Accurate fracture classification using X-ray images is crucial for the timely diagnosis and treatment of potential fractures, and yet it remains a labor-intensive and error-prone process for radiologists, especially in busy clinical environments or under-resourced medical facilities. Traditional convolutional neural networks (CNNs) show promise in automating this task, but they often require large amounts of data to perform well and struggle with spatial hierarchies. This study presents a hybrid deep learning model that combines capsule networks (CapsNet) with conventional CNN architectures to improve the classification of pediatric wrist fractures in X-ray images. All tested hybrid CNN architectures show notable improvement in accuracy, the best improvement being 0.85% compared to the baseline CNNs. The proposed hybrid architecture addresses the limitations of standard CNNs by refining feature representations and improves upon regular capsule networks by leveraging the features extracted by the CNN backbones. These enhancements have potential for computer-aided diagnostic tools, helping radiologists provide more accurate and efficient assessments, particularly in fast-paced or resource-limited clinical settings.