Development and Validation of a Machine Learning-Based Model for Predicting the Risk of Unplanned 30-Day Rehospitalization in Patients with Inflammatory Bowel Disease

Hong Dai, Yao Zeng, Lu Cao

Abstract


To develop and validate a machine learning-based predictive model for the risk of unplanned readmission within 30 days among patients with inflammatory bowel disease (IBD), thereby providing a quantitative tool to aid in the early clinical identification of high-risk patients and the optimization of post-discharge management strategies, a multicenter retrospective study design was employed to collect clinical data on inpatients with IBD from three Grade A Level 3 hospitals (the top-tier in mainland China) in Deyang City and Chengdu City of Sichuan Province between October 2021 and October 2024. Using 30-day unplanned readmission as the outcome variable, we employed LASSO regression with five-fold cross-validation to select the optimal predictors and constructed three machine learning prediction models: Support Vector Machine (SVM), XGBoost, and LightGBM. The dataset was randomly split into a training set and a validation set in a 7:3 ratio, and internal validation was performed using the Bootstrap method (2,000 iterations) to compare the accuracy, sensitivity, specificity, and area under the ROC curve (AUC) of the three models. Results demonstrated that a total of 560 IBD patients were enrolled, with 78 cases (13.9%) experiencing unplanned readmission within 30 days. LASSO regression identified seven predictive factors: daily bowel movement frequency, C-reactive protein, albumin, Barthel Index, number of medications prescribed at discharge, regular follow-up, and length of hospital stay. Among the three models, XGBoost performed the best, with an AUC of 0.842 (95% CI: 0.781–0.903) on the validation set, an accuracy of 80.4%, a sensitivity of 76.9%, and a specificity of 81.5%; LightGBM ranked second (AUC = 0.818), while SVM performed the worst (AUC = 0.763). Bootstrap internal validation demonstrated good stability of the XGBoost model (AUC = 0.836, 95% CI: 0.775–0.897). In conclusion, the XGBoost-based risk prediction model for 30-day unplanned readmission in IBD patients exhibits good predictive performance and stability, and could serve as a reference for clinical nursing risk assessment and the formulation of post-discharge follow-up strategies.


Full Text:

PDF


DOI: https://doi.org/10.5296/jbls.v18i1.24163

Refbacks

  • There are currently no refbacks.


Copyright (c) 2026 Hong Dai, Yao Zeng, Lu Cao

Creative Commons License
This work is licensed under a Creative Commons Attribution 4.0 International License.

Journal of Biology and Life Science  ISSN 2157-6076

Copyright © Macrothink Institute

To make sure that you can receive messages from us, please add the 'macrothink.org' domain to your e-mail 'safe list'. If you do not receive e-mail in your 'inbox', check your 'bulk mail' or 'junk mail' folders.

If you have questions, please contact jbls@macrothink.org.

------------------------------------------------------------------------------------------------------------------------------------------------------