Artificial intelligence for predicting transient hypocalcemia after total thyroidectomy

Area of Science:

• Endocrinology
• Surgical Oncology
• Artificial Intelligence in Medicine

Background:

• Transient hypocalcemia is a frequent complication following total thyroidectomy.
• Accurate early prediction of hypocalcemia is crucial for effective patient management.

Purpose of the Study:

• To develop and validate machine learning (ML) models for early prediction of transient hypocalcemia post-thyroidectomy.
• To create a clinical decision support system leveraging ML for enhanced postoperative care.

Main Methods:

• A comprehensive ML framework was developed using real-world health record data from 7,066 patients.
• Models utilized preoperative, intraoperative, and postoperative data.
• Prospective validation was conducted on 149 newly operated patients.

Main Results:

• The postoperative ML model achieved an AUC of 0.94 and precision of 0.88.
• Confirmed known risk factors like low postoperative parathyroid hormone (PTH) and bilateral neck dissection.
• Identified novel risk factors including prolonged surgery duration, earlier start times, and surgeon workload.

Conclusions:

• ML models, especially with explainable artificial intelligence (XAI), provide a reliable and interpretable tool for hypocalcemia prediction.
• These models can enhance personalized postoperative care and enable timely, risk-based interventions.