009-0008-8907-8010Bahaaldeen Alshraideh4https://orcid.org/0009-0008-9293-6426Bayan Mohamed Al-Fayoumi5Maged Nasser6https://orcid.org/0000-0003-3788-5722Artificial Intelligence Department, The University of Jordan, Amman, JordanSchool of Computer Science (SCS), Taylor’s University, Subang Jaya, MalaysiaDepartment of Computer Sciences, Faculty of Computing and Information Technology, Northern Border University, Rafha, Saudi ArabiaEast Midlands Deanery, NHS, Nottingham, England, U.K.Department of Special Surgery, Division of Urology, The University of Jordan, Amman, JordanInformation Technology College, Lusail University, Lusail, QatarComputer and Information Sciences Department, Universiti Teknologi PETRONAS, Seri Iskandar, MalaysiaRecurrent fragility fractures are a big challenge in managing bone health, especially in older adults and people with osteoporosis or related disorders. The prediction of such fractures depends on the overall understanding of various clinical and lifestyle factors contributing to bone fragility. The goal of this analysis is to carry out advanced predictive analytics related to the problem of recurrent fragility fractures concerning several key features about each patient: age, sex, body mass index, physical activities, smoking status, and several others: T-score, along with biomarkers such as Vitamin D3, calcium levels, and the rest. Indeed, the results show that the features most indicative of recurrent fractures include age, T-score, physical activity, and glucocorticoid usage; greater emphasis has been laid on bone mineral density and lifestyle factors. The detailed feature importance analysis showed that age and T-score have the highest important values for fracture recurrence prediction, followed by physical activity and glucocorticoid treatment. The study finds that the proposed models achieved an accuracy of 91.3% (LLaMA 7GB) and 90.2% (Gamma 7GB), with ROC AUC scores of 93% and 92%, respectively. The study highlights that age, T-score, and glucocorticoid usage were the most significant predictive factors. These findings enable improved clinical decision-making for fracture prevention. This model can facilitate the identification of high-risk patients with recurrent fragility fractures by clinicians, thus allowing for targeted intervention and personalized treatment to reduce the risk of future fractures. The findings contribute to the development of fracture prediction, incorporating a wide array of clinical data that, in turn, will improve patient outcomes and their quality of life.https://ieeexplore.ieee.org/document/10969766/Osteoporosisfractureslarge language modelsmachine learningrecurrent fragilitygamma model
spellingShingle Mohammad Alshraideh
Arafat Al-Dhaqm
Ahmad Alshammari
Abedalrahman Alshraideh
Bahaaldeen Alshraideh
Bayan Mohamed Al-Fayoumi
Maged Nasser
Advanced Prediction of Recurrent Fragility Fractures Using Large Language Models
IEEE Access
Osteoporosis
fractures
large language models
machine learning
recurrent fragility
gamma model
title Advanced Prediction of Recurrent Fragility Fractures Using Large Language Models
title_full Advanced Prediction of Recurrent Fragility Fractures Using Large Language Models
title_fullStr Advanced Prediction of Recurrent Fragility Fractures Using Large Language Models
title_full_unstemmed Advanced Prediction of Recurrent Fragility Fractures Using Large Language Models
title_short Advanced Prediction of Recurrent Fragility Fractures Using Large Language Models
title_sort advanced prediction of recurrent fragility fractures using large language models
topic Osteoporosis
fractures
large language models
machine learning
recurrent fragility
gamma model
url https://ieeexplore.ieee.org/document/10969766/
work_keys_str_mv AT mohammadalshraideh advancedpredictionofrecurrentfragilityfracturesusinglargelanguagemodels
AT arafataldhaqm advancedpredictionofrecurrentfragilityfracturesusinglargelanguagemodels
AT ahmadalshammari advancedpredictionofrecurrentfragilityfracturesusinglargelanguagemodels
AT abedalrahmanalshraideh advancedpredictionofrecurrentfragilityfracturesusinglargelanguagemodels
AT bahaaldeenalshraideh advancedpredictionofrecurrentfragilityfracturesusinglargelanguagemodels
AT bayanmohamedalfayoumi advancedpredictionofrecurrentfragilityfracturesusinglargelanguagemodels
AT magednasser advancedpredictionofrecurrentfragilityfracturesusinglargelanguagemodels