Data-Driven Strategies for Carbimazole Titration: Exploring Machine Learning Solutions in Hyperthyroidism Control.

Area of Science:

• Endocrinology
• Medical Informatics
• Machine Learning

Background:

• University Hospitals Dorset manages over 1000 thyroid patients annually, primarily those with autoimmune hyperthyroidism requiring carbimazole titration.
• Current manual dose adjustments by healthcare professionals (HCPs) based on thyroid function tests are time-consuming and cause treatment delays.

Purpose of the Study:

• To develop and evaluate a machine learning (ML) model for automated carbimazole dosing in hyperthyroidism.
• To enhance the efficiency and timeliness of thyroid disorder treatment and reduce HCP workload.

Main Methods:

• Data from 353 hyperthyroidism patients were anonymized and used to train and test various ML classification algorithms.
• An iterative model and feature selection process was employed, with performance evaluated using weighted F1 and Brier scores.

Main Results:

• A random forest (RF) model achieved the best performance, with an F1 score of 0.751 and a Brier score of 0.38.
• The selected model demonstrated a balance between prediction accuracy and confidence.

Conclusions:

• The study suggests that deploying this ML model in practice is expected to improve the accuracy of carbimazole dosing.
• Further assessment indicated potential for error reduction over time with practical application.