Fig. 6: The process of Clostridioides difficile infection prediction using tree-based and deep learning models. | npj Digital Medicine

Fig. 6: The process of Clostridioides difficile infection prediction using tree-based and deep learning models.

From: Deep learning-based prediction of Clostridioides difficile infection caused by antibiotics using longitudinal electronic health records

Fig. 6: The process of Clostridioides difficile infection prediction using tree-based and deep learning models.

Timeline variables (vital signs and laboratory tests) were fed into the RNN or attention layer, while patient information variables were fed into the fully connected (dense) layer. The output vectors of those two layers were concatenated and fed into another fully connected layer to predict CDI. For tree-based models, we transformed timeline and patient information data into a one-dimensional vector by concatenating the first and last columns of the timeline and patient information vector.

Back to article page