A Clinically Aligned Two-Stage Machine Learning Framework for Predicting Hungry Bone Syndrome After Parathyroidectomy

Area of Science:

• Nephrology and Endocrinology
• Artificial Intelligence in Medicine
• Clinical Prediction Modeling

Background:

• Hungry bone syndrome (HBS) is a frequent complication post-parathyroidectomy (PTX) for secondary hyperparathyroidism (SHPT).
• Existing prediction models for HBS have limitations due to small sample sizes and inability to capture complex interactions.
• This study addresses the need for improved HBS prediction in SHPT patients undergoing PTX.

Purpose of the Study:

• To develop and validate a clinically aligned, two-stage machine learning (ML) framework for predicting HBS after PTX.
• To identify key preoperative and intraoperative predictors of HBS.
• To provide a tool for early identification of high-risk patients.

Main Methods:

• Retrospective analysis of 882 patients undergoing PTX for SHPT.
• A two-stage ML framework incorporating preoperative risk scores and intraoperative features.
• Evaluation of ML models using AUROC, calibration metrics, and resampling techniques.

Main Results:

• HBS incidence was 69.9% in the cohort.
• The EasyEnsemble model showed the highest discrimination (AUROC 0.712) and overall performance.
• Key predictors included elevated preoperative alkaline phosphatase, higher intact parathyroid hormone, and lower serum calcium.

Conclusions:

• The developed two-stage ML framework demonstrates acceptable predictive performance for HBS post-PTX.
• This model can aid in the early identification of high-risk patients.
• It supports individualized perioperative management to mitigate HBS and its complications.