Chest X-ray(CXR) is one of the most commonly used imaging techniques in clinical diagnosis. However, with the increasing number of examinations and the challenges in diagnostic accuracy, healthcare professionals face a significant workload. Currently, Vision Transformer (ViT) models and their variants have shown excellent performance in medical image classification tasks. However, the computational complexity significantly increases as the number of image patches increases, limiting their application in high-resolution medical images. This paper proposes a Hybrid Attention Convolutional Transformer (HA-ConvFormer), which introduces Attention Augmented Convolution (AA-Conv) to generate feature representations rich in details, reducing input sequence length in the ViT model while retaining key diagnostic information. Experiments on the CheXpert dataset demonstrate that our model outperforms ResNet152, DenseNet121, and the standard ViT-B/16 in classifying five common chest diseases such as Cardiomegaly and Consolidation, while significantly reducing computational complexity compared to ViT-B/16. This study provides a new approach for efficient analysis of high-resolution medical images.