Enhancing the prediction of disease outcomes using electronic health records and pretrained deep learning models
This work addresses the challenge of accurate disease outcome prediction for healthcare providers, but it is incremental as it builds on existing pretrained models and methods.
The study tackled the problem of predicting patient outcomes from electronic health records by using a pretrained encoder-decoder architecture, achieving improved performance over state-of-the-art models like BERT across outcomes such as intentional self-harm and pancreatic cancer on a dataset of 6.8 million patients.
Question: Can an encoder-decoder architecture pretrained on a large dataset of longitudinal electronic health records improves patient outcome predictions? Findings: In this prognostic study of 6.8 million patients, our denoising sequence-to-sequence prediction model of multiple outcomes outperformed state-of-the-art models scuh pretrained BERT on a broad range of patient outcomes, including intentional self-harm and pancreatic cancer. Meaning: Deep bidirectional and autoregressive representation improves patient outcome prediction.