Predicting health-related quality of life change using natural language processing in thyroid cancer

•A natural language processing program based on deep learning methods demonstrated an area-under-curve (AUC) of 76.3% in predicting trajectory of health-related quality of life metrics in thyroid cancer patients undergoing surgery, compared to around 60% using classical machine learning models.•Changing the data representation by pairing sequences from different time points (preoperative; 2 weeks, 6 weeks, 6 months, and 1 year postoperative) improved the performance of the model above.•Augmentation of the data using GPT-2, an artificial intelligence which generates representative text sequences, also improved classification performance.