AI-Enabled Virtual Nursing Assistants

Background: The rapid growth of artificial intelligence (AI) is reshaping the quality of delivering the healthcare services, creating new opportunities to support patients beyond traditional clinical environments. Unfortunately, not all individuals can access medical services equally and efficiently. This research aims to develop a virtual nursing assistant by utilizing a neural network model to provide patients with immediate and effective answers to medical inquiries. Methods: In this study we adopted a generative artificial intelligence (AI) powered by neural network layers. The architecture starts with encoding the input of health enquiry into the embedding layer. We utilized the MedQud dataset from Kaggle website, it includes 47,457 medical question-answer pairs covering questions about treatment, diagnosis and side effects. Results: The proposed model performs well during the training phase. However, during the execution phase, the performance metric shows a higher precision score (40%) compared to the recall and F1-score. Conclusions: The healthcare industry can implement an affordable conversational assistant to help patients with medical inquiries by adopting a virtual nursing assistant that utilizes a sequence model with encoder-decoder architecture. In the future, the model will be enhanced to expand the dataset and incorporate a speech mechanism, enabling users to interact with the system through voice commands.