Stress is a significant concern in the work-place, as it contributes to 80% of workplace injuries, and medical professionals are particularly susceptible, especially during emergencies such as the Covid-19 outbreak. Nurses play a crucial role in providing care in hospitals. Stress prediction for nurses in a hospital environment is essential to work optimally without suffering from the chronic effects of stress. This paper aims to investigate the application of body sensors and machine learning algorithms in monitoring physiological indicators to detect stress levels in nurses. It uses a multimodal sensor dataset of Blood Volume Pulse (BVP), Electrodermal Activity (EDA), and Inter-Beat Interval (IBI) signals from 15 nurses. The study trains and evaluates several classification models separately for each physiological signal and the multimodal data. Performance evaluation reveals that the decision tree outperforms other models with a weighted F1 score of 0.99 for both 2-level and 3-level multimodal stress classification. Furthermore, the IBI-derived Heart Rate Variability (HRV) features is more promising than other physiological signals, with a weighted F1 score of 0.84 and 0.79 for 2-level and 3-level stress classification, respectively. The findings show that multimodal data can help reliably detect stress for nurses and ensure stress-free working in hospital environments to reduce human errors.