Unlike natural images, medical images often have intrinsic characteristics that can be leveraged for neural network learning. For example, images that belong to different stages of a disease may continuously follow a certain progression pattern. In this paper, we propose a novel method that leverages disease progression learning for medical image recognition. In our method, sequences of images ordered by disease stages are learned by a neural network that consists of a shared vision model for feature extraction and a long short-term memory network for the learning of stage sequences. Auxiliary vision outputs are also included to capture stage features that tend to be discrete along the disease progression. Our proposed method is evaluated on a public diabetic retinopathy dataset, and achieves about 3.3% improvement in disease staging accuracy, compared to the baseline method that does not use disease progression learning.