Forecasting the future clinical events of a patient through contrastive learning.

This investigation indicates that clinical event forecasting can be improved significantly through contrastive representation learning, especially when the number of positive training instances is small.