Prediction of Transient Hypocalcemia Following Total Thyroidectomy Using Machine Learning Methods

: To apply machine learning (ML) to forecast transient hypocalcemia after total thyroidectomy. : Predicting transient hypocalcemia remains difficult despite advances in surgery and perioperative care. ML can model complex relationships among clinical and laboratory variables to estimate individual risk. : We retrospectively analyzed 571 adults who underwent total thyroidectomy at Çukurova University (January 2015-December 2021). Demographic, clinical, pathological, and laboratory variables that were available before outcome ascertainment (second postoperative week) were evaluated. Multiple ML algorithms were trained and compared using receiver operating characteristic area under the curve (ROC-AUC), F1, sensitivity, precision, and accuracy on a held-out test set. : Transient hypocalcemia occurred in 19.8% of patients. It was more frequent in patients with incidental parathyroidectomy and in those with tracheal or recurrent laryngeal nerve injury (p<0.05). Logistic regression achieved the highest ROC-AUC (0.610), while linear discriminant analysis (LDA) yielded the highest accuracy (0.805). Day-1 calcium and parathyroid hormone were strongly associated with subsequent hypocalcemia. : ML provides a feasible approach to estimate post-thyroidectomy hypocalcemia risk, although discrimination was modest in this single-center cohort. Larger, well-curated, multicenter datasets and external validation are needed.